Merge pull request #2826 from zhayujie/feat-multi-model

feat: multi-provider model console

agent/memory/conversation_store.py

@@ -44,6 +44,7 @@ CREATE TABLE IF NOT EXISTS messages (
     role         TEXT    NOT NULL,
     content      TEXT    NOT NULL,
     created_at   INTEGER NOT NULL,
+    extras       TEXT    NOT NULL DEFAULT '',
     UNIQUE (session_id, seq)
 );
 
@@ -67,6 +68,12 @@ _MIGRATION_ADD_CONTEXT_START_SEQ = """
 ALTER TABLE sessions ADD COLUMN context_start_seq INTEGER NOT NULL DEFAULT 0;
 """
 
+# Generic JSON sidecar for per-message attachments (TTS audio URL, future use).
+# Always optional — readers must tolerate missing column / empty / invalid JSON.
+_MIGRATION_ADD_MSG_EXTRAS = """
+ALTER TABLE messages ADD COLUMN extras TEXT NOT NULL DEFAULT '';
+"""
+
 DEFAULT_MAX_AGE_DAYS: int = 30
 
 
@@ -169,20 +176,26 @@ def _group_into_display_turns(
     cur_rest: List[tuple] = []
     started = False
 
-    for role, raw_content, created_at in rows:
+    for role, raw_content, created_at, raw_extras in rows:
         try:
             content = json.loads(raw_content)
         except Exception:
             content = raw_content
+        try:
+            extras = json.loads(raw_extras) if raw_extras else {}
+            if not isinstance(extras, dict):
+                extras = {}
+        except Exception:
+            extras = {}
 
         if role == "user" and _is_visible_user_message(content):
             if started:
                 groups.append((cur_user, cur_rest))
-            cur_user = (content, created_at)
+            cur_user = (content, created_at, extras)
             cur_rest = []
             started = True
         else:
-            cur_rest.append((role, content, created_at))
+            cur_rest.append((role, content, created_at, extras))
 
     if started:
         groups.append((cur_user, cur_rest))
@@ -195,7 +208,7 @@ def _group_into_display_turns(
     for user_row, rest in groups:
         # User turn
         if user_row:
-            content, created_at = user_row
+            content, created_at, _u_extras = user_row
             text = _extract_display_text(content)
             if text:
                 turns.append({"role": "user", "content": text, "created_at": created_at})
@@ -206,8 +219,11 @@ def _group_into_display_turns(
         tool_results: Dict[str, str] = {}
         final_text = ""
         final_ts: Optional[int] = None
+        merged_extras: Dict[str, Any] = {}
 
-        for role, content, created_at in rest:
+        for role, content, created_at, extras in rest:
+            if role == "assistant" and isinstance(extras, dict):
+                merged_extras.update(extras)
             if role == "user":
                 tool_results.update(_extract_tool_results(content))
             elif role == "assistant":
@@ -256,6 +272,8 @@ def _group_into_display_turns(
                 "steps": steps,
                 "created_at": final_ts or (user_row[1] if user_row else 0),
             }
+            if merged_extras:
+                turn["extras"] = merged_extras
             turns.append(turn)
 
     return turns
@@ -411,13 +429,15 @@ class ConversationStore:
                         content = json.dumps(
                             msg.get("content", ""), ensure_ascii=False
                         )
+                        extras_obj = msg.get("extras") or {}
+                        extras = json.dumps(extras_obj, ensure_ascii=False) if extras_obj else ""
                         conn.execute(
                             """
                             INSERT OR IGNORE INTO messages
-                                (session_id, seq, role, content, created_at)
-                            VALUES (?, ?, ?, ?, ?)
+                                (session_id, seq, role, content, created_at, extras)
+                            VALUES (?, ?, ?, ?, ?, ?)
                             """,
-                            (session_id, next_seq, role, content, now),
+                            (session_id, next_seq, role, content, now, extras),
                         )
                         next_seq += 1
 
@@ -651,6 +671,55 @@ class ConversationStore:
             logger.info(f"[ConversationStore] Pruned {deleted} expired sessions")
         return deleted
 
+    def attach_extras_to_last_assistant(
+        self,
+        session_id: str,
+        extras: Dict[str, Any],
+    ) -> Optional[int]:
+        """
+        Merge ``extras`` into the latest assistant message of a session.
+
+        Used by post-processing (e.g. TTS) that needs to annotate an already
+        persisted bot reply with attachments such as audio URLs.
+
+        Returns the message seq that was updated, or ``None`` if no assistant
+        message exists or the update could not be applied.
+        """
+        if not extras:
+            return None
+        with self._lock:
+            conn = self._connect()
+            try:
+                row = conn.execute(
+                    """
+                    SELECT seq, extras FROM messages
+                    WHERE session_id = ? AND role = 'assistant'
+                    ORDER BY seq DESC LIMIT 1
+                    """,
+                    (session_id,),
+                ).fetchone()
+                if not row:
+                    return None
+                seq, raw = row
+                try:
+                    cur = json.loads(raw) if raw else {}
+                    if not isinstance(cur, dict):
+                        cur = {}
+                except Exception:
+                    cur = {}
+                cur.update(extras)
+                conn.execute(
+                    "UPDATE messages SET extras = ? WHERE session_id = ? AND seq = ?",
+                    (json.dumps(cur, ensure_ascii=False), session_id, seq),
+                )
+                conn.commit()
+                return seq
+            except Exception as e:
+                logger.warning(f"[ConversationStore] attach_extras failed: {e}")
+                return None
+            finally:
+                conn.close()
+
     def load_history_page(
         self,
         session_id: str,
@@ -698,7 +767,22 @@ class ConversationStore:
                 ).fetchone()
                 ctx_start = ctx_row[0] if ctx_row else 0
 
+                # extras column is added by migration; tolerate older DBs that
+                # might miss it by falling back to a NULL literal.
+                try:
                     rows = conn.execute(
+                        """
+                        SELECT seq, role, content, created_at, extras
+                        FROM messages
+                        WHERE session_id = ?
+                        ORDER BY seq ASC
+                        """,
+                        (session_id,),
+                    ).fetchall()
+                except sqlite3.OperationalError:
+                    rows = [
+                        (seq, role, content, created_at, "")
+                        for (seq, role, content, created_at) in conn.execute(
                             """
                             SELECT seq, role, content, created_at
                             FROM messages
@@ -707,6 +791,7 @@ class ConversationStore:
                             """,
                             (session_id,),
                         ).fetchall()
+                    ]
             finally:
                 conn.close()
 
@@ -719,13 +804,16 @@ class ConversationStore:
             include_thinking = False
 
         # Strip seq for display grouping, but record max seq per visible user group
-        plain_rows = [(role, content, created_at) for _seq, role, content, created_at in rows]
+        plain_rows = [
+            (role, content, created_at, extras_raw)
+            for _seq, role, content, created_at, extras_raw in rows
+        ]
         visible = _group_into_display_turns(plain_rows, include_thinking=include_thinking)
 
         # Build a mapping: find the seq of each visible user message to annotate context boundary.
         # Walk through rows to find visible user message seqs in order.
         visible_user_seqs: List[int] = []
-        for seq, role, raw_content, _ts in rows:
+        for seq, role, raw_content, _ts, _extras in rows:
             if role != "user":
                 continue
             try:
@@ -911,6 +999,18 @@ class ConversationStore:
             except Exception as e:
                 logger.warning(f"[ConversationStore] Migration (context_start_seq) failed: {e}")
 
+        msg_cols = {
+            row[1]
+            for row in conn.execute("PRAGMA table_info(messages)").fetchall()
+        }
+        if "extras" not in msg_cols:
+            try:
+                conn.execute(_MIGRATION_ADD_MSG_EXTRAS)
+                conn.commit()
+                logger.info("[ConversationStore] Migrated: added messages.extras column")
+            except Exception as e:
+                logger.warning(f"[ConversationStore] Migration (extras) failed: {e}")
+
     def _connect(self) -> sqlite3.Connection:
         conn = sqlite3.connect(str(self._db_path), timeout=10)
         conn.execute("PRAGMA journal_mode=WAL")

agent/protocol/agent_stream.py

@@ -603,15 +603,24 @@ class AgentStreamExecutor:
         except Exception as e:
             logger.debug(f"[Agent] MCP sync skipped: {e}")
 
-        # Prepare tool definitions (OpenAI/Claude format)
+        # Prepare tool definitions. Prefer get_json_schema() when it yields
+        # real properties (lets tools augment schema at runtime), otherwise
+        # fall back to the static `tool.params` (MCP tools rely on this).
         tools_schema = None
         if self.tools:
             tools_schema = []
             for tool in self.tools.values():
+                input_schema = tool.params
+                try:
+                    dynamic = (tool.get_json_schema() or {}).get("parameters") or {}
+                    if dynamic.get("properties"):
+                        input_schema = dynamic
+                except Exception:
+                    pass
                 tools_schema.append({
                     "name": tool.name,
                     "description": tool.description,
-                    "input_schema": tool.params  # Claude uses input_schema
+                    "input_schema": input_schema,
                 })
 
         # Create request

agent/tools/vision/vision.py

@@ -3,7 +3,7 @@ Vision tool - Analyze images using Vision API.
 Supports local files (auto base64-encoded) and HTTP URLs.
 
 Provider resolution:
-  - tool.vision.model (if set) means "prefer this model first; fall back to
+  - tools.vision.model (if set) means "prefer this model first; fall back to
     other configured providers if it fails". The model name is mapped to its
     native provider (e.g. doubao-* → Doubao, kimi-* → Moonshot, gpt-* →
     OpenAI/LinkAI). That provider is tried first, then the standard auto
@@ -30,7 +30,7 @@ from common import const
 from common.log import logger
 from config import conf
 
-DEFAULT_MODEL = const.GPT_41_MINI
+DEFAULT_MODEL = const.GPT_55
 DEFAULT_TIMEOUT = 60
 MAX_TOKENS = 1000
 COMPRESS_THRESHOLD = 1_048_576  # 1 MB
@@ -53,14 +53,14 @@ _DISCOVERABLE_MODELS = [
     ("ark_api_key", const.DOUBAO, const.DOUBAO_SEED_2_PRO, "Doubao"),
     ("dashscope_api_key", const.QWEN_DASHSCOPE, const.QWEN36_PLUS, "DashScope"),
     ("claude_api_key", const.CLAUDEAPI, const.CLAUDE_4_6_SONNET, "Claude"),
-    ("gemini_api_key", const.GEMINI, const.GEMINI_31_FLASH_LITE_PRE, "Gemini"),
+    ("gemini_api_key", const.GEMINI, const.GEMINI_35_FLASH, "Gemini"),
     ("qianfan_api_key", const.QIANFAN, const.ERNIE_45_TURBO_VL, "Qianfan"),
     ("zhipu_ai_api_key", const.ZHIPU_AI, const.GLM_4_7, "ZhipuAI"),
     ("minimax_api_key", const.MiniMax, const.MINIMAX_M2_7, "MiniMax"),
 ]
 
 # Model name prefix → discoverable provider display_name.
-# Used to auto-route tool.vision.model to its native provider.
+# Used to auto-route tools.vision.model to its native provider.
 # Matched case-insensitively; longest prefix wins.
 _MODEL_PREFIX_TO_PROVIDER = [
     ("doubao-", "Doubao"),
@@ -154,7 +154,7 @@ class Vision(BaseTool):
 
         # Default model is only used as a last-resort placeholder for providers
         # whose VisionProvider.model_override is None (e.g. raw OpenAI provider
-        # when the user did not configure tool.vision.model).
+        # when the user did not configure tools.vision.model).
         return self._call_with_fallback(providers, DEFAULT_MODEL, question, image_content)
 
     def _call_with_fallback(self, providers: List[VisionProvider], model: str,
@@ -193,12 +193,12 @@ class Vision(BaseTool):
         """
         Build an ordered list of providers to try.
 
-        Semantics of `tool.vision.model`:
+        Semantics of `tools.vision.model`:
           "Prefer this model first; fall back to other configured providers
            if it fails."
 
         Order:
-          1. The provider that natively serves `tool.vision.model` (if any
+          1. The provider that natively serves `tools.vision.model` (if any
              and its API key is configured) — using the user-specified model
              name verbatim.
           2. Auto-discovery chain as fallback:
@@ -213,7 +213,7 @@ class Vision(BaseTool):
         user_model = self._resolve_user_vision_model()
         providers: List[VisionProvider] = []
 
-        # Step 1: preferred provider derived from tool.vision.model
+        # Step 1: preferred provider derived from tools.vision.model
         if user_model:
             preferred = self._route_by_model_name(user_model)
             if preferred:
@@ -251,11 +251,11 @@ class Vision(BaseTool):
 
     @staticmethod
     def _resolve_user_vision_model() -> Optional[str]:
-        """Read tool.vision.model from config; return None if unset/blank."""
-        tool_conf = conf().get("tool", {})
-        if not isinstance(tool_conf, dict):
+        """Read tools.vision.model (singular ``tool`` kept as runtime fallback)."""
+        tools_conf = conf().get("tools") or conf().get("tool") or {}
+        if not isinstance(tools_conf, dict):
             return None
-        vision_conf = tool_conf.get("vision", {})
+        vision_conf = tools_conf.get("vision", {})
         if not isinstance(vision_conf, dict):
             return None
         m = vision_conf.get("model")
@@ -303,7 +303,7 @@ class Vision(BaseTool):
                 self._append_provider(providers, lambda: self._build_linkai_provider(user_model))
             if providers:
                 return providers
-            logger.warning(f"[Vision] tool.vision.model='{user_model}' looks like an OpenAI "
+            logger.warning(f"[Vision] tools.vision.model='{user_model}' looks like an OpenAI "
                            f"model but neither OPENAI_API_KEY nor LINKAI_API_KEY is configured.")
             return None  # fall through to auto
 
@@ -317,7 +317,7 @@ class Vision(BaseTool):
                 continue
             api_key = conf().get(config_key, "")
             if not api_key or not api_key.strip():
-                logger.warning(f"[Vision] tool.vision.model='{user_model}' routes to "
+                logger.warning(f"[Vision] tools.vision.model='{user_model}' routes to "
                                f"'{display_name}' but '{config_key}' is not configured. "
                                f"Falling back to auto-discovery.")
                 return None  # fall through to auto
@@ -452,8 +452,8 @@ class Vision(BaseTool):
         if not self._main_bot_supports_vision(bot):
             return None
 
-        # Use the configured main model name; do NOT inject tool.vision.model
-        # here, because by the time we reach this branch the tool.vision.model
+        # Use the configured main model name; do NOT inject tools.vision.model
+        # here, because by the time we reach this branch the tools.vision.model
         # routing has already been attempted (and either matched the main bot
         # or failed to find a provider).
         main_model_name = conf().get("model") or None

agent/tools/web_search/web_search.py

@@ -1,13 +1,27 @@
-"""
-Web Search tool - Search the web using Bocha or LinkAI search API.
-Supports two backends with unified response format:
-  1. Bocha Search (primary, requires BOCHA_API_KEY)
-  2. LinkAI Search (fallback, requires LINKAI_API_KEY)
+"""Web Search tool. Supports four backends with a unified response format:
+  - bocha   (https://open.bochaai.com)
+  - zhipu   (https://docs.bigmodel.cn/cn/guide/tools/web-search)
+  - qianfan (https://cloud.baidu.com/doc/qianfan/s/2mh4su4uy)
+  - linkai  (https://link-ai.tech, fallback)
+
+Provider selection
+  - strategy 'auto' (default): pick the first configured provider in the
+    canonical order [bocha, zhipu, qianfan, linkai]. When the caller passes
+    an explicit `provider` it overrides the pick; an invalid/unconfigured
+    one silently falls back to the auto order.
+  - strategy 'fixed': use the configured provider; if its credential is
+    missing at call time, silently fall back to auto order (no card hint).
+
+Credentials
+  - bocha   : tools.web_search.bocha_api_key  ->  env BOCHA_API_KEY
+  - zhipu   : conf.zhipu_ai_api_key            ->  env ZHIPUAI_API_KEY
+  - qianfan : conf.qianfan_api_key             ->  env QIANFAN_API_KEY
+  - linkai  : conf.linkai_api_key              ->  env LINKAI_API_KEY
 """
 
-import os
 import json
-from typing import Dict, Any, Optional
+import os
+from typing import Any, Dict, List, Optional
 
 import requests
 
@@ -16,12 +30,63 @@ from common.log import logger
 from config import conf
 
 
-# Default timeout for API requests (seconds)
 DEFAULT_TIMEOUT = 30
 
+# Canonical fallback order. Empirically ordered by Chinese real-time
+# quality + relevance: bocha (best overall), qianfan (best for hot news),
+# zhipu (strong on long-form articles), linkai (cloud aggregator, last
+# resort).
+PROVIDER_ORDER = ("bocha", "qianfan", "zhipu", "linkai")
+
+PROVIDER_LABELS = {
+    "bocha":   "Bocha",
+    "zhipu":   "Zhipu",
+    "qianfan": "Baidu Qianfan",
+    "linkai":  "LinkAI",
+}
+
+
+def _tools_web_search_conf() -> dict:
+    """Return the tools.web_search config block (dict-like)."""
+    tools_cfg = conf().get("tools") or {}
+    if not isinstance(tools_cfg, dict):
+        return {}
+    block = tools_cfg.get("web_search") or {}
+    return block if isinstance(block, dict) else {}
+
+
+def _get_api_key(provider: str) -> str:
+    """Resolve API key for a provider, with conf -> env fallback."""
+    if provider == "bocha":
+        key = (_tools_web_search_conf().get("bocha_api_key") or "").strip()
+        return key or os.environ.get("BOCHA_API_KEY", "").strip()
+    if provider == "zhipu":
+        key = (conf().get("zhipu_ai_api_key") or "").strip()
+        return key or os.environ.get("ZHIPUAI_API_KEY", "").strip()
+    if provider == "qianfan":
+        key = (conf().get("qianfan_api_key") or "").strip()
+        return key or os.environ.get("QIANFAN_API_KEY", "").strip()
+    if provider == "linkai":
+        key = (conf().get("linkai_api_key") or "").strip()
+        return key or os.environ.get("LINKAI_API_KEY", "").strip()
+    return ""
+
+
+def configured_providers() -> List[str]:
+    """Return configured providers in canonical order."""
+    return [p for p in PROVIDER_ORDER if _get_api_key(p)]
+
+
+def _configured_strategy() -> str:
+    return (_tools_web_search_conf().get("strategy") or "auto").strip().lower()
+
+
+def _configured_provider() -> str:
+    return (_tools_web_search_conf().get("provider") or "").strip().lower()
+
 
 class WebSearch(BaseTool):
-    """Tool for searching the web using Bocha or LinkAI search API"""
+    """Tool for searching the web across multiple providers."""
 
     name: str = "web_search"
     description: str = "Search the web for real-time information. Returns titles, URLs, and snippets."
@@ -55,264 +120,368 @@ class WebSearch(BaseTool):
 
     def __init__(self, config: dict = None):
         self.config = config or {}
-        self._backend = None  # Will be resolved on first execute
 
     @staticmethod
     def is_available() -> bool:
-        """Check if web search is available (at least one API key is configured)"""
-        return bool(os.environ.get("BOCHA_API_KEY") or os.environ.get("LINKAI_API_KEY"))
+        """Tool is offered to the agent when at least one provider has a key."""
+        return bool(configured_providers())
 
-    def _resolve_backend(self) -> Optional[str]:
-        """
-        Determine which search backend to use.
-        Priority: Bocha > LinkAI
+    @classmethod
+    def get_json_schema(cls) -> dict:
+        """Augment the static schema with a `provider` field — only when the
+        user has ≥2 providers configured AND strategy is 'auto'. Otherwise
+        the backend picks silently and exposing the field would only waste
+        the agent's tokens."""
+        schema = {
+            "name": cls.name,
+            "description": cls.description,
+            "parameters": json.loads(json.dumps(cls.params)),  # deep copy
+        }
+        if _configured_strategy() != "auto":
+            return schema
+        available = configured_providers()
+        if len(available) < 2:
+            return schema
 
-        :return: 'bocha', 'linkai', or None
+        schema["parameters"]["properties"]["provider"] = {
+            "type": "string",
+            "enum": available,
+            "description": "Optional. Specifies the search backend. You may switch between providers when the user wants results from a particular source or from multiple sources.",
+        }
+        return schema
+
+    # ------------------------------------------------------------------
+    # Provider resolution
+    # ------------------------------------------------------------------
+
+    def _resolve_provider(self, requested: Optional[str]) -> Optional[str]:
+        """Pick a provider for this call.
+
+        Priority: caller-supplied (if configured) > fixed strategy (if
+        configured) > first configured in PROVIDER_ORDER. Silent fallback
+        when the desired one has no key.
         """
-        if os.environ.get("BOCHA_API_KEY"):
-            return "bocha"
-        if os.environ.get("LINKAI_API_KEY"):
-            return "linkai"
+        available = configured_providers()
+        if not available:
             return None
 
-    def execute(self, args: Dict[str, Any]) -> ToolResult:
-        """
-        Execute web search
+        if requested:
+            req = requested.strip().lower()
+            if req in available:
+                return req
+            logger.warning(f"[WebSearch] requested provider '{requested}' unavailable, falling back")
 
-        :param args: Search parameters (query, count, freshness, summary)
-        :return: Search results
-        """
-        query = args.get("query", "").strip()
+        if _configured_strategy() == "fixed":
+            pinned = _configured_provider()
+            if pinned in available:
+                return pinned
+            if pinned:
+                logger.warning(f"[WebSearch] pinned provider '{pinned}' unavailable, falling back to auto")
+
+        return available[0]
+
+    @staticmethod
+    def _resolution_reason(requested: Optional[str], chosen: str) -> str:
+        """Human-readable explanation for why `chosen` won the resolver."""
+        if requested and requested.strip().lower() == chosen:
+            return "caller-requested"
+        strategy = _configured_strategy()
+        if strategy == "fixed" and _configured_provider() == chosen:
+            return "fixed-strategy"
+        return "auto-fallback"
+
+    # ------------------------------------------------------------------
+    # Entry point
+    # ------------------------------------------------------------------
+
+    def execute(self, args: Dict[str, Any]) -> ToolResult:
+        query = (args.get("query") or "").strip()
         if not query:
             return ToolResult.fail("Error: 'query' parameter is required")
 
         count = args.get("count", 10)
         freshness = args.get("freshness", "noLimit")
         summary = args.get("summary", False)
-
-        # Validate count
         if not isinstance(count, int) or count < 1 or count > 50:
             count = 10
 
-        # Resolve backend
-        backend = self._resolve_backend()
-        if not backend:
+        requested = args.get("provider")
+        provider = self._resolve_provider(requested)
+        if not provider:
             return ToolResult.fail(
-                "Error: No search API key configured. "
-                "Please set BOCHA_API_KEY or LINKAI_API_KEY using env_config tool.\n"
-                "  - Bocha Search: https://open.bocha.cn\n"
-                "  - LinkAI Search: https://link-ai.tech"
+                "Error: No search provider configured. "
+                "Configure one of BOCHA_API_KEY / zhipu_ai_api_key / qianfan_api_key / linkai_api_key."
+            )
+
+        # Always log the routing decision so multi-provider deployments can
+        # tell at a glance which backend served any given query.
+        available = configured_providers()
+        reason = self._resolution_reason(requested, provider)
+        q_preview = query if len(query) <= 60 else (query[:57] + "...")
+        logger.info(
+            f"[WebSearch] provider={provider} reason={reason} "
+            f"available={list(available)} query={q_preview!r} count={count} freshness={freshness}"
         )
 
         try:
-            if backend == "bocha":
+            if provider == "bocha":
                 return self._search_bocha(query, count, freshness, summary)
-            else:
+            if provider == "zhipu":
+                return self._search_zhipu(query, count, freshness)
+            if provider == "qianfan":
+                return self._search_qianfan(query, count, freshness)
+            if provider == "linkai":
                 return self._search_linkai(query, count, freshness)
+            return ToolResult.fail(f"Error: Unknown provider '{provider}'")
         except requests.Timeout:
             return ToolResult.fail(f"Error: Search request timed out after {DEFAULT_TIMEOUT}s")
         except requests.ConnectionError:
             return ToolResult.fail("Error: Failed to connect to search API")
         except Exception as e:
-            logger.error(f"[WebSearch] Unexpected error: {e}", exc_info=True)
+            logger.error(f"[WebSearch] Unexpected error ({provider}): {e}", exc_info=True)
             return ToolResult.fail(f"Error: Search failed - {str(e)}")
 
+    # ------------------------------------------------------------------
+    # Bocha
+    # ------------------------------------------------------------------
+
     def _search_bocha(self, query: str, count: int, freshness: str, summary: bool) -> ToolResult:
-        """
-        Search using Bocha API
-
-        :param query: Search query
-        :param count: Number of results
-        :param freshness: Time range filter
-        :param summary: Whether to include summary
-        :return: Formatted search results
-        """
-        api_key = os.environ.get("BOCHA_API_KEY", "")
-        url = "https://api.bocha.cn/v1/web-search"
-
+        api_key = _get_api_key("bocha")
+        url = "https://api.bochaai.com/v1/web-search"
         headers = {
             "Authorization": f"Bearer {api_key}",
             "Content-Type": "application/json",
-            "Accept": "application/json"
+            "Accept": "application/json",
         }
+        payload = {"query": query, "count": count, "freshness": freshness, "summary": summary}
 
-        payload = {
-            "query": query,
-            "count": count,
-            "freshness": freshness,
-            "summary": summary
-        }
+        logger.debug(f"[WebSearch] bocha: query='{query}', count={count}")
+        resp = requests.post(url, headers=headers, json=payload, timeout=DEFAULT_TIMEOUT)
 
-        logger.debug(f"[WebSearch] Bocha search: query='{query}', count={count}")
+        if resp.status_code == 401:
+            return ToolResult.fail("Error: Invalid bocha API key.")
+        if resp.status_code == 403:
+            return ToolResult.fail("Error: bocha API — insufficient balance. Top up at https://open.bochaai.com")
+        if resp.status_code == 429:
+            return ToolResult.fail("Error: bocha API rate limit reached.")
+        if resp.status_code != 200:
+            return ToolResult.fail(f"Error: bocha API returned HTTP {resp.status_code}")
 
-        response = requests.post(url, headers=headers, json=payload, timeout=DEFAULT_TIMEOUT)
-
-        if response.status_code == 401:
-            return ToolResult.fail("Error: Invalid BOCHA_API_KEY. Please check your API key.")
-        if response.status_code == 403:
-            return ToolResult.fail("Error: Bocha API - insufficient balance. Please top up at https://open.bocha.cn")
-        if response.status_code == 429:
-            return ToolResult.fail("Error: Bocha API rate limit reached. Please try again later.")
-        if response.status_code != 200:
-            return ToolResult.fail(f"Error: Bocha API returned HTTP {response.status_code}")
-
-        data = response.json()
-
-        # Check API-level error code
+        data = resp.json()
         api_code = data.get("code")
         if api_code is not None and api_code != 200:
             msg = data.get("msg") or "Unknown error"
-            return ToolResult.fail(f"Error: Bocha API error (code={api_code}): {msg}")
-
-        # Extract and format results
-        return self._format_bocha_results(data, query)
-
-    def _format_bocha_results(self, data: dict, query: str) -> ToolResult:
-        """
-        Format Bocha API response into unified result structure
-
-        :param data: Raw API response
-        :param query: Original query
-        :return: Formatted ToolResult
-        """
-        search_data = data.get("data", {})
-        web_pages = search_data.get("webPages", {})
-        pages = web_pages.get("value", [])
-
-        if not pages:
-            return ToolResult.success({
-                "query": query,
-                "backend": "bocha",
-                "total": 0,
-                "results": [],
-                "message": "No results found"
-            })
+            return ToolResult.fail(f"Error: bocha API error (code={api_code}): {msg}")
 
+        pages = (data.get("data") or {}).get("webPages", {}).get("value", []) or []
         results = []
-        for page in pages:
-            result = {
-                "title": page.get("name", ""),
-                "url": page.get("url", ""),
-                "snippet": page.get("snippet", ""),
-                "siteName": page.get("siteName", ""),
-                "datePublished": page.get("datePublished") or page.get("dateLastCrawled", ""),
+        for p in pages:
+            item = {
+                "title": p.get("name", ""),
+                "url": p.get("url", ""),
+                "snippet": p.get("snippet", ""),
+                "siteName": p.get("siteName", ""),
+                "datePublished": p.get("datePublished") or p.get("dateLastCrawled", ""),
             }
-            # Include summary only if present
-            if page.get("summary"):
-                result["summary"] = page["summary"]
-            results.append(result)
-
-        total = web_pages.get("totalEstimatedMatches", len(results))
-
+            if p.get("summary"):
+                item["summary"] = p["summary"]
+            results.append(item)
+        total = (data.get("data") or {}).get("webPages", {}).get("totalEstimatedMatches", len(results))
         return ToolResult.success({
-            "query": query,
-            "backend": "bocha",
-            "total": total,
-            "count": len(results),
-            "results": results
+            "query": query, "backend": "bocha",
+            "total": total, "count": len(results), "results": results,
         })
 
+    # ------------------------------------------------------------------
+    # Zhipu
+    # ------------------------------------------------------------------
+
+    def _search_zhipu(self, query: str, count: int, freshness: str) -> ToolResult:
+        api_key = _get_api_key("zhipu")
+        api_base = (conf().get("zhipu_ai_api_base") or "https://open.bigmodel.cn/api/paas/v4").rstrip("/")
+        url = f"{api_base}/web_search"
+        headers = {
+            "Authorization": f"Bearer {api_key}",
+            "Content-Type": "application/json",
+        }
+
+        # Zhipu Web Search expects `search_query` <= 70 chars; truncate
+        # gracefully so a long agent-supplied query doesn't get rejected.
+        trimmed_query = (query or "")[:70]
+        engine = (_tools_web_search_conf().get("zhipu_search_engine") or "search_pro").strip().lower()
+        if engine not in ("search_std", "search_pro", "search_pro_sogou", "search_pro_quark"):
+            engine = "search_pro"
+
+        payload: Dict[str, Any] = {
+            "search_engine": engine,
+            "search_query": trimmed_query,
+            "search_intent": False,
+            "count": max(1, min(int(count or 10), 50)),
+            "search_recency_filter": freshness if freshness in (
+                "oneDay", "oneWeek", "oneMonth", "oneYear", "noLimit"
+            ) else "noLimit",
+        }
+        content_size = (_tools_web_search_conf().get("zhipu_content_size") or "").strip().lower()
+        if content_size in ("medium", "high"):
+            payload["content_size"] = content_size
+
+        logger.debug(f"[WebSearch] zhipu: query='{trimmed_query}', count={payload['count']}, engine={engine}")
+        resp = requests.post(url, headers=headers, json=payload, timeout=DEFAULT_TIMEOUT)
+
+        if resp.status_code == 401:
+            return ToolResult.fail("Error: Invalid Zhipu API key.")
+        if resp.status_code != 200:
+            return ToolResult.fail(f"Error: Zhipu API returned HTTP {resp.status_code}: {resp.text[:200]}")
+
+        data = resp.json()
+        # Business-level errors (1701/1702/1703 etc.) come back as
+        # {"error": {"code","message"}} even on HTTP 200.
+        if isinstance(data, dict) and data.get("error"):
+            err = data["error"] or {}
+            return ToolResult.fail(f"Error: Zhipu returned {err.get('code')}: {err.get('message','')}")
+
+        items = data.get("search_result") or (data.get("data") or {}).get("search_result") or []
+        results = []
+        for it in items:
+            results.append({
+                "title": it.get("title", ""),
+                "url": it.get("link") or it.get("url", ""),
+                "snippet": it.get("content") or it.get("snippet", ""),
+                "siteName": it.get("media") or it.get("siteName", ""),
+                "datePublished": it.get("publish_date") or it.get("datePublished", ""),
+            })
+        return ToolResult.success({
+            "query": query, "backend": "zhipu",
+            "total": len(results), "count": len(results), "results": results,
+        })
+
+    # ------------------------------------------------------------------
+    # Qianfan (Baidu)
+    # ------------------------------------------------------------------
+
+    def _search_qianfan(self, query: str, count: int, freshness: str) -> ToolResult:
+        api_key = _get_api_key("qianfan")
+        api_base = (conf().get("qianfan_api_base") or "https://qianfan.baidubce.com/v2").rstrip("/")
+        url = f"{api_base}/ai_search/web_search"
+        headers = {
+            "Authorization": f"Bearer {api_key}",
+            "Content-Type": "application/json",
+            "X-Appbuilder-From": "cow",
+        }
+
+        count = max(1, min(int(count or 10), 50))
+        payload: Dict[str, Any] = {
+            "messages": [{"role": "user", "content": query}],
+            "search_source": "baidu_search_v2",
+            "resource_type_filter": [{"type": "web", "top_k": count}],
+        }
+
+        # Baidu AI Search expects freshness as a date-range filter, not a
+        # named recency token. Translate our shared vocabulary into the
+        # underlying page_time range expected by the API.
+        search_filter = self._qianfan_build_freshness_filter(freshness)
+        if search_filter:
+            payload["search_filter"] = search_filter
+
+        logger.debug(f"[WebSearch] qianfan: query='{query}', count={count}, freshness={freshness!r}")
+        resp = requests.post(url, headers=headers, json=payload, timeout=DEFAULT_TIMEOUT)
+
+        if resp.status_code == 401:
+            return ToolResult.fail("Error: Invalid Qianfan API key.")
+        if resp.status_code != 200:
+            return ToolResult.fail(f"Error: Qianfan API returned HTTP {resp.status_code}: {resp.text[:200]}")
+
+        data = resp.json()
+        # Even on HTTP 200 Baidu surfaces business errors as {"code","message"}.
+        if isinstance(data, dict) and data.get("code"):
+            return ToolResult.fail(f"Error: Qianfan returned {data.get('code')}: {data.get('message','')}")
+
+        refs = data.get("references") or []
+        results = []
+        for d in refs:
+            results.append({
+                "title": d.get("title", ""),
+                "url": d.get("url", ""),
+                "snippet": (d.get("content") or "")[:200],
+                "siteName": d.get("web_anchor") or d.get("website") or "",
+                "datePublished": d.get("date", ""),
+            })
+        return ToolResult.success({
+            "query": query, "backend": "qianfan",
+            "total": len(results), "count": len(results), "results": results,
+        })
+
+    @staticmethod
+    def _qianfan_build_freshness_filter(freshness: str) -> Optional[Dict[str, Any]]:
+        if not freshness or freshness == "noLimit":
+            return None
+        delta_days = {"oneDay": 1, "oneWeek": 7, "oneMonth": 30, "oneYear": 365}.get(freshness)
+        if not delta_days:
+            return None
+        from datetime import datetime, timedelta
+        now = datetime.now()
+        end_date = (now + timedelta(days=1)).strftime("%Y-%m-%d")
+        start_date = (now - timedelta(days=delta_days)).strftime("%Y-%m-%d")
+        return {"range": {"page_time": {"gte": start_date, "lt": end_date}}}
+
+    # ------------------------------------------------------------------
+    # LinkAI (plugin)
+    # ------------------------------------------------------------------
+
     def _search_linkai(self, query: str, count: int, freshness: str) -> ToolResult:
-        """
-        Search using LinkAI plugin API
-
-        :param query: Search query
-        :param count: Number of results
-        :param freshness: Time range filter
-        :return: Formatted search results
-        """
-        api_key = os.environ.get("LINKAI_API_KEY", "")
-        api_base = conf().get("linkai_api_base", "https://api.link-ai.tech")
-        url = f"{api_base.rstrip('/')}/v1/plugin/execute"
+        api_key = _get_api_key("linkai")
+        api_base = (conf().get("linkai_api_base") or "https://api.link-ai.tech").rstrip("/")
+        url = f"{api_base}/v1/plugin/execute"
 
         from common.utils import get_cloud_headers
         headers = get_cloud_headers(api_key)
 
-        payload = {
-            "code": "web-search",
-            "args": {
-                "query": query,
-                "count": count,
-                "freshness": freshness
-            }
-        }
+        payload = {"code": "web-search", "args": {"query": query, "count": count, "freshness": freshness}}
+        logger.debug(f"[WebSearch] linkai: query='{query}', count={count}")
+        resp = requests.post(url, headers=headers, json=payload, timeout=DEFAULT_TIMEOUT)
 
-        logger.debug(f"[WebSearch] LinkAI search: query='{query}', count={count}")
-
-        response = requests.post(url, headers=headers, json=payload, timeout=DEFAULT_TIMEOUT)
-
-        if response.status_code == 401:
-            return ToolResult.fail("Error: Invalid LINKAI_API_KEY. Please check your API key.")
-        if response.status_code != 200:
-            return ToolResult.fail(f"Error: LinkAI API returned HTTP {response.status_code}")
-
-        data = response.json()
+        if resp.status_code == 401:
+            return ToolResult.fail("Error: Invalid LinkAI API key.")
+        if resp.status_code != 200:
+            return ToolResult.fail(f"Error: LinkAI API returned HTTP {resp.status_code}")
 
+        data = resp.json()
         if not data.get("success"):
             msg = data.get("message") or "Unknown error"
             return ToolResult.fail(f"Error: LinkAI search failed: {msg}")
 
-        return self._format_linkai_results(data, query)
-
-    def _format_linkai_results(self, data: dict, query: str) -> ToolResult:
-        """
-        Format LinkAI API response into unified result structure.
-        LinkAI returns the search data in data.data field, which follows
-        the same Bing-compatible format as Bocha.
-
-        :param data: Raw API response
-        :param query: Original query
-        :return: Formatted ToolResult
-        """
-        raw_data = data.get("data", "")
-
-        # LinkAI may return data as a JSON string
-        if isinstance(raw_data, str):
+        raw = data.get("data", "")
+        if isinstance(raw, str):
             try:
-                raw_data = json.loads(raw_data)
+                raw = json.loads(raw)
             except (json.JSONDecodeError, TypeError):
-                # If data is plain text, return it as a single result
                 return ToolResult.success({
-                    "query": query,
-                    "backend": "linkai",
-                    "total": 1,
-                    "count": 1,
-                    "results": [{"content": raw_data}]
+                    "query": query, "backend": "linkai",
+                    "total": 1, "count": 1, "results": [{"content": raw}],
                 })
 
-        # If the response follows Bing-compatible structure
-        if isinstance(raw_data, dict):
-            web_pages = raw_data.get("webPages", {})
-            pages = web_pages.get("value", [])
-
+        if isinstance(raw, dict):
+            pages = (raw.get("webPages") or {}).get("value", []) or []
             if pages:
                 results = []
-                for page in pages:
-                    result = {
-                        "title": page.get("name", ""),
-                        "url": page.get("url", ""),
-                        "snippet": page.get("snippet", ""),
-                        "siteName": page.get("siteName", ""),
-                        "datePublished": page.get("datePublished") or page.get("dateLastCrawled", ""),
+                for p in pages:
+                    item = {
+                        "title": p.get("name", ""),
+                        "url": p.get("url", ""),
+                        "snippet": p.get("snippet", ""),
+                        "siteName": p.get("siteName", ""),
+                        "datePublished": p.get("datePublished") or p.get("dateLastCrawled", ""),
                     }
-                    if page.get("summary"):
-                        result["summary"] = page["summary"]
-                    results.append(result)
-
-                total = web_pages.get("totalEstimatedMatches", len(results))
+                    if p.get("summary"):
+                        item["summary"] = p["summary"]
+                    results.append(item)
+                total = (raw.get("webPages") or {}).get("totalEstimatedMatches", len(results))
                 return ToolResult.success({
-                    "query": query,
-                    "backend": "linkai",
-                    "total": total,
-                    "count": len(results),
-                    "results": results
+                    "query": query, "backend": "linkai",
+                    "total": total, "count": len(results), "results": results,
                 })
 
-        # Fallback: return raw data
         return ToolResult.success({
-            "query": query,
-            "backend": "linkai",
-            "total": 1,
-            "count": 1,
-            "results": [{"content": str(raw_data)}]
+            "query": query, "backend": "linkai",
+            "total": 1, "count": 1, "results": [{"content": str(raw)}],
         })

bridge/agent_initializer.py

@@ -521,7 +521,7 @@ class AgentInitializer:
                 if tool_name == "web_search":
                     from agent.tools.web_search.web_search import WebSearch
                     if not WebSearch.is_available():
-                        logger.debug("[AgentInitializer] WebSearch skipped - no BOCHA_API_KEY or LINKAI_API_KEY")
+                        logger.debug("[AgentInitializer] WebSearch skipped - no search provider configured")
                         continue
 
                 # Special handling for EnvConfig tool

bridge/bridge.py

@@ -14,7 +14,9 @@ class Bridge(object):
     def __init__(self):
         self.btype = {
             "chat": const.OPENAI,
-            "voice_to_text": conf().get("voice_to_text", "openai"),
+            # Empty `voice_to_text` (the default in new configs) triggers
+            # the auto-pick below — see _auto_pick_voice_to_text for order.
+            "voice_to_text": conf().get("voice_to_text") or self._auto_pick_voice_to_text(),
             "text_to_voice": conf().get("text_to_voice", "google"),
             "translate": conf().get("translate", "baidu"),
         }
@@ -84,6 +86,46 @@ class Bridge(object):
         self.chat_bots = {}
         self._agent_bridge = None
 
+    def refresh_voice(self):
+        """Re-read voice_to_text / text_to_voice from config and drop the
+        cached voice bots so the next call picks up the new provider.
+        Used by the web console after the user edits voice settings.
+        Does NOT touch the agent_bridge / agent state.
+        """
+        new_v2t = conf().get("voice_to_text") or self._auto_pick_voice_to_text()
+        new_t2v = conf().get("text_to_voice", "google")
+        if conf().get("use_linkai") and conf().get("linkai_api_key"):
+            if not conf().get("voice_to_text") or conf().get("voice_to_text") in ["openai"]:
+                new_v2t = const.LINKAI
+            if not conf().get("text_to_voice") or conf().get("text_to_voice") in ["openai", const.TTS_1, const.TTS_1_HD]:
+                new_t2v = const.LINKAI
+        self.btype["voice_to_text"] = new_v2t
+        self.btype["text_to_voice"] = new_t2v
+        self.bots.pop("voice_to_text", None)
+        self.bots.pop("text_to_voice", None)
+        logger.info(f"[Bridge] voice refreshed: voice_to_text={new_v2t}, text_to_voice={new_t2v}")
+
+    @staticmethod
+    def _auto_pick_voice_to_text() -> str:
+        """Pick an ASR provider by configured api keys when voice_to_text is
+        unset. Order matches the web console: openai → dashscope → zhipu →
+        linkai. Falls back to 'openai' when nothing is configured so the
+        original "missing key" error is preserved.
+        """
+        def has(k: str) -> bool:
+            v = (conf().get(k) or "").strip()
+            return v != "" and v not in ("YOUR API KEY", "YOUR_API_KEY")
+
+        for key, provider in (
+            ("open_ai_api_key", "openai"),
+            ("dashscope_api_key", "dashscope"),
+            ("zhipu_ai_api_key", "zhipu"),
+            ("linkai_api_key", "linkai"),
+        ):
+            if has(key):
+                return provider
+        return "openai"
+
     # 模型对应的接口
     def get_bot(self, typename):
         if self.bots.get(typename) is None:

channel/chat_channel.py

@@ -171,7 +171,13 @@ class ChatChannel(Channel):
             if "desire_rtype" not in context and conf().get("always_reply_voice") and ReplyType.VOICE not in self.NOT_SUPPORT_REPLYTYPE:
                 context["desire_rtype"] = ReplyType.VOICE
         elif context.type == ContextType.VOICE:
-            if "desire_rtype" not in context and conf().get("voice_reply_voice") and ReplyType.VOICE not in self.NOT_SUPPORT_REPLYTYPE:
+            # Voice input replies with voice when either voice_reply_voice
+            # (mirror voice) or the global always_reply_voice toggle is on.
+            if (
+                "desire_rtype" not in context
+                and (conf().get("voice_reply_voice") or conf().get("always_reply_voice"))
+                and ReplyType.VOICE not in self.NOT_SUPPORT_REPLYTYPE
+            ):
                 context["desire_rtype"] = ReplyType.VOICE
         return context
 
@@ -264,6 +270,8 @@ class ChatChannel(Channel):
                 if reply.type == ReplyType.TEXT:
                     reply_text = reply.content
                     if desire_rtype == ReplyType.VOICE and ReplyType.VOICE not in self.NOT_SUPPORT_REPLYTYPE:
+                        # Preserve original text for the "text-then-voice" pattern in _send_reply.
+                        context["voice_reply_text"] = reply.content
                         reply = super().build_text_to_voice(reply.content)
                         return self._decorate_reply(context, reply)
                     if context.get("isgroup", False):
@@ -311,6 +319,15 @@ class ChatChannel(Channel):
                     # 短暂延迟后发送图片
                     time.sleep(0.3)
                     self._send(reply, context)
+                # Send text bubble before voice, unless channel already streamed
+                # the text (feishu) or natively renders STT under the voice (wechatcom).
+                elif reply.type == ReplyType.VOICE and context.get("voice_reply_text") \
+                        and not context.get("feishu_streamed") \
+                        and context.get("channel_type") not in ("wechatcom_app",):
+                    text_reply = Reply(ReplyType.TEXT, context.get("voice_reply_text"))
+                    self._send(text_reply, context)
+                    time.sleep(0.3)
+                    self._send(reply, context)
                 else:
                     self._send(reply, context)
     

channel/dingtalk/dingtalk_channel.py

@@ -86,6 +86,8 @@ def _check(func):
 
 @singleton
 class DingTalkChanel(ChatChannel, dingtalk_stream.ChatbotHandler):
+    NOT_SUPPORT_REPLYTYPE = []
+
     dingtalk_client_id = conf().get('dingtalk_client_id')
     dingtalk_client_secret = conf().get('dingtalk_client_secret')
 
@@ -870,6 +872,48 @@ class DingTalkChanel(ChatChannel, dingtalk_stream.ChatbotHandler):
                     self.reply_text("抱歉，文件上传失败", incoming_message)
             return
         
+        # Native sampleAudio. Upload only accepts ogg/amr, so convert TTS mp3/wav to amr.
+        elif reply.type == ReplyType.VOICE:
+            logger.info(f"[DingTalk] Sending voice: {reply.content}")
+            access_token = self.get_access_token()
+            if not access_token:
+                logger.error("[DingTalk] Cannot get access token for voice")
+                self.reply_text("抱歉，语音发送失败（无法获取token）", incoming_message)
+                return
+
+            voice_path = reply.content
+            if voice_path.startswith("file://"):
+                voice_path = voice_path[7:]
+
+            amr_path = voice_path
+            duration_ms = 0
+            if not voice_path.lower().endswith((".amr", ".ogg")):
+                try:
+                    from voice.audio_convert import any_to_amr
+                    amr_path = os.path.splitext(voice_path)[0] + ".amr"
+                    duration_ms = int(any_to_amr(voice_path, amr_path) or 0)
+                except Exception as e:
+                    logger.error(f"[DingTalk] Failed to convert voice to amr: {e}")
+                    self.reply_text("抱歉，语音转码失败", incoming_message)
+                    return
+
+            media_id = self.upload_media(amr_path, media_type="voice")
+            if not media_id:
+                logger.error("[DingTalk] Failed to upload voice media")
+                self.reply_text("抱歉，语音上传失败", incoming_message)
+                return
+
+            msg_param = {
+                "mediaId": media_id,
+                "duration": str(duration_ms or 1000),
+            }
+            success = self._send_file_message(
+                access_token, incoming_message, "sampleAudio", msg_param, isgroup
+            )
+            if not success:
+                self.reply_text("抱歉，语音发送失败", incoming_message)
+            return
+
         # 处理文本消息
         elif reply.type == ReplyType.TEXT:
             logger.info(f"[DingTalk] Sending text message, length={len(reply.content)}")

channel/feishu/feishu_channel.py

@@ -1515,10 +1515,16 @@ class FeiShuChanel(ChatChannel):
             else:
                 context.type = ContextType.TEXT
             context.content = content.strip()
+            # Text input opts into voice replies only when the always-on toggle is set.
+            if "desire_rtype" not in context and conf().get("always_reply_voice"):
+                context["desire_rtype"] = ReplyType.VOICE
 
         elif context.type == ContextType.VOICE:
-            # 2.语音请求
-            if "desire_rtype" not in context and conf().get("voice_reply_voice"):
+            # 2.语音请求: voice input replies with voice if either
+            # voice_reply_voice (mirror reply) or always_reply_voice is on.
+            if "desire_rtype" not in context and (
+                conf().get("voice_reply_voice") or conf().get("always_reply_voice")
+            ):
                 context["desire_rtype"] = ReplyType.VOICE
 
         return context

channel/web/chat.html

@@ -137,6 +137,11 @@
                             <i class="fas fa-sliders item-icon text-xs w-5 text-center"></i>
                             <span data-i18n="menu_config">配置</span>
                         </a>
+                        <a class="sidebar-item flex items-center gap-3 px-3 py-2 rounded-lg cursor-pointer transition-all duration-150 hover:bg-white/5 hover:text-neutral-200 text-[14px]"
+                           data-view="models">
+                            <i class="fas fa-microchip item-icon text-xs w-5 text-center"></i>
+                            <span data-i18n="menu_models">模型</span>
+                        </a>
                         <a class="sidebar-item flex items-center gap-3 px-3 py-2 rounded-lg cursor-pointer transition-all duration-150 hover:bg-white/5 hover:text-neutral-200 text-[14px]"
                            data-view="skills">
                             <i class="fas fa-bolt item-icon text-xs w-5 text-center"></i>
@@ -417,8 +422,9 @@
                                     </button>
                                 </div>
                                 <div id="slash-menu" class="slash-menu hidden"></div>
+                                <div class="flex-1 min-w-0 relative flex items-center">
                                     <textarea id="chat-input"
-                                          class="flex-1 min-w-0 px-4 py-[10px] rounded-xl border border-slate-200 dark:border-slate-600
+                                              class="w-full pl-4 pr-11 py-[10px] rounded-xl border border-slate-200 dark:border-slate-600
                                                      bg-slate-50 dark:bg-white/5 text-slate-800 dark:text-slate-100
                                                      placeholder:text-slate-400 dark:placeholder:text-slate-500
                                                      focus:outline-none focus:ring-0 focus:border-primary-600
@@ -426,6 +432,14 @@
                                               rows="1"
                                               data-i18n-placeholder="input_placeholder"
                                               placeholder="输入消息，或输入 / 使用指令"></textarea>
+                                    <button id="mic-btn" type="button"
+                                            class="absolute right-2 top-1/2 -translate-y-1/2 w-8 h-8 flex items-center justify-center rounded-lg
+                                                   text-slate-400 hover:text-primary-500 hover:bg-primary-50 dark:hover:bg-primary-900/20
+                                                   cursor-pointer transition-colors duration-150"
+                                            data-i18n-title="mic_idle_title" title="点击录音 / 再按一次结束">
+                                        <i class="fas fa-microphone text-sm"></i>
+                                    </button>
+                                </div>
                                 <button id="send-btn"
                                         class="flex-shrink-0 w-10 h-10 flex items-center justify-center rounded-lg
                                                bg-primary-400 text-white hover:bg-primary-500
@@ -460,6 +474,11 @@
                                             <i class="fas fa-microchip text-primary-500 text-sm"></i>
                                         </div>
                                         <h3 class="font-semibold text-slate-800 dark:text-slate-100" data-i18n="config_model">模型配置</h3>
+                                        <a class="ml-auto text-xs text-slate-500 dark:text-slate-400 hover:text-primary-500 dark:hover:text-primary-400 cursor-pointer transition-colors flex items-center gap-1"
+                                           onclick="navigateTo('models')">
+                                            <span data-i18n="config_model_advanced">高级配置</span>
+                                            <i class="fas fa-arrow-right text-[10px]"></i>
+                                        </a>
                                     </div>
                                     <div class="space-y-5">
                                         <!-- Provider -->
@@ -850,6 +869,41 @@
                     </div>
                 </div>
 
+                <!-- ====================================================== -->
+                <!-- VIEW: Models                                            -->
+                <!-- ====================================================== -->
+                <div id="view-models" class="view">
+                    <!-- Tailwind JIT safelist: capability-card icon colors are
+                         emitted from JS template strings. Listing them here
+                         (display:none) guarantees the CDN-side compiler picks
+                         them up regardless of render timing. -->
+                    <div class="hidden bg-blue-50 dark:bg-blue-900/30 text-blue-500
+                                       bg-orange-50 dark:bg-orange-900/30 text-orange-500
+                                       bg-purple-50 dark:bg-purple-900/30 text-purple-500
+                                       bg-amber-50 dark:bg-amber-900/30 text-amber-500
+                                       bg-primary-50 dark:bg-primary-900/30 text-primary-500"></div>
+                    <div class="flex-1 overflow-y-auto p-6">
+                        <div class="max-w-4xl mx-auto">
+                            <div class="flex items-center justify-between mb-6">
+                                <div>
+                                    <h2 class="text-xl font-bold text-slate-800 dark:text-slate-100" data-i18n="models_title">模型管理</h2>
+                                    <p class="text-sm text-slate-500 dark:text-slate-400 mt-1" data-i18n="models_desc">统一管理对话、视觉、语音、向量、图像、搜索能力</p>
+                                </div>
+                                <button id="models-add-vendor-btn" onclick="openVendorModal('')"
+                                        class="flex items-center gap-2 px-4 py-2 rounded-lg bg-primary-500 hover:bg-primary-600
+                                               text-white text-sm font-medium cursor-pointer transition-colors duration-150">
+                                    <i class="fas fa-plus text-xs"></i>
+                                    <span data-i18n="models_add_vendor">添加厂商</span>
+                                </button>
+                            </div>
+                            <div id="models-loading" class="flex items-center gap-2 py-12 justify-center text-slate-400 dark:text-slate-500 text-sm">
+                                <i class="fas fa-spinner fa-spin text-xs"></i><span>Loading...</span>
+                            </div>
+                            <div id="models-content" class="grid gap-6 hidden"></div>
+                        </div>
+                    </div>
+                </div>
+
                 <!-- ====================================================== -->
                 <!-- VIEW: Channels                                          -->
                 <!-- ====================================================== -->
@@ -959,7 +1013,7 @@
     </div><!-- /app -->
 
     <!-- Confirm Dialog -->
-    <div id="confirm-dialog-overlay" class="fixed inset-0 bg-black/50 z-[100] hidden flex items-center justify-center">
+    <div id="confirm-dialog-overlay" class="fixed inset-0 bg-black/50 z-[200] hidden flex items-center justify-center">
         <div class="bg-white dark:bg-[#1A1A1A] rounded-2xl border border-slate-200 dark:border-white/10 shadow-xl
                     w-full max-w-sm mx-4 overflow-hidden">
             <div class="p-6">
@@ -984,6 +1038,77 @@
         </div>
     </div>
 
+    <!-- Vendor Credentials Modal -->
+    <div id="vendor-modal-overlay" class="fixed inset-0 bg-black/50 z-[100] hidden flex items-center justify-center">
+        <div class="bg-white dark:bg-[#1A1A1A] rounded-2xl border border-slate-200 dark:border-white/10 shadow-xl
+                    w-full max-w-md mx-4">
+            <div class="p-6">
+                <div class="flex items-center gap-3 mb-5">
+                    <div class="w-10 h-10 rounded-xl bg-primary-50 dark:bg-primary-900/20 flex items-center justify-center flex-shrink-0">
+                        <i class="fas fa-key text-primary-500"></i>
+                    </div>
+                    <div class="min-w-0 flex-1">
+                        <h3 id="vendor-modal-title" class="font-semibold text-slate-800 dark:text-slate-100 text-base"></h3>
+                        <p id="vendor-modal-subtitle" class="text-xs text-slate-500 dark:text-slate-400 mt-0.5 font-mono"></p>
+                    </div>
+                </div>
+
+                <!-- Provider selector (only visible when adding via top button) -->
+                <div id="vendor-modal-picker-wrap" class="mb-4 hidden">
+                    <label class="block text-sm font-medium text-slate-600 dark:text-slate-400 mb-1.5" data-i18n="models_provider">厂商</label>
+                    <div id="vendor-modal-picker" class="cfg-dropdown" tabindex="0">
+                        <div class="cfg-dropdown-selected">
+                            <span class="cfg-dropdown-text">--</span>
+                            <i class="fas fa-chevron-down cfg-dropdown-arrow"></i>
+                        </div>
+                        <div class="cfg-dropdown-menu"></div>
+                    </div>
+                </div>
+
+                <div class="space-y-4">
+                    <div>
+                        <label class="block text-sm font-medium text-slate-600 dark:text-slate-400 mb-1.5">API Key</label>
+                        <input id="vendor-modal-key" type="text" autocomplete="off" data-1p-ignore data-lpignore="true"
+                               class="w-full px-3 py-2 rounded-lg border border-slate-200 dark:border-slate-600
+                                      bg-slate-50 dark:bg-white/5 text-sm text-slate-800 dark:text-slate-100
+                                      focus:outline-none focus:border-primary-500 font-mono transition-colors"
+                               placeholder="sk-...">
+                    </div>
+                    <div id="vendor-modal-base-wrap">
+                        <label class="block text-sm font-medium text-slate-600 dark:text-slate-400 mb-1.5">API Base</label>
+                        <input id="vendor-modal-base" type="text"
+                               class="w-full px-3 py-2 rounded-lg border border-slate-200 dark:border-slate-600
+                                      bg-slate-50 dark:bg-white/5 text-sm text-slate-800 dark:text-slate-100
+                                      focus:outline-none focus:border-primary-500 font-mono transition-colors"
+                               placeholder="https://...../v1">
+                        <p id="vendor-modal-base-hint" class="mt-1.5 text-xs text-slate-400 dark:text-slate-500 hidden">
+                            <i class="fas fa-info-circle mr-1"></i><span data-i18n="models_base_default_hint">留空将使用官方默认地址</span>
+                        </p>
+                    </div>
+                </div>
+            </div>
+            <div class="flex items-center justify-between gap-3 px-6 py-4 border-t border-slate-100 dark:border-white/5 rounded-b-2xl">
+                <button id="vendor-modal-clear"
+                        class="px-3 py-2 rounded-lg text-xs
+                               text-red-500 dark:text-red-400 hover:bg-red-50 dark:hover:bg-red-900/20
+                               cursor-pointer transition-colors duration-150 hidden"
+                        data-i18n="models_clear_credential">清除凭据</button>
+                <span id="vendor-modal-status"
+                      class="flex-1 text-xs text-primary-500 opacity-0 transition-opacity duration-300 text-center"></span>
+                <button id="vendor-modal-cancel"
+                        class="px-4 py-2 rounded-lg border border-slate-200 dark:border-white/10
+                               text-slate-600 dark:text-slate-300 text-sm font-medium
+                               hover:bg-slate-50 dark:hover:bg-white/5
+                               cursor-pointer transition-colors duration-150"
+                        data-i18n="cancel">取消</button>
+                <button id="vendor-modal-save"
+                        class="px-4 py-2 rounded-lg bg-primary-500 hover:bg-primary-600 text-white text-sm font-medium
+                               cursor-pointer transition-colors duration-150 disabled:opacity-50 disabled:cursor-not-allowed"
+                        data-i18n="save">保存</button>
+            </div>
+        </div>
+    </div>
+
     <script defer src="assets/js/console.js"></script>
 </body>
 </html>

channel/web/static/css/console.css

@@ -725,6 +725,58 @@
     background: rgba(74, 190, 110, 0.15);
     color: #74E9A4;
 }
+/* When an item carries a hint (e.g. brand alias next to a technical model
+   id), label/hint are split into two spans so the hint sits on the right in
+   a dim, smaller weight. Without a hint the row stays a plain text node and
+   uses the default ellipsis behaviour, so no layout regressions for old call
+   sites. */
+.cfg-dropdown-label {
+    flex: 1 1 auto;
+    min-width: 0;
+    overflow: hidden;
+    text-overflow: ellipsis;
+}
+.cfg-dropdown-hint {
+    flex-shrink: 0;
+    margin-left: auto;
+    padding-left: 12px;
+    color: #94a3b8;
+    font-size: 12px;
+    font-weight: 400;
+}
+.dark .cfg-dropdown-hint {
+    color: #64748b;
+}
+.cfg-dropdown-item.active .cfg-dropdown-hint {
+    /* Tint the hint toward the brand colour on the active row so it doesn't
+       fight with the highlighted label tone. */
+    color: rgba(34, 133, 71, 0.65);
+}
+.dark .cfg-dropdown-item.active .cfg-dropdown-hint {
+    color: rgba(116, 233, 164, 0.6);
+}
+/* The active row gets a trailing brand-green checkmark via a Font Awesome
+   pseudo-element so every dropdown (chat / vision / image / asr / tts / etc.)
+   surfaces "this is what's currently selected" without per-call JS plumbing.
+   When a hint is present, the ✓ sits to its right with a small gap; without
+   a hint, margin-left:auto pushes the ✓ flush against the right edge. */
+.cfg-dropdown-item.active::after {
+    content: '\f00c';                  /* FontAwesome check glyph */
+    font-family: 'Font Awesome 6 Free', 'Font Awesome 5 Free', 'FontAwesome';
+    font-weight: 900;
+    margin-left: auto;
+    padding-left: 12px;
+    color: #4abe6e;
+    font-size: 11px;
+    flex-shrink: 0;
+}
+.cfg-dropdown-item.active:has(.cfg-dropdown-hint)::after {
+    /* When hint occupies the auto-margin slot, the ✓ no longer benefits
+       from `margin-left: auto`; replace it with a small fixed gap so the
+       ✓ trails the hint cleanly. */
+    margin-left: 0;
+    padding-left: 10px;
+}
 
 /* API Key masking via CSS (avoids browser password prompts) */
 .cfg-key-masked {
@@ -732,6 +784,77 @@
     text-security: disc;
 }
 
+/* Provider logo image — vendors flagged as `provider-logo-invert-dark`
+   ship a black wordmark that disappears on the dark canvas; we invert their
+   luminance only in dark mode so the brand stays recognizable without
+   touching multi-color marks like Google/MiniMax. */
+.provider-logo-img {
+    object-fit: contain;
+    object-position: center;
+}
+.dark .provider-logo-invert-dark {
+    filter: invert(1) brightness(1.15);
+}
+
+/* Models page — provider dropdown rows.
+   Configured rows look like ordinary picker entries; the .active row's
+   trailing brand-green ✓ already announces "this is what's selected"
+   (handled globally by .cfg-dropdown-item.active::after above).
+   Unconfigured rows are visually subdued and carry a trailing gear icon
+   as a "click to set up" affordance. */
+.cap-provider-label {
+    flex: 1 1 auto;
+    overflow: hidden;
+    text-overflow: ellipsis;
+}
+.cap-provider-gear {
+    margin-left: auto;
+    padding-left: 12px;
+    color: #94a3b8;
+    font-size: 11px;
+    flex-shrink: 0;
+}
+.cap-provider-item.cap-provider-unconfigured {
+    color: #94a3b8;
+}
+.dark .cap-provider-item.cap-provider-unconfigured {
+    color: #64748b;
+}
+.cap-provider-item.cap-provider-unconfigured:hover {
+    color: #475569;
+}
+.dark .cap-provider-item.cap-provider-unconfigured:hover {
+    color: #cbd5e1;
+}
+.cap-provider-item.cap-provider-unconfigured:hover .cap-provider-gear {
+    color: #475569;
+}
+.dark .cap-provider-item.cap-provider-unconfigured:hover .cap-provider-gear {
+    color: #cbd5e1;
+}
+/* If the active row ever lands on an unconfigured vendor (defensive — the
+   click handler normally diverts to the modal), suppress the global ✓ so
+   the gear remains the sole trailing icon and the row keeps reading as
+   "needs setup" rather than "already selected". */
+.cap-provider-item.cap-provider-unconfigured.active::after {
+    content: none;
+}
+
+/* "Add vendor" modal picker — each configured row carries a static
+   brand-green ✓ via decorateVendorModalPicker so users can see what's set
+   up at a glance. The active row's global ✓ is suppressed here to avoid
+   showing two checks side by side on configured + selected rows. */
+.vendor-picker-item.active::after {
+    content: none;
+}
+.vendor-picker-configured-mark {
+    margin-left: auto;
+    padding-left: 12px;
+    color: #4abe6e;
+    font-size: 11px;
+    flex-shrink: 0;
+}
+
 /* Chat Input */
 #chat-input {
     resize: none; height: 42px; max-height: 180px;
@@ -1171,3 +1294,76 @@
     overflow: hidden;
     min-height: 2.5em;  /* ~2 lines at text-sm leading-relaxed */
 }
+
+/* --------------------------------------------------------------------
+ * Voice pill — compact custom audio player used by mic uploads and TTS
+ * replies. Replaces the bulky native <audio controls> with a play/pause
+ * icon + thin progress bar + duration counter so it blends into chat
+ * bubbles without the chrome-grey browser default look.
+ * ------------------------------------------------------------------ */
+.voice-pill {
+    display: inline-flex;
+    align-items: center;
+    gap: 8px;
+    padding: 6px 10px;
+    border-radius: 999px;
+    background: rgba(15, 23, 42, 0.05);
+    color: rgb(71, 85, 105);
+    font-size: 12px;
+    line-height: 1;
+    max-width: 240px;
+    user-select: none;
+    cursor: default;
+}
+.dark .voice-pill {
+    background: rgba(255, 255, 255, 0.08);
+    color: rgb(203, 213, 225);
+}
+.voice-pill[data-loading="1"] {
+    opacity: 0.65;
+}
+.voice-pill-btn {
+    width: 22px;
+    height: 22px;
+    border-radius: 999px;
+    display: inline-flex;
+    align-items: center;
+    justify-content: center;
+    background: var(--color-primary-500, #2563eb);
+    color: #fff;
+    flex-shrink: 0;
+    cursor: pointer;
+    transition: transform 0.1s ease;
+}
+.voice-pill-btn:hover { transform: scale(1.05); }
+.voice-pill-btn i { font-size: 9px; margin-left: 1px; }
+.voice-pill-btn[data-state="play"] i { margin-left: 2px; }
+.voice-pill-btn[data-state="pause"] i { margin-left: 0; }
+.voice-pill-track {
+    flex: 1;
+    height: 3px;
+    border-radius: 999px;
+    background: rgba(100, 116, 139, 0.25);
+    overflow: hidden;
+    min-width: 70px;
+}
+.dark .voice-pill-track {
+    background: rgba(148, 163, 184, 0.25);
+}
+.voice-pill-fill {
+    height: 100%;
+    width: 0%;
+    background: var(--color-primary-500, #2563eb);
+    border-radius: inherit;
+    transition: width 0.1s linear;
+}
+.voice-pill-time {
+    font-variant-numeric: tabular-nums;
+    font-size: 11px;
+    color: inherit;
+    opacity: 0.75;
+    flex-shrink: 0;
+    min-width: 28px;
+    text-align: right;
+}
+.voice-pill audio { display: none; }

channel/web/static/logos/claudeAPI.svg

@@ -0,0 +1 @@
+<?xml version="1.0" standalone="no"?><!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd"><svg t="1779251656961" class="icon" viewBox="0 0 1024 1024" version="1.1" xmlns="http://www.w3.org/2000/svg" p-id="18432" xmlns:xlink="http://www.w3.org/1999/xlink" width="200" height="200"><path d="M252.8 652.8l167.893333-94.293333 2.773334-8.106667-2.773334-4.48h-8.106666l-28.16-1.706667-96-2.56-83.2-3.413333-80.64-4.266667-20.266667-4.266666L85.333333 504.746667l1.92-12.586667 17.066667-11.52 24.32 2.133333 53.973333 3.626667 81.066667 5.546667 58.666667 3.413333 87.04 9.173333h13.866666l1.92-5.546666-4.693333-3.413334-3.626667-3.413333-83.84-56.746667-90.666666-60.16-47.573334-34.56-25.813333-17.493333-13.013333-16.426667-5.546667-35.84 23.253333-25.813333 31.36 2.133333 7.893334 2.133334 31.786666 24.32 67.84 52.48L401.066667 391.466667l13.013333 10.88 5.12-3.626667 0.64-2.56-5.76-9.813333-48.213333-87.04L314.453333 210.773333l-22.826666-36.693333-5.973334-21.973333a107.861333 107.861333 0 0 1-3.626666-26.026667l26.666666-36.053333L323.413333 85.333333l35.413334 4.693334 14.933333 13.013333 21.973333 50.346667 35.626667 79.36 55.253333 107.733333 16.213334 32 8.746666 29.653333 3.2 9.173334h5.546667v-5.12l4.48-60.8 8.32-74.453334 8.106667-96 2.773333-27.093333 13.44-32.426667 26.666667-17.493333 20.693333 10.026667 17.066667 24.32-2.346667 15.786666-10.24 65.92-19.84 103.253334-13.013333 69.12h7.466666l8.746667-8.746667 34.986667-46.506667 58.666666-73.386666 26.026667-29.226667 30.293333-32.213333 19.413334-15.36h36.693333l27.093333 40.106666-12.16 41.386667-37.76 48-31.36 40.533333-45.013333 60.586667-28.16 48.426667 2.56 3.84 6.613333-0.64 101.546667-21.546667 54.826667-10.026667 65.493333-11.306666 29.653333 13.866666 3.2 14.08-11.733333 28.8-69.973333 17.28-82.133334 16.426667-122.24 29.013333-1.493333 1.066667 1.706667 2.133333 55.04 5.12 23.466666 1.28h57.6l107.306667 7.893334 28.16 18.56 16.853333 22.613333-2.773333 17.28-43.306667 21.973333-58.24-13.866666-136.106666-32.426667-46.72-11.733333h-6.4v3.84l38.826666 37.973333 71.253334 64.426667 89.173333 82.986666 4.48 20.48-11.52 16.213334-12.16-1.706667-78.506667-58.88-30.293333-26.666667-68.48-57.6h-4.48v5.973334l15.786667 23.04 83.413333 125.226666 4.266667 38.4-5.973334 12.586667-21.546666 7.466667-23.68-4.266667-48.853334-68.48-50.346666-77.226667-40.533334-69.12-4.906666 2.773334-23.893334 258.133333-11.306666 13.226667-26.026667 10.026666-21.546667-16.426666-11.52-26.666667 11.52-52.48 13.866667-68.48 11.306667-54.4 10.24-67.626667 5.973333-22.4-0.426667-1.493333-4.906666 0.64-50.986667 69.973333-77.653333 104.746667-61.44 65.706667-14.72 5.76-25.386667-13.226667 2.346667-23.466667 14.293333-20.906666 84.906667-107.946667 51.2-66.986667 33.066666-38.613333v-5.546667h-2.133333l-225.493333 146.56-40.106667 5.12-17.28-16.213333 2.133333-26.666667 8.106667-8.746666 67.84-46.72h-0.213333l0.853333 0.853333z" fill="#D97757" p-id="18433"></path></svg>

channel/web/static/logos/custom.svg

@@ -0,0 +1,10 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" width="200" height="200" fill="none" stroke="#475569" stroke-width="1.8" stroke-linecap="round" stroke-linejoin="round">
+  <!-- Horizontal slider tracks -->
+  <line x1="4" y1="7" x2="20" y2="7"/>
+  <line x1="4" y1="12" x2="20" y2="12"/>
+  <line x1="4" y1="17" x2="20" y2="17"/>
+  <!-- Knobs (filled circles) -->
+  <circle cx="9" cy="7"  r="2.2" fill="#475569" stroke="none"/>
+  <circle cx="15" cy="12" r="2.2" fill="#475569" stroke="none"/>
+  <circle cx="7" cy="17"  r="2.2" fill="#475569" stroke="none"/>
+</svg>

channel/web/static/logos/dashscope.svg

@@ -0,0 +1 @@
+<?xml version="1.0" standalone="no"?><!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd"><svg t="1779251621200" class="icon" viewBox="0 0 1024 1024" version="1.1" xmlns="http://www.w3.org/2000/svg" p-id="17444" xmlns:xlink="http://www.w3.org/1999/xlink" width="200" height="200"><path d="M1019.364785 620.816931L891.797142 397.807295 946.450846 293.15069a29.097778 29.097778 0 0 0 6.399732-36.393472l-70.184053-126.586684a30.078737 30.078737 0 0 0-24.574968-13.652427H597.4945L539.171949 14.549389a27.348852 27.348852 0 0 0-20.906122-14.549389H380.628607a29.139776 29.139776 0 0 0-24.616967 14.549389v5.545767L225.797108 243.062793H100.919352a29.182775 29.182775 0 0 0-25.513928 13.653427L3.428446 384.11187a32.766624 32.766624 0 0 0 0 29.182775L132.831012 638.096205 74.508461 740.064923a32.766624 32.766624 0 0 0 0 29.05478l66.514207 116.561105a29.905744 29.905744 0 0 0 25.513929 14.505391H427.132654l62.845361 109.222414A30.078737 30.078737 0 0 0 512.762058 1024H660.382859a29.139776 29.139776 0 0 0 24.574968-14.549389l128.463606-224.843558h114.76818a31.91366 31.91366 0 0 0 24.660965-15.444352l66.471208-117.414069a28.158818 28.158818 0 0 0 0-30.9747l0.042999 0.042999z m-161.273228 14.591387L791.57735 512.490479 518.265827 993.964261l-74.748861-122.87484h-273.268525l65.618244-119.205994h139.386147L101.856313 272.244568h143.055993L380.671605 30.121735l68.34913 119.247993-70.184053 122.87484H925.501726l-69.202094 121.936879 137.594222 241.183873H858.134555z" fill="#605BEC" p-id="17445"></path><path d="M499.962596 699.320634l174.371677-274.719464H324.694955z" fill="#605BEC" p-id="17446"></path></svg>

channel/web/static/logos/doubao.svg

@@ -0,0 +1 @@
+<?xml version="1.0" standalone="no"?><!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd"><svg t="1779261485522" class="icon" viewBox="0 0 1024 1024" version="1.1" xmlns="http://www.w3.org/2000/svg" p-id="5381" xmlns:xlink="http://www.w3.org/1999/xlink" width="200" height="200"><path d="M958.976 439.808C804.864 336.896 642.56 321.536 642.56 321.536s8.192 235.008-10.752 306.176c-0.512 9.728-11.776 75.264-43.008 157.696-10.752 28.16-24.064 55.296-39.424 81.408-40.96 74.24-89.6 127.488-89.6 127.488 119.808-48.64 205.312-92.672 309.76-175.616 122.88-96.768 229.376-254.464 189.44-378.88z" fill="#37E1BE" p-id="5382"></path><path d="M329.728 395.776c158.208-100.864 308.736-78.848 312.32-74.752 0.512 0.512 1.024 0.512 1.024 0.512 0-14.336-6.656-60.928-13.312-106.496-11.776-60.928-22.528-124.928-23.04-133.632-170.496-139.264-356.864-78.336-448 25.6-61.44 70.144-103.424 169.984-102.4 224.256V762.88c0.512-12.8 1.536-20.48 2.048-20.48 17.92-197.12 271.36-346.624 271.36-346.624z" fill="#A569FF" p-id="5383"></path><path d="M792.064 272.384c-41.984-43.52-87.552-88.576-122.368-125.44-33.28-34.816-59.392-60.928-62.976-65.536 0.512 8.704 11.264 72.704 23.04 133.632 6.656 45.568 12.8 92.672 13.312 106.496 0 0 162.304 15.36 316.416 118.272-0.512 0-83.456-80.384-167.424-167.424zM549.888 866.816c-2.56 1.024-198.656 107.008-292.352-30.72-20.992-30.72-31.744-68.096-33.28-106.496-3.072-74.752 5.12-227.84 105.472-333.824 0 0-253.44 149.504-270.848 346.624-0.512 0.512-2.048 8.192-2.048 20.48-1.024 32.768 4.608 98.304 43.008 155.136 52.224 78.336 193.024 138.752 328.192 85.504l33.28-9.728c-1.024 0.512 47.616-52.224 88.576-126.976z" fill="#1E37FC" p-id="5384"></path></svg>

channel/web/static/logos/gemini.svg

@@ -0,0 +1 @@
+<?xml version="1.0" standalone="no"?><!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd"><svg t="1779251750646" class="icon" viewBox="0 0 1024 1024" version="1.1" xmlns="http://www.w3.org/2000/svg" p-id="29551" xmlns:xlink="http://www.w3.org/1999/xlink" width="200" height="200"><path d="M214.101333 512c0-32.512 5.546667-63.701333 15.36-92.928L57.173333 290.218667A491.861333 491.861333 0 0 0 4.693333 512c0 79.701333 18.858667 154.88 52.394667 221.610667l172.202667-129.066667A290.56 290.56 0 0 1 214.101333 512" fill="#FBBC05" p-id="29552"></path><path d="M516.693333 216.192c72.106667 0 137.258667 25.002667 188.458667 65.962667L854.101333 136.533333C763.349333 59.178667 646.997333 11.392 516.693333 11.392c-202.325333 0-376.234667 113.28-459.52 278.826667l172.373334 128.853333c39.68-118.016 152.832-202.88 287.146666-202.88" fill="#EA4335" p-id="29553"></path><path d="M516.693333 807.808c-134.357333 0-247.509333-84.864-287.232-202.88l-172.288 128.853333c83.242667 165.546667 257.152 278.826667 459.52 278.826667 124.842667 0 244.053333-43.392 333.568-124.757333l-163.584-123.818667c-46.122667 28.458667-104.234667 43.776-170.026666 43.776" fill="#34A853" p-id="29554"></path><path d="M1005.397333 512c0-29.568-4.693333-61.44-11.648-91.008H516.650667V614.4h274.602666c-13.696 65.962667-51.072 116.650667-104.533333 149.632l163.541333 123.818667c93.994667-85.418667 155.136-212.650667 155.136-375.850667" fill="#4285F4" p-id="29555"></path></svg>

channel/web/static/logos/minimax.svg

@@ -0,0 +1 @@
+<?xml version="1.0" standalone="no"?><!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd"><svg t="1779251514432" class="icon" viewBox="0 0 1024 1024" version="1.1" xmlns="http://www.w3.org/2000/svg" p-id="11888" xmlns:xlink="http://www.w3.org/1999/xlink" width="200" height="200"><path d="M415.392 475.808v329.984c-22.304 111.744-170.56 82.944-171.2 1.92-0.672-101.824 0-202.976 0-304.064v-117.184c0-14.656-3.2-26.24-16-35.392-24.96-18.72-54.944 3.264-55.584 30.208-1.408 36.16-0.704 71.616-1.408 107.264 0 28.16 0 55.52 0.64 83.648-18.368 123.776-168.32 103.232-171.808 0.704V487.04c0-28.032 54.944-34.624 52.256 7.36-1.792 20.8-0.64 42.272-1.344 62.912-0.64 36.8 55.648 61.6 68.896 1.408 0.64-49.632 0.64-99.264 0.64-149.344 0-62.752 17.824-113.856 84.352-118.624 28.8-2.56 47.968 9.504 66.336 30.304 7.04 7.36 23.68 30.72 24.32 56.16 0 23.456 0.64 46.752 0.64 70.464 0 46.72-0.64 93.76-0.64 140.48 0 30.304 0.64 60.256 0.64 89.856 0 37.536 0 75.552-0.64 113.152-0.64 48.864 58.816 48.16 68.352-0.768 0-57.632 0.64-114.56 0.64-172.192 0-141.984-0.64-283.968-0.64-425.856 0-14.72-2.048-55.584 5.76-70.464 41.504-101.12 167.392-56.96 168.544 26.72 2.432 171.52 0 344.896 0.64 516.8 0 59.616-48.416 46.816-51.104 23.488 0-178.88 0-358.4 0.64-537.024-2.368-44.832-68.832-38.72-72.672-6.592-1.28 36.864-0.64 74.4-1.28 111.232v219.008h0.64l0.448 0.256h-0.064z" fill="#D4367A" p-id="11889"></path><path d="M610.016 473.184v242.336V143.648c21.632-112.512 169.824-83.264 170.464-2.176 0.704 101.12 0 202.912 0.704 304 0 38.784 0 77.728-0.64 116.544 0 15.36 3.776 26.176 16.64 36.032 24.32 18.24 54.24-3.2 55.584-30.592 1.344-35.488 0.64-70.976 0.64-107.328V376.96c18.56-123.776 168.128-103.232 171.264-0.704v310.592c0 28.16-54.304 34.848-51.872-7.296 1.472-21.44 0-267.104 0.768-288.64 1.28-36.16-55.712-61.664-68.928-0.768v148.576c0 63.68-17.856 113.92-84.96 119.36-63.264 1.504-88.704-42.24-90.752-86.432V271.328c0-38.24 0-75.552 0.64-113.088 0.64-48.864-58.784-48.864-68.896 0.704V831.36c0 14.592 2.048 55.52-5.184 70.432-41.44 101.056-168 56.864-169.152-26.752v-79.616c3.136-53.6 48.416-40.864 50.464-18.176v94.464c2.432 44.928 68.928 39.488 72.064 6.656 1.344-36.896 1.344-73.728 1.344-111.296v-293.824h-0.192v-0.064z" fill="#ED6D48" p-id="11890"></path></svg>

channel/web/static/logos/moonshot.svg

@@ -0,0 +1 @@
+<?xml version="1.0" standalone="no"?><!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd"><svg t="1779251592968" class="icon" viewBox="0 0 1024 1024" version="1.1" xmlns="http://www.w3.org/2000/svg" p-id="16416" xmlns:xlink="http://www.w3.org/1999/xlink" width="200" height="200"><path d="M117.9648 684.6464l342.30272 93.57312v75.34592l209.7152 58.5728A428.99456 428.99456 0 0 1 512 942.08c-176.128 0-327.53664-105.8816-394.0352-257.4336zM83.29216 477.42976l407.30624 112.64-9.6256 37.00736-6.0416 35.0208 383.3856 104.96a432.5376 432.5376 0 0 1-65.10592 70.32832l-688.18944-185.9584A429.4656 429.4656 0 0 1 81.92 512c0-11.63264 0.47104-23.1424 1.37216-34.54976z m57.344-182.4768l429.07648 114.21696a279.94112 279.94112 0 0 0-23.06048 35.55328 201.17504 201.17504 0 0 0-14.70464 34.93888l403.08736 110.26432a426.8032 426.8032 0 0 1-23.552 81.7152L86.54848 448.7168a427.25376 427.25376 0 0 1 54.0672-153.76384z m158.47424-156.75392l404.23424 108.31872a190.2592 190.2592 0 0 0-32.80896 24.90368c-9.13408 8.8064-19.8656 21.4016-32.1536 37.74464l285.24544 77.78304c9.216 30.45376 15.03232 61.8496 17.32608 93.5936L156.61056 269.68064a432.27136 432.27136 0 0 1 142.49984-131.4816zM512 81.92c142.90944 0 269.55776 69.71392 347.7504 176.98816L337.26464 118.90688A428.50304 428.50304 0 0 1 512 81.92z" fill="#000000" p-id="16417"></path></svg>

channel/web/static/logos/openai.svg

@@ -0,0 +1 @@
+<?xml version="1.0" standalone="no"?><!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd"><svg t="1779251225589" class="icon" viewBox="0 0 1024 1024" version="1.1" xmlns="http://www.w3.org/2000/svg" p-id="9015" xmlns:xlink="http://www.w3.org/1999/xlink" width="200" height="200"><path d="M881.664 431.488a218.88 218.88 0 0 0-18.176-177.088A218.624 218.624 0 0 0 628.992 149.76c-40.576-45.824-100.288-71.424-162.176-71.424a219.136 219.136 0 0 0-208 150.4 215.68 215.68 0 0 0-144 104.512 218.944 218.944 0 0 0 26.688 254.912 218.752 218.752 0 0 0 19.2 177.152 217.088 217.088 0 0 0 234.624 104.512 219.136 219.136 0 0 0 162.112 72.512 219.136 219.136 0 0 0 208-150.4 215.68 215.68 0 0 0 144-104.512 219.008 219.008 0 0 0-27.712-256z m-324.288 454.4a158.08 158.08 0 0 1-103.424-37.376c1.088-1.088 4.288-2.176 5.376-3.2l171.712-99.2a28.16 28.16 0 0 0 13.824-24.512V479.488l72.576 41.6c1.024 0 1.024 1.024 1.024 2.112v200.512a160.512 160.512 0 0 1-161.088 162.112z m-347.712-148.288c-19.2-33.088-25.6-71.488-19.2-108.8 1.088 1.024 3.2 2.176 5.376 3.2l171.712 99.2a25.984 25.984 0 0 0 27.712 0l210.112-121.6v84.224c0 1.152 0 2.176-1.024 2.176L430.464 796.16c-76.8 44.8-176 18.176-220.8-58.624z m-44.736-375.424c19.2-32.64 48.896-57.856 84.224-71.488v204.8c0 9.6 5.376 19.2 13.888 24.512l210.176 121.6-72.576 41.6c-1.024 0-2.112 1.088-2.112 0L224.64 582.912a160.448 160.448 0 0 1-59.776-220.8h0.064z m597.312 138.688l-210.112-121.6 72.512-41.6c1.088 0 2.176-1.088 2.176 0l173.824 100.224a161.088 161.088 0 0 1-25.6 291.2V525.44a26.304 26.304 0 0 0-12.8-24.512z m71.488-108.8a23.232 23.232 0 0 0-5.312-3.2L656.64 289.536a26.048 26.048 0 0 0-27.712 0l-210.176 121.6V326.912c0-1.088 0-2.176 1.088-2.176l173.824-100.224a161.152 161.152 0 0 1 220.8 59.712c19.2 32 25.6 70.4 19.2 107.776z m-454.4 149.248l-72.64-41.6c-1.024 0-1.024-1.088-1.024-2.176V297.088A162.048 162.048 0 0 1 467.84 135.04a158.08 158.08 0 0 1 103.424 37.312 22.848 22.848 0 0 1-5.312 3.2L394.24 274.688a28.16 28.16 0 0 0-13.888 24.512v242.112h-1.088z m39.424-85.312l93.824-54.4 93.888 54.4v107.712l-93.888 54.4-93.824-54.4V456z" fill="#000000" p-id="9016"></path></svg>

channel/web/static/logos/qianfan.svg

@@ -0,0 +1 @@
+<?xml version="1.0" standalone="no"?><!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd"><svg t="1779251568791" class="icon" viewBox="0 0 1024 1024" version="1.1" xmlns="http://www.w3.org/2000/svg" p-id="14450" xmlns:xlink="http://www.w3.org/1999/xlink" width="200" height="200"><path d="M96.20121136 636.3124965c-0.1472897-113.41305959-0.29457937-226.8261192-0.29457937-340.23917879 0-14.87625845 7.65906378-26.51214381 20.4732666-34.02391789 45.51251353-26.65943349 91.02502705-53.31886698 136.83211997-79.53643141 71.1409192-40.94653321 142.42912809-81.59848704 213.71733698-122.39773055 7.36448439-4.12411126 14.58167909-8.3955122 21.50429441-13.2560719 19.44223878-13.40336159 39.03176725-16.05457598 60.09419263-3.53495252 27.39588193 16.34915535 54.93905355 32.25644163 82.48222516 48.16372793 88.0792333 50.96223197 176.30575629 101.77717426 264.38498958 152.59211653 9.86840908 5.74429781 19.88410785 11.19401627 29.60522725 17.0856038 14.13981003 8.54280189 21.50429441 21.06242535 21.50429443 37.70616007 0 147.73155685 0.29457937 295.46311371-0.1472897 443.19467057 0 15.46541722-7.2171947 28.57419943-21.7988738 36.96971163-34.7603663 20.17868721-70.55176044 38.88447758-104.57567833 59.94690293-48.90017634 30.19438599-100.00969801 56.11737105-148.76258466 86.60633642-29.01606849 18.11663161-59.50503387 34.02391789-89.11026112 50.96223197-13.10878221 7.51177407-26.07027474 15.17083783-39.03176726 22.9771913-13.84523065 8.3955122-27.83775099 8.83738127-41.97756102 0.73644843-56.41195043-32.55102101-112.82390085-65.10204201-169.38314098-97.653063-61.86166887-35.64410444-123.72333775-71.1409192-185.4377169-106.78502365-11.19401627-6.48074626-22.24074286-12.81420285-32.99289009-19.88410785-11.48859565-7.65906378-17.08560379-19.14765941-17.08560378-32.69831069-0.1472897-34.7603663 0.1472897-69.52073264 0.29457938-104.28109895 1.62018657-0.58915875 1.62018657-1.62018657-0.29457938-2.65121438z m356.58833414-225.500512c2.20934532-1.76747625 4.41869063-3.68224221 6.77532565-5.15513907 68.93157389-39.62092601 137.86314777-79.24185204 206.94201135-118.86277807 2.79850407-1.62018657 6.48074626-1.62018657 6.62803594-6.18616688 0.1472897-4.8605597-4.12411126-4.71327001-6.77532564-6.18616688-40.65195383-23.56635005-81.59848704-46.83812071-122.10315117-70.84633984-16.79102442-10.01569877-32.84560039-8.54280189-48.45830728 0.58915876-45.9543826 26.51214381-91.46689612 53.61344636-137.27398903 80.42016953-31.96186226 18.70579035-64.21830387 37.11700133-96.32745581 55.67550198-18.41121097 10.60485751-27.54317163 25.33382629-27.24859225 47.72185885 0.88373813 89.55213018 0.58915875 179.10426036 0.14728969 268.65639053-0.1472897 20.17868721 9.27925033 33.58204881 25.33382629 43.15587853 31.3727035 18.70579035 63.18727606 37.11700133 95.14913832 54.93905355 10.89943689 6.03887719 21.06242535 13.99252034 35.79139414 18.41121096V505.51925374c6.48074626 19.58952848 18.55850066 34.02391789 36.67513226 44.6287754 27.83775099 16.20186565 63.18727606 12.51962347 86.31175705-10.45756784 26.95401286-26.65943349 28.72148912-62.89269668 12.81420282-90.14128893-16.34915535-28.42690974-43.59774757-37.55887038-74.38129233-38.73718787z m82.48222517 429.64401928c14.28709972-3.82953187 25.92298506-13.99252034 38.88447758-21.35700473 40.94653321-23.27177067 81.30390766-47.72185885 122.54502023-70.55176046 26.95401286-15.02354815 52.87699792-31.66728287 80.71474891-45.21793415 16.79102442-8.10093283 29.60522723-22.53532223 29.60522726-43.4504579 0.1472897-92.939793 0.29457937-185.73229631 0.14728969-278.6720893 0-11.19401627-5.15513907-13.99252034-13.84523067-7.06990501-26.51214381 20.76784598-57.29568854 34.46578693-86.16446735 51.25681135-54.49718448 31.81457257-109.14165865 63.33456576-163.78613282 95.00184862-8.54280189 4.8605597-11.78317502 10.45756784-11.63588535 20.47326662 0.29457937 96.18016613 0.1472897 192.50762194 0.1472897 288.68778806-0.29457937 3.5349525-1.47289687 7.65906378 3.38766282 10.8994369z" fill="#066AF3" p-id="14451"></path><path d="M96.20121136 636.3124965c1.91476594 1.03102783 1.91476594 2.06205563 0 3.09308345v-3.09308345z" fill="#4372E0" p-id="14452"></path><path d="M391.3697457 505.37196405c-5.44971845-44.33419602 13.84523065-74.08671296 61.4197998-94.55997955 30.93083443 1.17831749 58.03213699 10.31027814 74.38129233 38.5898982 15.75999659 27.39588193 14.13981003 63.48185543-12.81420282 90.14128893-23.27177067 22.97719129-58.47400606 26.65943349-86.31175705 10.45756783-18.11663161-10.60485751-30.34167568-25.03924691-36.67513226-44.62877541z" fill="#002A9A" p-id="14453"></path></svg>

channel/web/static/logos/zhipu.svg

@@ -0,0 +1 @@
+<?xml version="1.0" standalone="no"?><!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd"><svg t="1779251419020" class="icon" viewBox="0 0 1024 1024" version="1.1" xmlns="http://www.w3.org/2000/svg" p-id="10062" xmlns:xlink="http://www.w3.org/1999/xlink" width="200" height="200"><path d="M520.063496 0v77.563152c0 269.231173-144.758953 414.054122-434.212862 434.340854L86.106618 511.968002H76.827198V255.984001l443.236298-255.984001z" fill="#5B55F6" p-id="10063"></path><path d="M520.063496 1023.936004v-77.563152c0-269.231173-144.758953-414.054122-434.212862-434.340854L86.042622 511.968002H76.827198v255.984001l443.236298 255.984001z" fill="#376AF3" p-id="10064"></path><path d="M520.063496 0v77.563152c0 269.231173 144.758953 414.054122 434.276858 434.340854L954.08437 511.968002h9.215424V255.984001L520.063496 0z" fill="#5B55F6" p-id="10065"></path><path d="M520.063496 1023.936004v-77.563152c0-269.231173 144.758953-414.054122 434.276858-434.340854L954.08437 511.968002h9.27942v255.984001l-443.236298 255.984001z" fill="#376AF3" p-id="10066"></path></svg>

channel/wecom_bot/wecom_bot_channel.py

@@ -81,6 +81,8 @@ def _loads_wecom_ws_json(raw):
 @singleton
 class WecomBotChannel(ChatChannel):
 
+    NOT_SUPPORT_REPLYTYPE = []
+
     def __init__(self):
         super().__init__()
         self.bot_id = ""
@@ -472,6 +474,8 @@ class WecomBotChannel(ChatChannel):
             else:
                 context.type = ContextType.TEXT
             context.content = content.strip()
+            if "desire_rtype" not in context and conf().get("always_reply_voice"):
+                context["desire_rtype"] = ReplyType.VOICE
 
         return context
 
@@ -498,6 +502,8 @@ class WecomBotChannel(ChatChannel):
             self._send_file(reply.content, receiver, is_group, req_id)
         elif reply.type == ReplyType.VIDEO or reply.type == ReplyType.VIDEO_URL:
             self._send_file(reply.content, receiver, is_group, req_id, media_type="video")
+        elif reply.type == ReplyType.VOICE:
+            self._send_voice(reply.content, receiver, is_group, req_id)
         else:
             logger.warning(f"[WecomBot] Unsupported reply type: {reply.type}, falling back to text")
             self._send_text(str(reply.content), receiver, is_group, req_id)
@@ -730,6 +736,65 @@ class WecomBotChannel(ChatChannel):
                 },
             })
 
+    def _send_voice(self, voice_path: str, receiver: str, is_group: bool, req_id: str = None):
+        """Send native voice reply. WeCom voice media must be amr."""
+        local_path = voice_path
+        if local_path.startswith("file://"):
+            local_path = local_path[7:]
+
+        if local_path.startswith(("http://", "https://")):
+            try:
+                resp = requests.get(local_path, timeout=60)
+                resp.raise_for_status()
+                ext = os.path.splitext(local_path)[1] or ".mp3"
+                tmp_path = f"/tmp/wecom_voice_{uuid.uuid4().hex[:8]}{ext}"
+                with open(tmp_path, "wb") as f:
+                    f.write(resp.content)
+                local_path = tmp_path
+            except Exception as e:
+                logger.error(f"[WecomBot] Failed to download voice for sending: {e}")
+                return
+
+        if not os.path.exists(local_path):
+            logger.error(f"[WecomBot] Voice file not found: {local_path}")
+            return
+
+        amr_path = local_path
+        if not local_path.lower().endswith(".amr"):
+            try:
+                from voice.audio_convert import any_to_amr
+                amr_path = os.path.splitext(local_path)[0] + ".amr"
+                any_to_amr(local_path, amr_path)
+            except Exception as e:
+                logger.error(f"[WecomBot] Failed to convert voice to amr: {e}")
+                return
+
+        media_id = self._upload_media(amr_path, "voice")
+        if not media_id:
+            logger.error("[WecomBot] Failed to upload voice media")
+            return
+
+        if req_id:
+            self._ws_send({
+                "cmd": "aibot_respond_msg",
+                "headers": {"req_id": req_id},
+                "body": {
+                    "msgtype": "voice",
+                    "voice": {"media_id": media_id},
+                },
+            })
+        else:
+            self._ws_send({
+                "cmd": "aibot_send_msg",
+                "headers": {"req_id": self._gen_req_id()},
+                "body": {
+                    "chatid": receiver,
+                    "chat_type": 2 if is_group else 1,
+                    "msgtype": "voice",
+                    "voice": {"media_id": media_id},
+                },
+            })
+
     def _active_send_markdown(self, content: str, receiver: str, is_group: bool):
         """Proactively send markdown message (for scheduled tasks, no req_id)."""
         self._ws_send({

channel/weixin/weixin_channel.py

@@ -60,6 +60,9 @@ def _save_credentials(cred_path: str, data: dict):
 @singleton
 class WeixinChannel(ChatChannel):
 
+    # ilink bot protocol has no outbound voice item; deliver TTS as a file.
+    NOT_SUPPORT_REPLYTYPE = []
+
     LOGIN_STATUS_IDLE = "idle"
     LOGIN_STATUS_WAITING = "waiting_scan"
     LOGIN_STATUS_SCANNED = "scanned"
@@ -464,6 +467,14 @@ class WeixinChannel(ChatChannel):
             else:
                 context.type = ContextType.TEXT
             context.content = content.strip()
+            if "desire_rtype" not in context and conf().get("always_reply_voice"):
+                context["desire_rtype"] = ReplyType.VOICE
+
+        elif ctype == ContextType.VOICE:
+            if "desire_rtype" not in context and (
+                conf().get("voice_reply_voice") or conf().get("always_reply_voice")
+            ):
+                context["desire_rtype"] = ReplyType.VOICE
 
         return context
 
@@ -486,6 +497,9 @@ class WeixinChannel(ChatChannel):
             self._send_file(reply.content, receiver, context_token)
         elif reply.type in (ReplyType.VIDEO, ReplyType.VIDEO_URL):
             self._send_video(reply.content, receiver, context_token)
+        elif reply.type == ReplyType.VOICE:
+            # ilink has no outbound voice item; deliver TTS as a file attachment.
+            self._send_file(reply.content, receiver, context_token)
         else:
             logger.warning(f"[Weixin] Unsupported reply type: {reply.type}, fallback to text")
             self._send_text(str(reply.content), receiver, context_token)

cli/VERSION

@@ -1 +1 @@
-2.0.8
+2.0.9

common/const.py

@@ -47,6 +47,7 @@ GEMINI_3_FLASH_PRE = "gemini-3-flash-preview"  # Gemini 3 Flash Preview - Agent
 GEMINI_3_PRO_PRE = "gemini-3-pro-preview"  # Gemini 3 Pro Preview
 GEMINI_31_PRO_PRE = "gemini-3.1-pro-preview"  # Gemini 3.1 Pro Preview - Agent推荐模型
 GEMINI_31_FLASH_LITE_PRE = "gemini-3.1-flash-lite-preview"  # Gemini 3.1 Flash Lite Preview - Agent推荐模型
+GEMINI_35_FLASH = "gemini-3.5-flash"  # Gemini 3.5 Flash - Agent推荐模型
 
 # OpenAI
 GPT35 = "gpt-3.5-turbo"
@@ -74,6 +75,7 @@ GPT_5_NANO = "gpt-5-nano"
 GPT_54 = "gpt-5.4"  # GPT-5.4 - Agent recommended model
 GPT_54_MINI = "gpt-5.4-mini"
 GPT_54_NANO = "gpt-5.4-nano"
+GPT_55 = "gpt-5.5"  # GPT-5.5 - top-tier (expensive), not default
 O1 = "o1-preview"
 O1_MINI = "o1-mini"
 WHISPER_1 = "whisper-1"
@@ -104,10 +106,12 @@ QWEN_LONG = "qwen-long"
 QWEN3_MAX = "qwen3-max"  # Qwen3 Max - Agent推荐模型
 QWEN35_PLUS = "qwen3.5-plus"  # Qwen3.5 Plus - Omni model (MultiModalConversation)
 QWEN36_PLUS = "qwen3.6-plus"  # Qwen3.6 Plus - Omni model (MultiModalConversation)
+QWEN37_MAX = "qwen3.7-max"  # Qwen3.7 Max - Agent推荐模型
 QWQ_PLUS = "qwq-plus"
 
 # MiniMax
 MINIMAX_M2_7 = "MiniMax-M2.7"  # MiniMax M2.7 - Latest
+MINIMAX_TEXT_01 = "MiniMax-Text-01"  # MiniMax 多模态 (vision)
 MINIMAX_M2_7_HIGHSPEED = "MiniMax-M2.7-highspeed"  # MiniMax M2.7 highspeed
 MINIMAX_M2_5 = "MiniMax-M2.5"  # MiniMax M2.5
 MINIMAX_M2_1 = "MiniMax-M2.1"  # MiniMax M2.1
@@ -119,6 +123,7 @@ MINIMAX_ABAB6_5 = "abab6.5-chat"  # MiniMax abab6.5
 GLM_5_1 = "glm-5.1"  # 智谱 GLM-5.1 - Agent recommended model (default)
 GLM_5_TURBO = "glm-5-turbo"  # 智谱 GLM-5-Turbo
 GLM_5 = "glm-5"  # 智谱 GLM-5
+GLM_5V_TURBO = "glm-5v-turbo"  # 智谱多模态 (vision)
 GLM_4 = "glm-4"
 GLM_4_PLUS = "glm-4-plus"
 GLM_4_flash = "glm-4-flash"
@@ -183,7 +188,7 @@ MODEL_LIST = [
               "claude", "claude-3-haiku", "claude-3-sonnet", "claude-3-opus", "claude-3.5-sonnet",
 
               # Gemini
-              GEMINI_31_FLASH_LITE_PRE, GEMINI_31_PRO_PRE, GEMINI_3_PRO_PRE, GEMINI_3_FLASH_PRE, GEMINI_25_PRO_PRE, GEMINI_25_FLASH_PRE,
+              GEMINI_35_FLASH, GEMINI_31_FLASH_LITE_PRE, GEMINI_31_PRO_PRE, GEMINI_3_PRO_PRE, GEMINI_3_FLASH_PRE, GEMINI_25_PRO_PRE, GEMINI_25_FLASH_PRE,
               GEMINI_20_FLASH, GEMINI_20_flash_exp, GEMINI_15_PRO, GEMINI_15_flash, GEMINI_PRO, GEMINI,
 
               # OpenAI
@@ -193,7 +198,7 @@ MODEL_LIST = [
               GPT_4o, GPT_4O_0806, GPT_4o_MINI,
               GPT_41, GPT_41_MINI, GPT_41_NANO,
               GPT_5, GPT_5_MINI, GPT_5_NANO,
-              GPT_54, GPT_54_MINI, GPT_54_NANO,
+              GPT_54, GPT_55, GPT_54_MINI, GPT_54_NANO,
               O1, O1_MINI,
 
               # GLM (智谱AI)
@@ -201,7 +206,7 @@ MODEL_LIST = [
               GLM_4_0520, GLM_4_AIR, GLM_4_AIRX, GLM_4_7,
 
               # Qwen (通义千问)
-              QWEN36_PLUS, QWEN35_PLUS, QWEN3_MAX, QWEN_MAX, QWEN_PLUS, QWEN_TURBO, QWEN_LONG,
+              QWEN37_MAX, QWEN36_PLUS, QWEN35_PLUS, QWEN3_MAX, QWEN_MAX, QWEN_PLUS, QWEN_TURBO, QWEN_LONG,
 
               # Doubao (豆包)
               DOUBAO, DOUBAO_SEED_2_CODE, DOUBAO_SEED_2_PRO, DOUBAO_SEED_2_LITE, DOUBAO_SEED_2_MINI,

config-template.json

@@ -16,8 +16,8 @@
   "open_ai_api_base": "https://api.openai.com/v1",
   "gemini_api_key": "",
   "gemini_api_base": "https://generativelanguage.googleapis.com",
-  "voice_to_text": "openai",
-  "text_to_voice": "openai",
+  "voice_to_text": "",
+  "text_to_voice": "",
   "voice_reply_voice": false,
   "speech_recognition": true,
   "group_speech_recognition": false,

config.py

@@ -330,8 +330,18 @@ def load_config():
     config_str = read_file(config_path)
     logger.debug("[INIT] config str: {}".format(drag_sensitive(config_str)))
 
-    # 将json字符串反序列化为dict类型
-    config = Config(json.loads(config_str))
+    # 将json字符串反序列化为dict类型。
+    # `object_pairs_hook` lets us catch users who accidentally typed the
+    # same key twice (e.g. two `"tools"` blocks) — json.loads would
+    # otherwise silently drop all but the last occurrence.
+    config = Config(json.loads(config_str, object_pairs_hook=_merge_duplicate_keys))
+
+    # Migrate legacy singular keys (`tool`, `skill`) into the canonical
+    # plural buckets so the rest of the codebase only reads one schema.
+    # Deep-merge so existing `tools`/`skills` entries are preserved and
+    # only missing namespaces are filled in from the legacy section.
+    _merge_legacy_namespace(config, legacy="tool",  canonical="tools")
+    _merge_legacy_namespace(config, legacy="skill", canonical="skills")
 
     # override config with environment variables.
     # Some online deployment platforms (e.g. Railway) deploy project from github directly. So you shouldn't put your secrets like api key in a config file, instead use environment variables to override the default config.
@@ -422,7 +432,7 @@ def load_config():
                 os.environ[env_key] = str(val)
                 injected += 1
 
-    injected += _sync_skill_config_to_env(config.get("skill", {}))
+    injected += _sync_skill_config_to_env(config.get("skills", {}))
 
     if injected:
         logger.info("[INIT] Synced {} config values to environment variables".format(injected))
@@ -430,11 +440,90 @@ def load_config():
     config.load_user_datas()
 
 
+def _deep_merge_dicts(base: dict, incoming: dict) -> dict:
+    """Recursively merge ``incoming`` into ``base`` (incoming wins on leaves)."""
+    for key, val in incoming.items():
+        if (
+            key in base
+            and isinstance(base[key], dict)
+            and isinstance(val, dict)
+        ):
+            _deep_merge_dicts(base[key], val)
+        else:
+            base[key] = val
+    return base
+
+
+def _merge_duplicate_keys(pairs):
+    """object_pairs_hook for json.loads: deep-merge duplicate top-level keys
+    (lists concat, dicts merge, scalars take the latter) instead of dropping."""
+    out = {}
+    duplicates = []
+    for key, val in pairs:
+        if key not in out:
+            out[key] = val
+            continue
+        duplicates.append(key)
+        prev = out[key]
+        if isinstance(prev, dict) and isinstance(val, dict):
+            _deep_merge_dicts(prev, val)
+        elif isinstance(prev, list) and isinstance(val, list):
+            prev.extend(val)
+        else:
+            out[key] = val
+    if duplicates:
+        # logger may not be wired yet — fall back to print so we never lose the warning.
+        unique = sorted(set(duplicates))
+        try:
+            logger.warning("[INIT] config.json has duplicate keys (merged): %s", unique)
+        except Exception:
+            print("[INIT] config.json has duplicate keys (merged):", unique)
+    return out
+
+
+def _merge_legacy_namespace(cfg, legacy: str, canonical: str) -> None:
+    """Fold deprecated singular keys (``tool`` / ``skill``) into their plural
+    canonical counterparts at load time. Canonical entries always win."""
+    legacy_section = cfg.get(legacy)
+    if not isinstance(legacy_section, dict) or not legacy_section:
+        cfg.pop(legacy, None)
+        return
+    canonical_section = cfg.get(canonical)
+    if not isinstance(canonical_section, dict):
+        canonical_section = {}
+    merged_keys = []
+    for name, val in legacy_section.items():
+        if name in canonical_section:
+            if isinstance(canonical_section[name], dict) and isinstance(val, dict):
+                for sub_key, sub_val in val.items():
+                    if (
+                        sub_key in canonical_section[name]
+                        and isinstance(canonical_section[name][sub_key], dict)
+                        and isinstance(sub_val, dict)
+                    ):
+                        _deep_merge_dicts(sub_val, canonical_section[name][sub_key])
+                        canonical_section[name][sub_key] = sub_val
+                    else:
+                        canonical_section[name].setdefault(sub_key, sub_val)
+            continue
+        canonical_section[name] = val
+        merged_keys.append(name)
+    cfg[canonical] = canonical_section
+    cfg.pop(legacy, None)
+    if merged_keys:
+        logger.warning(
+            "[INIT] Legacy config key '{}' is deprecated; merged into '{}': {}. "
+            "Please rename '{}' to '{}' in your config.json.".format(
+                legacy, canonical, merged_keys, legacy, canonical,
+            )
+        )
+
+
 def _sync_skill_config_to_env(skill_section) -> int:
     """Flatten skill-namespaced config into environment variables.
 
-    Mapping rule: ``config["skill"][<name>][<key>]`` -> ``SKILL_<NAME>_<KEY>``
-    (e.g. ``skill["image-generation"].model`` -> ``SKILL_IMAGE_GENERATION_MODEL``).
+    Mapping rule: ``config["skills"][<name>][<key>]`` -> ``SKILL_<NAME>_<KEY>``
+    (e.g. ``skills["image-generation"].model`` -> ``SKILL_IMAGE_GENERATION_MODEL``).
 
     This lets subprocess-based skill scripts read their own settings without
     importing project code. Existing env vars are NOT overwritten so the

docs/channels/index.mdx

@@ -0,0 +1,39 @@
+---
+title: 通道概览
+description: CowAgent 支持的通道及能力矩阵
+---
+
+CowAgent 支持接入多种聊天通道，启动时通过 `channel_type` 切换。Web 控制台默认开启，可与其他接入通道并行运行。
+
+## 能力矩阵
+
+下表汇总各通道支持的入站消息类型、机器人回复类型与群聊能力，方便按场景选择。
+
+| 通道 | 文本 | 图片 | 文件 | 语音 | 群聊 |
+| --- | :-: | :-: | :-: | :-: | :-: |
+| [微信](/channels/weixin) | ✅ | ✅ | ✅ | ✅ |  |
+| [Web 控制台](/channels/web) | ✅ | ✅ | ✅ | ✅ | |
+| [飞书](/channels/feishu) | ✅ | ✅ | ✅ | ✅ | ✅ |
+| [钉钉](/channels/dingtalk) | ✅ | ✅ | ✅ | ✅ | ✅ |
+| [企微智能机器人](/channels/wecom-bot) | ✅ | ✅ | ✅ | ✅ | ✅ |
+| [QQ](/channels/qq) | ✅ | ✅ | ✅ | | ✅ |
+| [企业微信应用](/channels/wecom) | ✅ | ✅ | ✅ | ✅ | |
+| [公众号](/channels/wechatmp) | ✅ | ✅ | | ✅ | |
+
+- **图片 / 文件 / 语音**列表示通道支持收发对应消息类型，具体细节详见各通道文档
+- **群聊**列指可识别并响应群消息
+
+<Tip>
+  每个通道的语音 / 图像能力依赖对应模型厂商的配置，详见 [模型概览](/models)。
+</Tip>
+
+## 通道一览
+
+- [Web 控制台](/channels/web) — 内置浏览器对话和管理面板，默认开启
+- [微信](/channels/weixin) — 通过个人微信扫码登录
+- [飞书](/channels/feishu) — 飞书自建机器人
+- [钉钉](/channels/dingtalk) — 钉钉自建机器人
+- [企微智能机器人](/channels/wecom-bot) — 企业微信智能机器人
+- [QQ](/channels/qq) — QQ 官方机器人开放平台
+- [企业微信应用](/channels/wecom) — 企业微信自建应用接入
+- [公众号](/channels/wechatmp) — 微信公众号（订阅号 / 服务号）

docs/channels/web.mdx

@@ -59,9 +59,9 @@ Web 控制台是 CowAgent 的默认通道，启动后会自动运行，通过浏
 
 ### 模型管理
 
-支持在线管理模型配置，无需手动编辑配置文件：
+支持在线管理不同模型厂商的文本、图像、语音、向量模型配置，无需手动编辑配置文件：
 
-<img width="850" src="https://cdn.link-ai.tech/doc/20260227173811.png" />
+<img width="850" src="https://cdn.link-ai.tech/doc/20260521212949.png" />
 
 ### 技能管理
 

docs/docs.json

@@ -181,6 +181,7 @@
               {
                 "group": "接入渠道",
                 "pages": [
+                  "channels/index",
                   "channels/weixin",
                   "channels/web",
                   "channels/feishu",

docs/en/models/qianfan.mdx

@@ -40,7 +40,7 @@ To force a specific Vision model, set it explicitly in `config.json`:
 
 ```json
 {
-  "tool": {
+  "tools": {
     "vision": {
       "model": "ernie-4.5-turbo-vl"
     }

docs/en/releases/v2.0.7.mdx

@@ -11,7 +11,7 @@ New built-in `image-generation` skill supporting text-to-image, image-to-image,
 - **Zero model selection**: Just configure an API key and it works — no need to manually specify a model. You can also name a specific model in conversation (e.g. "draw a cat with seedream")
 - **Flexible control**: Supports `quality`, `size` (512/1K–4K), and `aspect_ratio` parameters, with each provider automatically mapping to its supported values
 - **Image editing**: Pass existing images for editing, style transfer, or multi-image fusion (Seedream supports up to 14 reference images)
-- **Skill-level config**: Pin a default model via `skill.image-generation.model` in `config.json`
+- **Skill-level config**: Pin a default model via `skills.image-generation.model` in `config.json`
 - **Image lightbox**: All images in the Web console now support click-to-enlarge preview
 
 Docs: [Image Generation Skill](https://docs.cowagent.ai/en/skills/image-generation)

docs/en/releases/v2.0.8.mdx

@@ -51,7 +51,7 @@ The voice and streaming building blocks come from a community contribution #2791
 
 ## 🔧 Tools and Safety
 
-- **Vision model selection**: `tool.vision.model` config now actually takes effect, with automatic fallback when unconfigured #2792
+- **Vision model selection**: `tools.vision.model` config now actually takes effect, with automatic fallback when unconfigured #2792
 - **Bash safety prompt**: The destructive-deletion confirm prompt is now scoped to paths outside the workspace — routine in-workspace operations are no longer interrupted
 
 ## 🐛 Other Fixes

docs/en/skills/image-generation.mdx

@@ -87,7 +87,7 @@ Configure ARK_API_KEY as xxx
 To force all image generation through a specific provider's model, add this to `config.json`:
 
 ```json
-"skill": {
+"skills": {
   "image-generation": {
     "model": "seedream-5.0-lite"
   }

docs/en/tools/vision.mdx

@@ -51,7 +51,7 @@ To specify a particular model for the vision tool, add to `config.json`:
 
 ```json
 {
-    "tool": {
+    "tools": {
         "vision": {
             "model": "ernie-4.5-turbo-vl"
         }

docs/ja/models/qianfan.mdx

@@ -40,7 +40,7 @@ description: Baidu Qianfan ERNIE モデル設定
 
 ```json
 {
-  "tool": {
+  "tools": {
     "vision": {
       "model": "ernie-4.5-turbo-vl"
     }

docs/ja/releases/v2.0.7.mdx

@@ -11,7 +11,7 @@ description: CowAgent 2.0.7 - 画像生成スキル（6プロバイダー自動
 - **モデル選択不要**：API Key を設定するだけで使用可能、モデルを手動で指定する必要なし。会話で特定モデルを指名することも可能（例：「seedream で猫を描いて」）
 - **柔軟な制御**：`quality`（画質）、`size`（解像度、512/1K〜4K）、`aspect_ratio`（アスペクト比）パラメータ対応、各プロバイダーが自動的に有効な値にマッピング
 - **画像編集**：既存の画像を渡して編集・スタイル変換・複数画像融合が可能（Seedream は最大 14 枚の参照画像をサポート）
-- **スキルレベル設定**：`config.json` の `skill.image-generation.model` でデフォルトモデルを固定可能
+- **スキルレベル設定**：`config.json` の `skills.image-generation.model` でデフォルトモデルを固定可能
 - **画像ライトボックス**：Web コンソールのすべての画像がクリックで拡大プレビュー対応
 
 ドキュメント：[画像生成スキル](https://docs.cowagent.ai/ja/skills/image-generation)

docs/ja/releases/v2.0.8.mdx

@@ -51,7 +51,7 @@ description: CowAgent 2.0.8 - 飛書チャネル全面アップグレード（
 
 ## 🔧 ツールと安全性
 
-- **Vision モデル選択**：`tool.vision.model` 設定が実際に反映されるようになり、未設定時は自動フォールバック #2792
+- **Vision モデル選択**：`tools.vision.model` 設定が実際に反映されるようになり、未設定時は自動フォールバック #2792
 - **Bash セーフティ確認**：破壊的削除の確認プロンプトをワークスペース外のパスに限定。ワークスペース内の通常操作は中断されません
 
 ## 🐛 その他の修正

docs/ja/skills/image-generation.mdx

@@ -87,7 +87,7 @@ ARK_API_KEY を xxx に設定して
 すべての画像生成を特定のプロバイダーのモデルで固定したい場合、`config.json` に以下を追加：
 
 ```json
-"skill": {
+"skills": {
   "image-generation": {
     "model": "seedream-5.0-lite"
   }

docs/ja/tools/vision.mdx

@@ -51,7 +51,7 @@ Vision ツールで使用するモデルを指定するには、`config.json`
 
 ```json
 {
-    "tool": {
+    "tools": {
         "vision": {
             "model": "ernie-4.5-turbo-vl"
         }

docs/models/claude.mdx

@@ -1,8 +1,16 @@
 ---
 title: Claude
-description: Claude 模型配置
+description: Anthropic Claude 模型配置（文本对话 + 图像理解）
 ---
 
+Claude 由 Anthropic 提供，支持文本对话与图像理解，主流 Sonnet / Opus 模型均原生支持视觉，无需额外指定 Vision 模型。
+
+<Tip>
+  通过 Web 控制台的「模型管理」页面可一站式配置以下全部能力，无需手动改配置文件。
+</Tip>
+
+## 文本对话
+
 ```json
 {
   "model": "claude-sonnet-4-6",
@@ -14,4 +22,28 @@ description: Claude 模型配置
 | --- | --- |
 | `model` | 支持 `claude-sonnet-4-6`、`claude-opus-4-7`、`claude-opus-4-6`、`claude-sonnet-4-5`、`claude-sonnet-4-0`、`claude-3-5-sonnet-latest` 等，参考 [官方模型](https://docs.anthropic.com/en/docs/about-claude/models/overview) |
 | `claude_api_key` | 在 [Claude 控制台](https://console.anthropic.com/settings/keys) 创建 |
-| `claude_api_base` | 可选，默认为 `https://api.anthropic.com/v1`，修改可接入第三方代理 |
+| `claude_api_base` | 可选，默认为 `https://api.anthropic.com/v1`，可改为第三方代理 |
+
+### 模型选择
+
+| 模型 | 适用场景 |
+| --- | --- |
+| `claude-sonnet-4-6` | 默认推荐，性价比与速度平衡 |
+| `claude-opus-4-7` | 复杂推理与长链路任务，效果最佳但成本更高 |
+| `claude-sonnet-4-5` / `claude-sonnet-4-0` | 上一代旗舰，价格更低 |
+
+## 图像理解
+
+配置 `claude_api_key` 后 Agent 的 Vision 工具会自动使用 Claude 主模型识别图像，无需额外配置。
+
+如需手动指定 Vision 模型，可在配置文件中显式配置：
+
+```json
+{
+  "tools": {
+    "vision": {
+      "model": "claude-sonnet-4-6"
+    }
+  }
+}
+```

docs/models/custom.mdx

@@ -13,7 +13,7 @@ description: 自定义厂商配置，适用于第三方 API 代理和本地模
   与 `openai` 厂商的区别：选择自定义厂商后，通过 `/config model` 切换模型时，不会自动切换厂商类型，始终使用自定义的 API 地址。
 </Note>
 
-## 配置方式
+## 文本对话
 
 ### 第三方 API 代理
 
@@ -35,7 +35,7 @@ description: 自定义厂商配置，适用于第三方 API 代理和本地模
 
 ### 本地模型
 
-本地模型通常不需要 API Key，只需填写 API Base 即可：
+本地模型通常不需要 API Key，只需填写 API Base：
 
 ```json
 {
@@ -53,7 +53,7 @@ description: 自定义厂商配置，适用于第三方 API 代理和本地模
 | [vLLM](https://docs.vllm.ai) | `http://localhost:8000/v1` |
 | [LocalAI](https://localai.io) | `http://localhost:8080/v1` |
 
-## 切换模型
+### 切换模型
 
 自定义厂商下切换模型时，只会修改 `model`，不会改变 `bot_type` 和 API 地址：
 

docs/models/deepseek.mdx

@@ -1,9 +1,11 @@
 ---
 title: DeepSeek
-description: DeepSeek 模型配置
+description: DeepSeek 模型配置（文本对话 + 思考模式）
 ---
 
-方式一：官方接入（推荐）：
+DeepSeek 是当前 Agent 模式默认推荐的厂商之一，主打高性价比的文本对话和任务规划能力。
+
+## 文本对话
 
 ```json
 {
@@ -18,20 +20,20 @@ description: DeepSeek 模型配置
 | `deepseek_api_key` | 在 [DeepSeek 平台](https://platform.deepseek.com/api_keys) 创建 |
 | `deepseek_api_base` | 可选，默认为 `https://api.deepseek.com/v1`，可修改为第三方代理地址 |
 
-## 模型选择
+### 模型选择
 
 | 模型 | 适用场景 |
 | --- | --- |
 | `deepseek-v4-flash` | 默认推荐，速度快、成本低 |
-| `deepseek-v4-pro` | 更智能、复杂任务效果更强 |
+| `deepseek-v4-pro` | 更智能，复杂任务效果更强 |
 
 ## 思考模式
 
-V4 系列（`deepseek-v4-flash` / `deepseek-v4-pro`）支持显式的"思考模式"：模型在输出最终回答前，先输出一段思维链（`reasoning_content`），从而提升答案质量。
+V4 系列（`deepseek-v4-flash` / `deepseek-v4-pro`）支持显式的「思考模式」：模型在输出最终回答前，先输出一段思维链（`reasoning_content`），从而提升答案质量。
 
 ### 开关
 
-通过全局配置 `enable_thinking` 控制：
+通过全局配置 `enable_thinking` 控制，也可在 web控制台 - 配置页面中进行切换：
 
 ```json
 {
@@ -66,16 +68,5 @@ V4 系列（`deepseek-v4-flash` / `deepseek-v4-pro`）支持显式的"思考模
 - **多轮工具调用**：当历史中包含工具调用时，DeepSeek 要求所有 assistant 消息必须回传 `reasoning_content`。CowAgent 会自动处理回传逻辑，跨轮次切换思考开关也不会出错。
 
 <Tip>
-  默认使用 `deepseek-v4-flash`；复杂任务可使用 `deepseek-v4-pro`；需要深度思考可开启 `enable_thinking`。
+  默认使用 `deepseek-v4-flash`；复杂任务可使用 `deepseek-v4-pro`；需要深度推理可开启 `enable_thinking`。
 </Tip>
-
-方式二：OpenAI 兼容方式接入：
-
-```json
-{
-  "model": "deepseek-v4-flash",
-  "bot_type": "openai",
-  "open_ai_api_key": "YOUR_API_KEY",
-  "open_ai_api_base": "https://api.deepseek.com/v1"
-}
-```

docs/models/doubao.mdx

@@ -1,17 +1,66 @@
 ---
 title: 豆包 Doubao
-description: 豆包 (火山方舟) 模型配置
+description: 豆包（火山方舟）模型配置（文本 / 图像理解 / 图像生成 / 向量）
 ---
 
+豆包（火山方舟）支持文本对话、图像理解、图像生成（Seedream）和向量能力，一份 `ark_api_key` 即可启用全部能力。
+
+<Tip>
+  通过 Web 控制台的「模型管理」页面可一站式配置以下全部能力，无需手动改配置文件。
+</Tip>
+
+## 文本对话
+
 ```json
 {
-  "model": "doubao-seed-2-0-code-preview-260215",
+  "model": "doubao-seed-2-0-pro-260215",
   "ark_api_key": "YOUR_API_KEY"
 }
 ```
 
 | 参数 | 说明 |
 | --- | --- |
-| `model` | 可填 `doubao-seed-2-0-code-preview-260215`、`doubao-seed-2-0-pro-260215`、`doubao-seed-2-0-lite-260215` 等 |
+| `model` | 可填 `doubao-seed-2-0-pro-260215`、`doubao-seed-2-0-code-preview-260215`、`doubao-seed-2-0-lite-260215` 等 |
 | `ark_api_key` | 在 [火山方舟控制台](https://console.volcengine.com/ark/region:ark+cn-beijing/apikey) 创建 |
 | `ark_base_url` | 可选，默认为 `https://ark.cn-beijing.volces.com/api/v3` |
+
+## 图像理解
+
+配置 `ark_api_key` 后 Agent 的 Vision 工具会自动使用 `doubao-seed-2-0-pro-260215` 识别图像，无需额外配置。
+
+如需手动指定 Vision 模型：
+
+```json
+{
+  "tools": {
+    "vision": {
+      "model": "doubao-seed-2-0-pro-260215"
+    }
+  }
+}
+```
+
+## 图像生成
+
+```json
+{
+  "skills": {
+    "image-generation": {
+      "model": "seedream-5.0-lite"
+    }
+  }
+}
+```
+
+可选模型：`seedream-5.0-lite`、`seedream-4.5`。
+
+## 向量
+
+```json
+{
+  "embedding_provider": "doubao",
+  "embedding_model": "doubao-embedding-vision-251215"
+}
+```
+
+默认模型 `doubao-embedding-vision-251215`（多模态 embedding），可在配置文件中通过 `embedding_dimensions` 指定 1024 或 2048 维。修改 embedding 后需执行 `/memory rebuild-index` 命令重建索引。

docs/models/gemini.mdx

@@ -1,16 +1,59 @@
 ---
 title: Gemini
-description: Google Gemini 模型配置
+description: Google Gemini 模型配置（文本对话 + 图像理解 + 图像生成）
 ---
 
+Google Gemini 支持文本对话、图像理解和图像生成（Nano Banana 系列），一个 `gemini_api_key` 即可启用全部能力。
+
+<Tip>
+  通过 Web 控制台的「模型管理」页面可一站式配置以下全部能力，无需手动改配置文件。
+</Tip>
+
+## 文本对话
+
 ```json
 {
-  "model": "gemini-3.1-pro-preview",
+  "model": "gemini-3.5-flash",
   "gemini_api_key": "YOUR_API_KEY"
 }
 ```
 
 | 参数 | 说明 |
 | --- | --- |
-| `model` | 支持 `gemini-3.1-flash-lite-preview`、`gemini-3.1-pro-preview`、`gemini-3-flash-preview`、`gemini-3-pro-preview` 等，参考 [官方文档](https://ai.google.dev/gemini-api/docs/models) |
+| `model` | 推荐 `gemini-3.5-flash`，亦支持 `gemini-3.1-pro-preview`、`gemini-3.1-flash-lite-preview`、`gemini-3-flash-preview`、`gemini-3-pro-preview` 等，参考 [官方文档](https://ai.google.dev/gemini-api/docs/models) |
 | `gemini_api_key` | 在 [Google AI Studio](https://aistudio.google.com/app/apikey) 创建 |
+| `gemini_api_base` | 可选，默认为 `https://generativelanguage.googleapis.com`，可改为第三方代理 |
+
+## 图像理解
+
+Gemini 全系列模型均原生支持视觉，配置 `gemini_api_key` 后 Agent 的 Vision 工具会自动使用主模型识别图像，无需额外配置。
+
+如需手动指定 Vision 模型：
+
+```json
+{
+  "tools": {
+    "vision": {
+      "model": "gemini-3.1-flash-lite-preview"
+    }
+  }
+}
+```
+
+## 图像生成
+
+```json
+{
+  "skills": {
+    "image-generation": {
+      "model": "gemini-3.1-flash-image-preview"
+    }
+  }
+}
+```
+
+| 模型 ID | 别名 |
+| --- | --- |
+| `gemini-3.1-flash-image-preview` | Nano Banana 2 |
+| `gemini-3-pro-image-preview` | Nano Banana Pro |
+| `gemini-2.5-flash-image` | Nano Banana |

docs/models/glm.mdx

@@ -1,8 +1,16 @@
 ---
 title: 智谱 GLM
-description: 智谱AI GLM 模型配置
+description: 智谱 AI GLM 模型配置（文本 / 图像理解 / 语音识别 / 向量）
 ---
 
+智谱 AI 支持文本对话、图像理解、语音识别（ASR）和向量（Embedding），一份 `zhipu_ai_api_key` 即可启用全部能力。
+
+<Tip>
+  通过 Web 控制台的「模型管理」页面可一站式配置以下全部能力，无需手动改配置文件。
+</Tip>
+
+## 文本对话
+
 ```json
 {
   "model": "glm-5.1",
@@ -14,14 +22,35 @@ description: 智谱AI GLM 模型配置
 | --- | --- |
 | `model` | 可填 `glm-5.1`、`glm-5-turbo`、`glm-5`、`glm-4.7`、`glm-4-plus`、`glm-4-flash`、`glm-4-air` 等，参考 [模型编码](https://bigmodel.cn/dev/api/normal-model/glm-4) |
 | `zhipu_ai_api_key` | 在 [智谱 AI 控制台](https://www.bigmodel.cn/usercenter/proj-mgmt/apikeys) 创建 |
+| `zhipu_ai_api_base` | 可选，默认为 `https://open.bigmodel.cn/api/paas/v4` |
 
-也支持 OpenAI 兼容方式接入：
+## 图像理解
+
+智谱 chat 系列模型（`glm-5.1`、`glm-5-turbo` 等）不支持视觉，视觉调用统一路由到 `glm-5v-turbo`。配置 `zhipu_ai_api_key` 后 Agent 的 Vision 工具会自动使用该模型，无需在配置文件中显式指定。
+
+## 语音识别
 
 ```json
 {
-  "bot_type": "openai",
-  "model": "glm-5.1",
-  "open_ai_api_base": "https://open.bigmodel.cn/api/paas/v4",
-  "open_ai_api_key": "YOUR_API_KEY"
+  "voice_to_text": "zhipu",
+  "voice_to_text_model": "glm-asr-2512"
 }
 ```
+
+| 参数 | 说明 |
+| --- | --- |
+| `voice_to_text` | 设为 `zhipu` 启用智谱 ASR |
+| `voice_to_text_model` | 可选，默认 `glm-asr-2512` |
+
+凭证自动复用 `zhipu_ai_api_key`。语音文件建议小于 25MB，超大文件可能被服务端拒绝。
+
+## 向量
+
+```json
+{
+  "embedding_provider": "zhipu",
+  "embedding_model": "embedding-3"
+}
+```
+
+可选模型：`embedding-3`、`embedding-2`。修改 embedding 后需执行 `/memory rebuild-index` 命令重建索引。

docs/models/index.mdx

@@ -1,67 +1,45 @@
 ---
 title: 模型概览
-description: CowAgent 支持的模型及推荐选择
+description: CowAgent 支持的模型厂商及能力矩阵
 ---
 
-CowAgent 支持国内外主流厂商的大语言模型，模型接口实现在项目的 `models/` 目录下。
+CowAgent 支持国内外主流厂商的大语言模型，模型接口实现在项目的 `models/` 目录下。除文本对话外，部分厂商还提供视觉理解、图像生成、语音识别、语音合成、向量等能力，可在 Agent 流程中按需调用。
 
 <Note>
-  Agent 模式下推荐使用以下模型，可根据效果及成本综合选择：deepseek-v4-flash、MiniMax-M2.7、claude-sonnet-4-6、gemini-3.1-pro-preview、glm-5.1、qwen3.6-plus、kimi-k2.6、ernie-5.1
+  Agent 模式下推荐使用以下模型，可根据效果及成本综合选择：deepseek-v4-flash、MiniMax-M2.7、claude-sonnet-4-6、gemini-3.5-flash、glm-5.1、qwen3.6-plus、kimi-k2.6、ernie-5.1。
 
-  同时支持使用 [LinkAI](https://link-ai.tech) 平台接口，可灵活切换多种模型，并支持知识库、工作流、插件等 Agent 能力。
+  同时支持使用 [LinkAI](https://link-ai.tech) 平台接口，一个 Key 即可灵活切换多家厂商，并附带知识库、工作流、插件等能力。
 </Note>
 
+
+## 模型能力总览
+
+各厂商提供的能力一览。「文本」指主对话模型，其余列表示该厂商可承担对应 Agent 能力。
+
+| 厂商 | 代表模型 | 文本 | 图像理解 | 图像生成 | 语音识别 | 语音合成 | 向量 |
+| --- | --- | :-: | :-: | :-: | :-: | :-: | :-: |
+| [DeepSeek](/models/deepseek) | deepseek-v4-flash / pro | ✅ | | | | | |
+| [MiniMax](/models/minimax) | MiniMax-M2.7 | ✅ | ✅ | ✅ | | ✅ | |
+| [Claude](/models/claude) | claude-opus-4-7 | ✅ | ✅ | | | | |
+| [Gemini](/models/gemini) | gemini-3.5-flash | ✅ | ✅ | ✅ | | | |
+| [OpenAI](/models/openai) | gpt-5.5、o 系列 | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |
+| [智谱 GLM](/models/glm) | glm-5.1、glm-5v-turbo | ✅ | ✅ | | ✅ | | ✅ |
+| [通义千问](/models/qwen) | qwen3.7-max | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |
+| [豆包 Doubao](/models/doubao) | doubao-seed-2.0 系列 | ✅ | ✅ | ✅ | | | ✅ |
+| [Kimi](/models/kimi) | kimi-k2.6 | ✅ | ✅ | | | | |
+| [百度千帆](/models/qianfan) | ernie-5.1 | ✅ | ✅ | | | | |
+| [LinkAI](/models/linkai) | 多厂商 100+ 模型统一接入 | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |
+| [自定义](/models/custom) |本地模型 / 三方代理 | ✅ | | | | | |
+
+<Tip>
+  Web 控制台中各项能力（视觉 / 图像 / 语音识别 / 语音合成 / 向量 / 网络搜索）均可独立配置厂商与模型，互相之间不强制绑定。
+</Tip>
+
+
 ## 配置方式
 
-**方式一（推荐）：** 通过 [Web 控制台](/channels/web) 在线管理模型配置，无需手动编辑配置文件：
+**方式一（推荐）：** 通过 [Web 控制台](/channels/web) 在线管理模型与各项能力，无需手动编辑配置文件：
 
-<img width="850" src="https://cdn.link-ai.tech/doc/20260227173811.png" />
+<img width="900" src="https://cdn.link-ai.tech/doc/20260521212527.png" />
 
 **方式二：** 手动编辑 `config.json`，根据所选模型填写对应的模型名称和 API Key。每个模型也支持 OpenAI 兼容方式接入，将 `bot_type` 设为 `openai`，配置 `open_ai_api_base` 和 `open_ai_api_key` 即可。
-
-
-## 支持的模型
-
-<CardGroup cols={2}>
-  <Card title="DeepSeek" href="/models/deepseek">
-    deepseek-v4-flash、deepseek-v4-pro 等
-  </Card>
-  <Card title="百度千帆 / ERNIE" href="/models/qianfan">
-    ernie-5.1、ernie-5.0、ernie-4.5-turbo-128k 等
-  </Card>
-  <Card title="MiniMax" href="/models/minimax">
-    MiniMax-M2.7 等系列模型
-  </Card>
-  <Card title="Claude" href="/models/claude">
-    claude-sonnet-4-6 等
-  </Card>
-  <Card title="Gemini" href="/models/gemini">
-    gemini-3.1-pro-preview 等
-  </Card>
-  <Card title="OpenAI" href="/models/openai">
-    gpt-5.4、gpt-4.1、o 系列等
-  </Card>
-  <Card title="智谱 GLM" href="/models/glm">
-    glm-5.1、glm-5-turbo、glm-5 等系列模型
-  </Card>
-  <Card title="通义千问 Qwen" href="/models/qwen">
-    qwen3.6-plus、qwen3-max 等
-  </Card>
-  <Card title="豆包 Doubao" href="/models/doubao">
-    doubao-seed 系列模型
-  </Card>
-  <Card title="Kimi" href="/models/kimi">
-    kimi-k2.6、kimi-k2.5、kimi-k2 等
-  </Card>
-  <Card title="LinkAI" href="/models/linkai">
-    多模型统一接口 + 知识库
-  </Card>
-  <Card title="自定义" href="/models/custom">
-    第三方代理、本地模型等
-  </Card>
-</CardGroup>
-
-
-<Tip>
-  全部模型名称可参考项目 [`common/const.py`](https://github.com/zhayujie/CowAgent/blob/master/common/const.py) 文件。
-</Tip>

docs/models/kimi.mdx

@@ -1,8 +1,16 @@
 ---
 title: Kimi
-description: Kimi (Moonshot) 模型配置
+description: Kimi（Moonshot）模型配置（文本对话 + 图像理解）
 ---
 
+Kimi 由 Moonshot 提供，支持文本对话与图像理解，`kimi-k2.x` 系列原生支持视觉。
+
+<Tip>
+  通过 Web 控制台的「模型管理」页面可一站式配置以下全部能力，无需手动改配置文件。
+</Tip>
+
+## 文本对话
+
 ```json
 {
   "model": "kimi-k2.6",
@@ -14,14 +22,20 @@ description: Kimi (Moonshot) 模型配置
 | --- | --- |
 | `model` | 可填 `kimi-k2.6`、`kimi-k2.5`、`kimi-k2`、`moonshot-v1-8k`、`moonshot-v1-32k`、`moonshot-v1-128k` |
 | `moonshot_api_key` | 在 [Moonshot 控制台](https://platform.moonshot.cn/console/api-keys) 创建 |
+| `moonshot_base_url` | 可选，默认为 `https://api.moonshot.cn/v1` |
 
-也支持 OpenAI 兼容方式接入：
+## 图像理解
+
+配置 `moonshot_api_key` 后 Agent 的 Vision 工具会自动使用 `kimi-k2.6` 识别图像，无需额外配置。
+
+如需手动指定 Vision 模型：
 
 ```json
 {
-  "bot_type": "openai",
-  "model": "kimi-k2.6",
-  "open_ai_api_base": "https://api.moonshot.cn/v1",
-  "open_ai_api_key": "YOUR_API_KEY"
+  "tools": {
+    "vision": {
+      "model": "kimi-k2.6"
+    }
+  }
 }
 ```

docs/models/linkai.mdx

@@ -1,9 +1,15 @@
 ---
 title: LinkAI
-description: 通过 LinkAI 平台统一接入多种模型
+description: 通过 LinkAI 平台统一接入文本、视觉、图像、语音与向量能力
 ---
 
-通过 [LinkAI](https://link-ai.tech) 平台可灵活切换 OpenAI、Claude、Gemini、DeepSeek、MiniMax、Qwen、Kimi 等多种模型，并支持知识库、工作流、插件等 Agent 能力。
+通过一份 `linkai_api_key` 即可访问 OpenAI、Claude、Gemini、DeepSeek、MiniMax、Qwen、Kimi、豆包 等主流厂商的全部能力。
+
+<Tip>
+  通过 Web 控制台的「模型管理」页面可一站式配置以下全部能力，无需手动改配置文件。
+</Tip>
+
+## 文本对话
 
 ```json
 {
@@ -14,8 +20,84 @@ description: 通过 LinkAI 平台统一接入多种模型
 
 | 参数 | 说明 |
 | --- | --- |
-| `use_linkai` | 设为 `true` 启用 LinkAI 接口 |
+| `use_linkai` | 设为 `true` 启用 |
 | `linkai_api_key` | 在 [控制台](https://link-ai.tech/console/interface) 创建 |
-| `model` | 留空则使用智能体默认模型，可在平台中灵活切换，[模型列表](https://link-ai.tech/console/models) 中的全部模型均可使用 |
+| `model` | 可填写 [模型列表](https://link-ai.tech/console/models) 中任意编码 |
 
-参考 [接口文档](https://docs.link-ai.tech/platform/api) 了解更多。
+前往 [模型服务](https://link-ai.tech/console/models) 了解更多。
+
+## 图像理解
+
+配置完成后 Agent 的 Vision 工具会自动调用网关上的多模态模型，无需额外配置。如需手动指定 Vision 模型：
+
+```json
+{
+  "tools": {
+    "vision": {
+      "model": "gpt-5.4-mini"
+    }
+  }
+}
+```
+
+可选模型：`gpt-4.1-mini`、`gpt-5.4-mini`、`qwen3.6-plus`、`doubao-seed-2-0-pro-260215`、`kimi-k2.6`、`claude-sonnet-4-6`、`gemini-3.1-flash-lite-preview` 等。
+
+## 图像生成
+
+```json
+{
+  "skills": {
+    "image-generation": {
+      "model": "gpt-image-2"
+    }
+  }
+}
+```
+
+| 模型 ID | 别名 |
+| --- | --- |
+| `gpt-image-2` | OpenAI |
+| `gemini-3.1-flash-image-preview` | Nano Banana 2 |
+| `gemini-3-pro-image-preview` | Nano Banana Pro |
+| `seedream-5.0-lite` | 字节豆包 Seedream |
+
+## 语音识别
+
+```json
+{
+  "voice_to_text": "linkai"
+}
+```
+
+ASR 固定使用 Whisper，凭证自动复用 `linkai_api_key`。
+
+## 语音合成
+
+语音合成网关下支持多个底层 TTS 引擎，按 `text_to_voice_model` 选择引擎，音色随引擎切换。
+
+```json
+{
+  "text_to_voice": "linkai",
+  "text_to_voice_model": "doubao",
+  "tts_voice_id": "BV001_streaming"
+}
+```
+
+| `text_to_voice_model` | 引擎说明 |
+| --- | --- |
+| `tts-1` | OpenAI · 多语种通用（音色 `alloy` / `nova` / `echo` 等） |
+| `doubao` | 字节豆包 · 中文音色丰富 |
+| `baidu` | 百度 · 中文主播音色 |
+
+不同引擎对应的音色不同，建议在 Web 控制台「模型管理 → 语音合成」中可视化选择。
+
+## 向量
+
+```json
+{
+  "embedding_provider": "linkai",
+  "embedding_model": "text-embedding-3-small"
+}
+```
+
+默认模型 `text-embedding-3-small`（OpenAI 兼容）。修改 embedding 后需执行 `/memory rebuild-index` 命令重建索引。

docs/models/minimax.mdx

@@ -1,8 +1,16 @@
 ---
 title: MiniMax
-description: MiniMax 模型配置
+description: MiniMax 模型配置（文本 / 图像理解 / 图像生成 / 语音合成）
 ---
 
+MiniMax 支持文本对话、图像理解、图像生成与语音合成，一份 `minimax_api_key` 即可启用全部能力。
+
+<Tip>
+  通过 Web 控制台的「模型管理」页面可一站式配置以下全部能力，无需手动改配置文件。
+</Tip>
+
+## 文本对话
+
 ```json
 {
   "model": "MiniMax-M2.7",
@@ -12,16 +20,52 @@ description: MiniMax 模型配置
 
 | 参数 | 说明 |
 | --- | --- |
-| `model` | 可填 `MiniMax-M2.7`、`MiniMax-M2.5`、`MiniMax-M2.1`、`MiniMax-M2.1-lightning`、`MiniMax-M2` 等 |
+| `model` | 可填 `MiniMax-M2.7`、`MiniMax-M2.7-highspeed`、`MiniMax-M2.5`、`MiniMax-M2.1`、`MiniMax-M2.1-lightning`、`MiniMax-M2` 等 |
 | `minimax_api_key` | 在 [MiniMax 控制台](https://platform.minimaxi.com/user-center/basic-information/interface-key) 创建 |
 
-也支持 OpenAI 兼容方式接入：
+## 图像理解
+
+MiniMax 的 M2.x 系列 chat 模型本身不支持视觉，视觉调用统一路由到 `MiniMax-Text-01`。配置 `minimax_api_key` 后 Agent 的 Vision 工具会自动使用该模型，无需在配置文件中显式指定。
+
+## 图像生成
 
 ```json
 {
-  "bot_type": "openai",
-  "model": "MiniMax-M2.7",
-  "open_ai_api_base": "https://api.minimaxi.com/v1",
-  "open_ai_api_key": "YOUR_API_KEY"
+  "skills": {
+    "image-generation": {
+      "model": "image-01"
+    }
+  }
 }
 ```
+
+可选模型：`image-01`。
+
+## 语音合成
+
+```json
+{
+  "text_to_voice": "minimax",
+  "text_to_voice_model": "speech-2.8-hd",
+  "tts_voice_id": "female-shaonv"
+}
+```
+
+| 参数 | 说明 |
+| --- | --- |
+| `text_to_voice_model` | `speech-2.8-hd`（情绪渲染、自然听感）、`speech-2.8-turbo`（极速）、`speech-2.6-hd`、`speech-2.6-turbo` |
+| `tts_voice_id` | 音色 ID，支持中文 / 粤语 / 英 / 日 / 韩，共 70+ 种 |
+
+常用音色示例：
+
+| 音色 ID | 说明 |
+| --- | --- |
+| `female-shaonv` | 中文 · 少女（女） |
+| `female-yujie` | 中文 · 御姐（女） |
+| `female-tianmei` | 中文 · 甜美女性（女） |
+| `male-qn-jingying` | 中文 · 精英青年（男） |
+| `male-qn-badao` | 中文 · 霸道青年（男） |
+| `Cantonese_GentleLady` | 粤语 · 温柔女声 |
+| `English_Graceful_Lady` | 英文 · Graceful Lady |
+
+完整音色（中文 / 粤语 / 英 / 日 / 韩共 70+ 种）可参考 [系统音色列表](https://platform.minimaxi.com/docs/faq/system-voice-id)，也可在 Web 控制台的「模型管理 → 语音合成」下拉框中可视化选择。

docs/models/openai.mdx

@@ -1,11 +1,20 @@
 ---
 title: OpenAI
-description: OpenAI 模型配置
+description: OpenAI 模型配置（文本 / 视觉 / 图像 / 语音 / 向量）
 ---
 
+OpenAI 是覆盖最完整的厂商，可同时承担文本对话、视觉理解、图像生成、语音识别（ASR）、语音合成（TTS）和向量（Embedding）能力。一份 `open_ai_api_key` 即可让 Agent 用到全部能力。
+
+<Tip>
+  通过 Web 控制台的「模型管理」页面可一站式配置以下全部能力，无需手动改配置文件。
+</Tip>
+
+
+## 文本对话
+
 ```json
 {
-  "model": "gpt-5.4",
+  "model": "gpt-5.5",
   "open_ai_api_key": "YOUR_API_KEY",
   "open_ai_api_base": "https://api.openai.com/v1"
 }
@@ -13,7 +22,82 @@ description: OpenAI 模型配置
 
 | 参数 | 说明 |
 | --- | --- |
-| `model` | 与 OpenAI 接口的 [model 参数](https://platform.openai.com/docs/models) 一致，支持 o 系列、gpt-5.4、gpt-5.4-mini、gpt-5.4-nano、gpt-5 系列、gpt-4.1 等，Agent 模式推荐使用 `gpt-5.4` |
+| `model` | 与 OpenAI 接口的 [model 参数](https://platform.openai.com/docs/models) 一致，支持 `gpt-5.5`、`gpt-5.4`、`gpt-5.4-mini`、`gpt-5.4-nano`、`gpt-5` 系列、`gpt-4.1`、o 系列等；Agent 模式默认 `gpt-5.5`，追求性价比可改为 `gpt-5.4` |
 | `open_ai_api_key` | 在 [OpenAI 平台](https://platform.openai.com/api-keys) 创建 |
-| `open_ai_api_base` | 可选，修改可接入第三方代理接口 |
-| `bot_type` | 使用 OpenAI 官方模型时无需填写。当通过代理接口使用 Claude 等非 OpenAI 模型时，设为 `openai` |
+| `open_ai_api_base` | 可选，修改可接入第三方代理 |
+| `bot_type` | 使用 OpenAI 官方模型时无需填写；通过兼容协议接入厂商模型时需设为 `openai` |
+
+## 图像理解
+
+`gpt-5.5`、`gpt-5.4`、`gpt-4o`、`gpt-4.1` 等 OpenAI 模型均原生支持视觉，配置 `open_ai_api_key` 后 Agent 的 Vision 工具会自动使用主模型识别图像。若主模型不支持视觉或希望显式指定，可在配置文件中配置：
+
+```json
+{
+  "tools": {
+    "vision": {
+      "model": "gpt-5.4-mini"
+    }
+  }
+}
+```
+
+支持的 Vision 模型：`gpt-5.5`、`gpt-5.4`、`gpt-5.4-mini`、`gpt-5.4-nano`、`gpt-5`、`gpt-4.1`、`gpt-4.1-mini`、`gpt-4o`。
+
+## 图像生成
+
+在配置文件中指定图像生成模型，Agent 调用图像生成技能时会自动路由到 OpenAI：
+
+```json
+{
+  "skills": {
+    "image-generation": {
+      "model": "gpt-image-2"
+    }
+  }
+}
+```
+
+支持的图像生成模型：`gpt-image-2`、`gpt-image-1`。
+
+## 语音识别
+
+```json
+{
+  "voice_to_text": "openai",
+  "voice_to_text_model": "gpt-4o-mini-transcribe"
+}
+```
+
+| 参数 | 说明 |
+| --- | --- |
+| `voice_to_text` | 设为 `openai` 启用 OpenAI 语音识别 |
+| `voice_to_text_model` | 可选，默认 `gpt-4o-mini-transcribe`；也可填 `gpt-4o-transcribe`、`whisper-1` |
+
+凭证自动复用 `open_ai_api_key`。
+
+## 语音合成
+
+```json
+{
+  "text_to_voice": "openai",
+  "text_to_voice_model": "tts-1",
+  "tts_voice_id": "alloy"
+}
+```
+
+| 参数 | 说明 |
+| --- | --- |
+| `text_to_voice_model` | `tts-1`、`tts-1-hd`、`gpt-4o-mini-tts` |
+| `tts_voice_id` | 音色：`alloy`、`echo`、`fable`、`onyx`、`nova`、`shimmer`、`ash`、`ballad`、`coral`、`sage`、`verse` |
+
+## 向量
+
+```json
+{
+  "embedding_provider": "openai",
+  "embedding_model": "text-embedding-3-small"
+}
+```
+
+可选模型：`text-embedding-3-small`、`text-embedding-3-large`、`text-embedding-ada-002`。修改 embedding 后需执行 `/memory rebuild-index` 命令重建索引。
+

docs/models/qianfan.mdx

@@ -1,14 +1,20 @@
 ---
 title: 百度千帆
-description: 百度千帆 ERNIE 模型配置
+description: 百度千帆 ERNIE 模型配置（文本对话 + 图像理解）
 ---
 
-方式一：官方接入（推荐）：
+百度千帆提供 ERNIE 系列模型，支持文本对话与图像理解。
+
+<Tip>
+  通过 Web 控制台的「模型管理」页面可一站式配置以下全部能力，无需手动改配置文件。
+</Tip>
+
+## 文本对话
 
 ```json
 {
   "model": "ernie-5.1",
-  "qianfan_api_key": "",
+  "qianfan_api_key": "YOUR_API_KEY",
   "qianfan_api_base": "https://qianfan.baidubce.com/v2"
 }
 ```
@@ -19,7 +25,7 @@ description: 百度千帆 ERNIE 模型配置
 | `qianfan_api_key` | 千帆 API Key，格式通常以 `bce-v3/` 开头 |
 | `qianfan_api_base` | 可选，默认为 `https://qianfan.baidubce.com/v2` |
 
-## 模型选择
+### 模型选择
 
 | 模型 | 适用场景 |
 | --- | --- |
@@ -29,18 +35,18 @@ description: 百度千帆 ERNIE 模型配置
 | `ernie-4.5-turbo-128k` | 长上下文和通用对话 |
 | `ernie-4.5-turbo-32k` | 通用对话，成本和上下文更均衡 |
 
-## Vision 工具
+## 图像理解
 
 配置 `qianfan_api_key` 后，Agent 的 Vision 工具可以自动使用千帆视觉模型：
 
 - 当主模型本身是多模态时（如 `ernie-5.1`、`ernie-5.0`、`ernie-x1.1`、`ernie-4.5-turbo-vl`），直接由主模型识别图像，无需额外配置
 - 当主模型是纯文本时（如 `ernie-4.5-turbo-128k`），Vision 工具会自动 fallback 到 `ernie-4.5-turbo-vl`
 
-如需手动指定 Vision 模型，可在 `config.json` 中显式配置：
+如需手动指定 Vision 模型，可在配置文件中显式配置：
 
 ```json
 {
-  "tool": {
+  "tools": {
     "vision": {
       "model": "ernie-4.5-turbo-vl"
     }
@@ -48,17 +54,6 @@ description: 百度千帆 ERNIE 模型配置
 }
 ```
 
-方式二：OpenAI 兼容方式接入：
-
-```json
-{
-  "model": "ernie-5.1",
-  "bot_type": "openai",
-  "open_ai_api_key": "",
-  "open_ai_api_base": "https://qianfan.baidubce.com/v2"
-}
-```
-
 <Tip>
   新配置推荐使用 `qianfan_api_key`。旧的 `wenxin`、`wenxin-4`、`baidu_wenxin_api_key`、`baidu_wenxin_secret_key` 配置仍保持兼容。
 </Tip>

docs/models/qwen.mdx

@@ -1,8 +1,16 @@
 ---
 title: 通义千问 Qwen
-description: 通义千问模型配置
+description: 通义千问模型配置（文本 / 图像理解 / 图像生成 / 语音识别 / 语音合成 / 向量）
 ---
 
+通义千问（DashScope / 百炼）是国内覆盖最完整的厂商之一，文本、图像理解、图像生成、语音识别、语音合成与向量能力均可用一份 `dashscope_api_key` 启用。
+
+<Tip>
+  通过 Web 控制台的「模型管理」页面可一站式配置以下全部能力，无需手动改配置文件。
+</Tip>
+
+## 文本对话
+
 ```json
 {
   "model": "qwen3.6-plus",
@@ -12,16 +20,93 @@ description: 通义千问模型配置
 
 | 参数 | 说明 |
 | --- | --- |
-| `model` | 可填 `qwen3.6-plus`、`qwen3.5-plus`、`qwen3-max`、`qwen-max`、`qwen-plus`、`qwen-turbo`、`qwq-plus` 等 |
+| `model` | 可填 `qwen3.6-plus`、`qwen3.7-max`、`qwen3.5-plus`、`qwen3-max`、`qwen-max`、`qwen-plus`、`qwen-turbo`、`qwq-plus` 等 |
 | `dashscope_api_key` | 在 [百炼控制台](https://bailian.console.aliyun.com/?tab=model#/api-key) 创建，参考 [官方文档](https://bailian.console.aliyun.com/?tab=api#/api) |
 
-也支持 OpenAI 兼容方式接入：
+## 图像理解
+
+配置 `dashscope_api_key` 后 Agent 的 Vision 工具会自动调用千问的视觉模型识别图像。`qwen3-max` / `qwen3.5-plus` / `qwen3.6-plus` 等模型本身就是多模态；若主模型是纯文本（如 `qwen-turbo`），会自动回落到 `qwen-vl-max`。
+
+如需手动指定 Vision 模型：
 
 ```json
 {
-  "bot_type": "openai",
-  "model": "qwen3.6-plus",
-  "open_ai_api_base": "https://dashscope.aliyuncs.com/compatible-mode/v1",
-  "open_ai_api_key": "YOUR_API_KEY"
+  "tools": {
+    "vision": {
+      "model": "qwen3.6-plus"
+    }
+  }
 }
 ```
+
+支持模型：`qwen3.6-plus`、`qwen3.5-plus`、`qwen3-max`。
+
+## 图像生成
+
+```json
+{
+  "skills": {
+    "image-generation": {
+      "model": "qwen-image-2.0"
+    }
+  }
+}
+```
+
+可选模型：`qwen-image-2.0`、`qwen-image-2.0-pro`。
+
+## 语音识别
+
+```json
+{
+  "voice_to_text": "dashscope",
+  "voice_to_text_model": "qwen3-asr-flash"
+}
+```
+
+| 参数 | 说明 |
+| --- | --- |
+| `voice_to_text` | 设为 `dashscope` 启用通义千问 ASR |
+| `voice_to_text_model` | 可选，默认 `qwen3-asr-flash` |
+
+凭证自动复用 `dashscope_api_key`。单段音频建议小于 10MB、时长不超过 300 秒。
+
+## 语音合成
+
+```json
+{
+  "text_to_voice": "dashscope",
+  "text_to_voice_model": "qwen3-tts-flash",
+  "tts_voice_id": "Cherry"
+}
+```
+
+| 参数 | 说明 |
+| --- | --- |
+| `text_to_voice_model` | 可选，默认 `qwen3-tts-flash`，覆盖普通话、方言与主流外语 |
+| `tts_voice_id` | 音色 ID，详见下方常用列表 |
+
+常用音色示例：
+
+| 音色 ID | 说明 |
+| --- | --- |
+| `Cherry` | 芊悦 · 阳光女声 |
+| `Serena` | 苏瑶 · 温柔女声 |
+| `Ethan` | 晨煦 · 阳光男声 |
+| `Chelsie` | 千雪 · 二次元少女 |
+| `Dylan` | 北京话 · 晓东 |
+| `Rocky` | 粤语 · 阿强 |
+| `Sunny` | 四川话 · 晴儿 |
+
+完整音色（普通话 / 各地方言 / 双语等）可在 Web 控制台的「模型管理 → 语音合成」下拉框中可视化选择。
+
+## 向量
+
+```json
+{
+  "embedding_provider": "dashscope",
+  "embedding_model": "text-embedding-v4"
+}
+```
+
+默认模型 `text-embedding-v4`。修改 embedding 后需执行 `/memory rebuild-index` 命令重建索引。

docs/releases/overview.mdx

@@ -5,6 +5,7 @@ description: CowAgent 版本更新历史
 
 | 版本 | 日期 | 说明 |
 | --- | --- | --- |
+| [2.0.9](/releases/v2.0.9) | 2026.05.21 | MCP 工具生态接入、模型管理页重构（厂商凭据共享 + 多能力统一调度）、语音系统升级、浏览器持久登录 |
 | [2.0.8](/releases/v2.0.8) | 2026.05.06 | 飞书渠道全面升级（语音、流式输出和Markdown、扫码一键接入）、DeepSeek V4和百度模型新增、定时任务工具增强 |
 | [2.0.7](/releases/v2.0.7) | 2026.04.22 | 图像生成技能（六厂商自动路由）、新模型支持（Kimi K2.6、Claude Opus 4.7、GLM 5.1）、知识库增强、Web 控制台优化 |
 | [2.0.6](/releases/v2.0.6) | 2026.04.14 | 项目更名、知识库系统、梦境记忆蒸馏、上下文智能压缩、Web 控制台多会话及多项优化 |

docs/releases/v2.0.7.mdx

@@ -11,7 +11,7 @@ description: CowAgent 2.0.7 - 图像生成技能（六厂商自动路由）、
 - **开箱即用**：配置 API Key 即可使用，无需手动指定模型。也支持在对话中指定特定模型
 - **灵活控制**：支持 `quality`（画质）、`size`（分辨率，512/1K~4K）、`aspect_ratio`（宽高比）等参数，各厂商自动适配有效值
 - **图片编辑**：传入已有图片即可进行编辑、风格迁移、多图融合
-- **Skill 级配置**：支持通过 `config.json` 中的 `skill.image-generation.model` 固定默认模型
+- **Skill 级配置**：支持通过 `config.json` 中的 `skills.image-generation.model` 固定默认模型
 
 相关文档：[图像生成技能](https://docs.cowagent.ai/skills/image-generation)
 

docs/releases/v2.0.8.mdx

@@ -46,7 +46,7 @@ description: CowAgent 2.0.8 - 飞书渠道全面升级（语音、流式打字
 
 ## 🔧 工具与安全
 
-- **图像识别模型**：让 `tool.vision.model` 配置真正生效，未配置时自动 fallback #2792 Thanks CNXudiandian
+- **图像识别模型**：让 `tools.vision.model` 配置真正生效，未配置时自动 fallback #2792 Thanks CNXudiandian
 - **Bash 安全确认**：仅对工作区外的破坏性删除做二次确认，工作区内常规操作不再打扰
 
 ## 🐛 其他修复

docs/releases/v2.0.9.mdx

@@ -0,0 +1,92 @@
+---
+title: v2.0.9
+description: CowAgent 2.0.9 - MCP 工具生态接入、模型管理页重构、语音系统升级、浏览器持久登录
+---
+
+## 🧩 MCP 工具生态接入
+
+新增 **MCP（Model Context Protocol）** 工具集成，CowAgent 从固定工具集扩展为开放可插拔的工具生态。任何兼容 MCP 协议的服务（高德地图、Chrome DevTools、Filesystem、Playwright 等）都可作为工具直接接入 Agent。
+
+- **零额外依赖**：原生 JSON-RPC 实现，同时支持 `stdio`（本地进程）和 `sse`（远程 URL）两种传输
+- **兼容主流配置**：兼容 Claude Desktop / Cursor 风格的 `mcpServers` 配置，优先读取 `~/cow/mcp.json`，未配置则回退 `config.json`
+- **异步启动**：MCP 服务在后台线程启动，不阻塞 Agent 初始化；单个服务失败不影响整体
+
+相关文档：[MCP 工具](https://docs.cowagent.ai/tools/mcp) · 社区贡献 #2801 Thanks @yangluxin613
+
+## 🖥️ 模型管理页面重构
+
+「模型」页面整体重新设计，从原来按 LLM 厂商堆叠的列表，重构为 **厂商凭据 + 能力调度** 两层结构：一处配置厂商凭据，对话、图像、语音、向量、搜索等多个能力共享。
+
+- **厂商凭据集中管理**：所有支持厂商（OpenAI / Claude / Gemini / DeepSeek / Qwen / 豆包 / Kimi / 智谱 / MiniMax / 千帆 / LinkAI / Custom 等）的 API Key / API Base 在顶部统一维护，编辑后下方所有能力立即生效
+- **能力卡片**：按主模型、图像理解、图像生成、语音识别、语音合成、向量、联网搜索分卡，每个能力可独立选择厂商和模型，未配置时自动跟随主模型或按默认顺序回退
+
+### 多厂商联网搜索
+
+联网搜索升级为多厂商架构，**输出格式统一**：
+
+- 四家可选：博查（bocha）、百度千帆（qianfan）、智谱（zhipu）、LinkAI
+- 两种调度策略：`auto`（按 bocha > qianfan > zhipu > linkai 顺序自动选第一个已配置的厂商）/ `fixed`（固定指定厂商）
+- 配置 ≥2 家且为 `auto` 时，Agent 可在单次调用中临时指定 `provider` 切换搜索源
+
+### 向量厂商热切换
+
+向量（Embedding）支持多厂商，告别对 OpenAI 的单一依赖：
+
+- 原生支持 `openai` / `dashscope` / `doubao` / `zhipu` / `linkai`
+- **在线重建索引**：切换厂商后执行 `/memory rebuild-index`，无需重启、不会中断当前对话
+- 梦境日记默认排除在向量索引之外，避免反复出现在检索结果中干扰对话
+
+## 🎙️ 语音系统升级
+
+- **TTS 适配更多通道**：个人微信（ilink）、钉钉、企微智能机器人现已原生支持语音回复，开关沿用 `always_reply_voice` / `voice_reply_voice`；触发 TTS 时先发文本气泡再发语音消息，方便对照阅读
+- **新增 ASR 厂商**：百炼（DashScope）、智谱
+- **TTS 多厂商重构**：MiniMax / LinkAI / DashScope / 智谱 TTS 在流式合成、长文本切分、错误回退上更稳
+- **网页麦克风输入**：Web 控制台聊天框新增麦克风按钮，可直接录音发送，自动走 ASR 转文本
+
+## 🌐 浏览器工具
+
+浏览器工具支持三种启动模式，告别"每次开会话都得重新登录"：
+
+- **持久化用户配置（默认）**：复用 `~/.cow/browser_profile`，登录一次后下次自动复用登录态
+- **CDP 模式**：通过 `cdp_endpoint` 附加到手动启动的真实 Chrome，享有完整指纹，适合反爬严格的站点
+- **Fresh 模式**：每次清空环境，适合做隔离任务
+
+此外，浏览器被用户中途关闭后下次调用会自动重新拉起，CDP 模式下不会误杀用户的 Chrome 进程。相关文档：[浏览器工具](https://docs.cowagent.ai/tools/browser) #2809
+
+## 🤖 新模型与模型增强
+
+- **百度 ERNIE 5.1**：新增 `ernie-5.1` 模型
+- **DeepSeek V4 reasoning_effort**：DeepSeek V4 系列支持 `reasoning_effort` 配置思考深度
+- **OpenRouter / Vercel AI Gateway 归因**：调用这两个平台时自动注入归因 Header，平台可正确识别 CowAgent 用量
+- 修复 MiMo 等思考模型在多轮对话中 `reasoning_content` 丢失的问题
+
+## 🚀 启动与运行体验
+
+来自社区的多项体验改进 Thanks @yangluxin613
+
+- **自动选端口 + 自动开浏览器**：默认端口被占用时自动切换，启动成功后默认打开控制台
+- **Ctrl+C 干净退出**：不再打印一长串堆栈
+- **日志面板**：差异化级别配色、多行日志继承级别、新增级别筛选 Checkbox
+
+## 🔒 部署与安全
+
+- **默认仅本机访问**：Web 控制台 `web_host` 默认 `127.0.0.1`，避免无密码情况下被外网直接访问；显式 `0.0.0.0` 且未设密码时给出提示
+- **前端资源完全本地化**：第三方 CSS / JS 全部本地分发，离线 / 内网环境也能正常加载控制台 #2816 Thanks @TryToMakeUsBetter
+- **支持文件夹上传**：上传区支持整目录一次性上传，路径校验适配 Windows #2815 Thanks @TryToMakeUsBetter
+
+## 🛠 其他改进与修复
+
+- **定时任务防重复执行**：调度器初始化做幂等处理
+- **工具失败状态持久化**：刷新页面或重载历史时失败的工具调用正确显示失败状态 #2822 Thanks @a1094174619
+- **企微机器人非法字符**：修复消息中包含非法控制字符导致投递失败的问题 #2810 Thanks @Jacques-Zhao
+- **飞书文件消息**：飞书通道支持文件消息接收
+- **工具配置合并**：修复用户自定义工具配置（如 `tools.browser`）被工作区默认值整体覆盖的问题，现按字段合并
+- 修复单文件上传偶发 TypeError、切换语言后 JS 动态视图未重渲染等问题
+
+## 📦 升级方式
+
+源码部署可执行 `cow update` 或 `./run.sh update` 一键升级，或手动拉取代码后重启。详见 [更新升级文档](https://docs.cowagent.ai/guide/upgrade)。
+
+> ⚠️ 切换向量厂商后，建议执行一次 `/memory rebuild-index`，让历史记忆按新的向量维度重新入库。
+
+**发布日期**：2026.05.21 | [Full Changelog](https://github.com/zhayujie/CowAgent/compare/2.0.8...2.0.9)

docs/skills/image-generation.mdx

@@ -3,149 +3,87 @@ title: image-generation - 图像生成
 description: 文生图 / 图生图 / 多图融合，支持多家厂商自动路由与回退
 ---
 
-通用的图像生成与编辑技能，支持 OpenAI、Gemini、Seedream（火山方舟）、Qwen（百炼）、MiniMax、LinkAI 共六家厂商。不需要手动选模型，脚本会按固定优先级自动挑选已配置的厂商来出图。
+通用的图像生成与编辑技能，支持 OpenAI、Gemini、Seedream（火山方舟）、Qwen（百炼）、MiniMax、LinkAI 共六家厂商。配好任意一家的 Key 即可使用，配多家可享受自动回退。
 
-## 模型选择
-
-`image-generation` 采用「固定优先级 + 自动回退」的策略，配好 Key 就能用：
-
-1. **优先级顺序**：`OpenAI → Gemini → Seedream → Qwen → MiniMax → LinkAI`
-2. **没配 Key 的跳过**：只有设了 API Key 的厂商才会参与
-3. **失败自动切下一家**：遇到 401、模型未开通、网络异常等错误时，会自动试下一个
-4. **指定模型时前置**：如果明确传了某个模型名，对应厂商会被提到最前面先试
-
-### 支持的模型
+## 支持的模型
 
 | 厂商 | 模型 / 别名 | 特点 |
 | --- | --- | --- |
-| OpenAI | `gpt-image-2`、`gpt-image-1` | 通用文生图，高质量、高智能，支持 `quality` 参数控制画质 |
+| OpenAI | `gpt-image-2`、`gpt-image-1` | 通用文生图，高质量，支持 `quality` 控制画质 |
 | Gemini Nano Banana | `nano-banana-2`、`nano-banana-pro`、`nano-banana` | 对应 `gemini-3.1-flash`、`gemini-3-pro`、`gemini-2.5-flash` 的图像版本 |
 | Seedream（火山方舟） | `seedream-5.0-lite`、`seedream-4.5` | 原生 2K–4K，最多 14 张图融合 |
 | Qwen（百炼） | `qwen-image-2.0`、`qwen-image-2.0-pro` | 擅长中文排版和图文融合 |
-| MiniMax | `image-01` | 简单快速的图片生成 |
-| LinkAI | 任意模型 | 通用代理，兜底用 |
+| MiniMax | `image-01` | 简单快速 |
+| LinkAI | 任意模型 | 统一网关，作为兜底 |
 
-<Note>
-默认情况下 Agent 不会主动选模型，而是走自动路由。如果你想用某个特定模型，直接在对话里说就行，比如「用 seedream 画一只猫」或「用 gpt-image-2 生成海报」。也可以通过下面的「自定义配置」固定默认模型。
-</Note>
+## 模型选择
 
-## 自定义配置
+默认走「自动路由 + 失败回退」：
 
-### API Key 配置
+1. 按 `OpenAI → Gemini → Seedream → Qwen → MiniMax → LinkAI` 顺序选第一个已配置的厂商
+2. 遇到 401、模型未开通、网络异常等错误时，自动切到下一家
+3. 用户在对话里指定模型时（如「用 seedream 画一只猫」），对应厂商会被提到最前优先尝试
 
-至少需要配**一个**厂商的 Key，配多个就能享受自动回退能力。有三种配置方式：
-
-#### 方式一：已有模型 Key 自动复用
-
-如果你在 web控制台 或 `config.json` 中配置了对话模型的 Key（比如 `openai_api_key`、`gemini_api_key` 等），启动时这些 Key 会被**自动同步**到对应的环境变量。也就是说，只要你的对话模型能用，图像生成就能直接用同一个 Key，不需要额外配置。
-
-#### 方式二：在 config.json 中配置
-
-在 `config.json` 中直接写对应的 Key 字段即可，支持的字段如下：
+如需固定使用某个模型：
 
 ```json
 {
-  "openai_api_key": "sk-xxx",
-  "openai_api_base": "https://api.openai.com/v1",
-  "gemini_api_key": "AIza-xxx",
-  "ark_api_key": "xxx",
-  "dashscope_api_key": "sk-xxx",
-  "minimax_api_key": "xxx",
-  "linkai_api_key": "xxx"
-}
-```
-
-修改后需要重启生效。每个 Key 还有对应的 `*_api_base` 字段可以自定义接口地址。
-
-#### 方式三：对话中直接配置
-
-在对话里发送 API Key，Agent 会通过 `env_config` 工具自动保存到 `~/cow/.env`，**不需要重启**就能生效。例如：
-
-```
-帮我配置 OPENAI_API_KEY 为 sk-xxx
-```
-
-或者：
-
-```
-设置 ARK_API_KEY 为 xxx
-```
-
-### API Key 一览
-
-| 环境变量 | config.json 字段 | 对应厂商 | 默认 Base URL |
-| --- | --- | --- | --- |
-| `OPENAI_API_KEY` | `openai_api_key` | OpenAI | `https://api.openai.com/v1` |
-| `GEMINI_API_KEY` | `gemini_api_key` | Gemini | `https://generativelanguage.googleapis.com` |
-| `ARK_API_KEY` | `ark_api_key` | 火山方舟（Seedream） | `https://ark.cn-beijing.volces.com/api/v3` |
-| `DASHSCOPE_API_KEY` | `dashscope_api_key` | 阿里百炼（Qwen） | `https://dashscope.aliyuncs.com` |
-| `MINIMAX_API_KEY` | `minimax_api_key` | MiniMax | `https://api.minimaxi.com` |
-| `LINKAI_API_KEY` | `linkai_api_key` | LinkAI | `https://api.link-ai.tech` |
-
-
-### 指定默认模型
-
-如果想让所有图像生成固定走某个厂商的模型，可以在 `config.json` 里加：
-
-```json
-"skill": {
+  "skills": {
     "image-generation": {
       "model": "seedream-5.0-lite"
     }
   }
+}
 ```
 
-启动时这段配置会被自动转成环境变量 `SKILL_IMAGE_GENERATION_MODEL`，脚本读到后会固定使用这个模型所在的厂商进行生成。
+## 配置 API Key
+
+<Tip>
+  推荐通过 [Web 控制台](/channels/web) 的「模型管理」页面配置，配好的对话模型 Key 会被图像生成技能自动复用，无需重复配置。也可手动编辑配置文件或在对话中通过 `env_config` 工具临时设置。
+</Tip>
+
+凭证统一复用主模型厂商的 Key：
+
+| 字段 | 对应厂商 |
+| --- | --- |
+| `openai_api_key` | OpenAI |
+| `gemini_api_key` | Gemini |
+| `ark_api_key` | 火山方舟（Seedream） |
+| `dashscope_api_key` | 阿里百炼（Qwen） |
+| `minimax_api_key` | MiniMax |
+| `linkai_api_key` | LinkAI |
 
 
 ## 开启和关闭
 
-`image-generation` 是内置技能，**会根据 API Key 自动调整状态**：
+技能会根据 API Key 自动调整状态：
 
-- **Key 已配置**：技能正常可用，Agent 收到画图请求时会直接调用
-- **Key 未配置**：技能仍然会出现在上下文中（标记为「需要配置」），Agent 会引导用户去配 Key，而不是直接调用失败
+- **已配置 Key**：Agent 收到画图请求时直接调用
+- **未配置 Key**：技能仍会出现在上下文中（标记为「需要配置」），Agent 会引导用户去配 Key
 
-如果想手动控制，也可以用命令：
+如需手动控制：
 
 ```text
-/skill disable image-generation    # 手动关闭（即使有 Key 也不会被调用）
+/skill disable image-generation    # 关闭
 /skill enable image-generation     # 重新开启
 ```
 
-终端里对应的命令是 `cow skill disable image-generation` / `cow skill enable image-generation`。
+终端等价命令：`cow skill disable image-generation` / `cow skill enable image-generation`。
 
 ## 参数
 
 | 参数 | 类型 | 必填 | 默认 | 说明 |
 | --- | --- | --- | --- | --- |
 | `prompt` | string | 是 | — | 图像描述 |
-| `image_url` | string / list | 否 | null | 编辑用的输入图，支持本地路径或 URL。传多个就是多图融合 |
-| `quality` | string | 否 | auto | `low` / `medium` / `high`，只有部分厂商支持 |
-| `size` | string | 否 | auto | `512` / `1K` / `2K` / `3K` / `4K`，也可以写像素值如 `1024x1024` |
+| `image_url` | string / list | 否 | null | 编辑用的输入图，本地路径或 URL；传列表为多图融合 |
+| `quality` | string | 否 | auto | `low` / `medium` / `high`，仅部分厂商支持 |
+| `size` | string | 否 | auto | `512` / `1K` / `2K` / `3K` / `4K`，或像素值如 `1024x1024` |
 | `aspect_ratio` | string | 否 | null | `1:1` / `3:2` / `2:3` / `16:9` / `9:16` / `21:9`；Gemini 还支持 `1:4` / `4:1` / `1:8` / `8:1` |
 
 <Warning>
-**质量越高、分辨率越大，花的钱越多、等的时间越长。**
-
-- 日常对话和快速预览直接用默认（`auto`），或者 `quality=low` + `size=1K`，大概 20 秒出图
-- 做海报、用户明确要高清的时候再上 `quality=high` + `size=2K/4K`，可能要等 1～5 分钟，取决于不同模型的速度
+  **质量越高、分辨率越大，耗时和成本越高。** 日常对话用默认（`auto`）或 `quality=low` + `size=1K` 即可，约 20 秒出图；做海报或明确要高清时再上 `high` + `2K/4K`，可能需要 1–5 分钟。
 </Warning>
 
-## 输出
-
-成功时返回：
-
-```json
-{
-  "model": "doubao-seedream-5-0-260128",
-  "images": [
-    {"url": "/path/to/output.png"}
-  ]
-}
-```
-
-失败时返回 `{ "error": "..." }`。出错后**不要直接重试**——大概率是配置问题（Key 填错、API 地址不对、模型没开通），让用户修好配置再试。
-
 ## 常见用法
 
 - **文生图**：根据描述生成插画、海报、图标、头像、分镜图等
@@ -153,8 +91,8 @@ description: 文生图 / 图生图 / 多图融合，支持多家厂商自动路
 - **多图融合**：把多张参考图合成一张（换装、角色合影等）
 
 <Note>
-- bash 超时建议设 600 秒。单个厂商的 HTTP 超时是 300 秒，但脚本可能依次尝试多个厂商
-- 输入的图片会自动压缩到 4MB 以内、最长边不超过 4096px
-- Gemini / Seedream / Qwen / MiniMax 不支持 `quality` 参数，传了也没用
-- Seedream 默认出 2K 图，`seedream-5.0-lite` 支持到 3K，`seedream-4.5` 支持到 4K
+- bash 超时建议设 600 秒：单厂商 HTTP 超时 300 秒，脚本可能依次尝试多家
+- 输入图片自动压缩到 4MB 以内、最长边不超过 4096px
+- Gemini / Seedream / Qwen / MiniMax 不支持 `quality` 参数
+- Seedream 默认出 2K 图；`seedream-5.0-lite` 支持到 3K，`seedream-4.5` 支持到 4K
 </Note>

docs/tools/vision.mdx

@@ -40,7 +40,7 @@ Vision 工具采用多级自动选择 + 自动兜底策略，无需手动配置
 
 ```json
 {
-    "tool": {
+    "tools": {
         "vision": {
             "model": "gpt-4.1"
         }

docs/tools/web-search.mdx

@@ -1,32 +1,51 @@
 ---
 title: web_search - 联网搜索
-description: 搜索互联网获取实时信息
+description: 搜索互联网获取实时信息，支持多个搜索厂商
 ---
 
-搜索互联网获取实时信息、新闻、研究等内容。支持两个搜索后端，自动选择可用的后端。
+搜索互联网获取实时信息、新闻、研究等内容。支持博查、百度千帆、智谱、LinkAI 四个后端，配置任意一家即可使用。
 
-## 依赖
+<Tip>
+  推荐通过 [Web 控制台](/channels/web) 的「模型管理 → 搜索」面板可视化配置厂商与策略，无需手动编辑配置文件。
+</Tip>
 
-需要配置至少一个搜索 API Key（通过 `env_config` 工具或工作空间 `.env` 文件配置）：
+## 厂商
 
-| 后端 | 环境变量 | 优先级 | 获取方式 |
-| --- | --- | --- | --- |
-| 博查搜索 | `BOCHA_API_KEY` | 优先使用 | [博查开放平台](https://open.bochaai.com/) |
-| LinkAI 搜索 | `LINKAI_API_KEY` | 可选 | [LinkAI 控制台](https://link-ai.tech/console/interface) |
+| 厂商 | 凭证 | 申请入口 |
+| --- | --- | --- |
+| 博查 Bocha | `tools.web_search.bocha_api_key` | [博查开放平台](https://open.bochaai.com/) |
+| 百度千帆 | 复用 `qianfan_api_key` | [千帆控制台](https://cloud.baidu.com/doc/qianfan/s/2mh4su4uy) |
+| 智谱 Zhipu | 复用 `zhipu_ai_api_key` | [智谱开放平台](https://docs.bigmodel.cn/cn/guide/tools/web-search) |
+| LinkAI | 复用 `linkai_api_key` | [LinkAI 控制台](https://link-ai.tech/console/interface) |
 
-## 参数
+除博查需要单独的 `bocha_api_key` 外，其他三家直接复用对应模型的 API Key，配好模型即同时获得搜索能力。
+
+## 路由策略
+
+```json
+{
+  "tools": {
+    "web_search": {
+      "strategy": "auto",
+      "provider": ""
+    }
+  }
+}
+```
+
+- `auto`（默认）：由 Agent 在已配置的厂商中智能选择，并可在一次任务中多次调用、切换不同厂商以获取更全面的结果；未指定时按 `bocha → qianfan → zhipu → linkai` 顺序兜底。
+- `fixed`：固定使用 `provider` 指定的厂商；该厂商凭证缺失时自动回落到 auto 顺序。
+
+## 工具参数
 
 | 参数 | 类型 | 必填 | 说明 |
 | --- | --- | --- | --- |
 | `query` | string | 是 | 搜索关键词 |
-| `count` | integer | 否 | 返回结果数量（1-50，默认 10） |
-| `freshness` | string | 否 | 时间范围：`noLimit`、`oneDay`、`oneWeek`、`oneMonth`、`oneYear`，或日期范围如 `2025-01-01..2025-02-01` |
+| `count` | integer | 否 | 返回结果数量（1–50，默认 10） |
+| `freshness` | string | 否 | 时间范围：`noLimit`（默认）、`oneDay`、`oneWeek`、`oneMonth`、`oneYear`，或日期范围如 `2025-01-01..2025-02-01` |
 | `summary` | boolean | 否 | 是否返回页面摘要（默认 false） |
-
-## 使用场景
-
-当用户询问最新信息、需要事实核查或获取实时数据时，Agent 会自动调用此工具。
+| `provider` | string | 否 | `auto` 策略下配置了多个厂商时可见，用于单次切换厂商 |
 
 <Note>
-  如果未配置任何搜索 API Key，该工具不会被加载。
+  四家凭证均未配置时，该工具不会注册到 Agent。
 </Note>

models/chatgpt/chat_gpt_bot.py

@@ -60,7 +60,7 @@ class ChatGPTBot(Bot, OpenAIImage, OpenAICompatibleBot):
             "timeout": conf().get("request_timeout", None),  # 重试超时时间，在这个时间内，将会自动重试
         }
         # 部分模型暂不支持一些参数，特殊处理
-        if conf_model in [const.O1, const.O1_MINI, const.GPT_5, const.GPT_5_MINI, const.GPT_5_NANO]:
+        if conf_model in [const.O1, const.O1_MINI, const.GPT_5, const.GPT_5_MINI, const.GPT_5_NANO, const.GPT_55]:
             remove_keys = ["temperature", "top_p", "frequency_penalty", "presence_penalty"]
             for key in remove_keys:
                 self.args.pop(key, None)  # 如果键不存在，使用 None 来避免抛出错、

models/gemini/google_gemini_bot.py

@@ -38,9 +38,9 @@ class GoogleGeminiBot(Bot):
 
     @property
     def model(self):
-        model_name = conf().get("model") or "gemini-3.1-pro-preview"
+        model_name = conf().get("model") or "gemini-3.5-flash"
         if model_name == "gemini":
-            model_name = "gemini-3.1-pro-preview"
+            model_name = "gemini-3.5-flash"
         return model_name
 
     @property

models/openai_compatible_bot.py

@@ -89,8 +89,9 @@ class OpenAICompatibleBot:
                     messages[0] = {"role": "system", "content": system_prompt}
             
             # Build request parameters
+            model_name = kwargs.get("model", api_config.get('model', 'gpt-5.4'))
             request_params = {
-                "model": kwargs.get("model", api_config.get('model', 'gpt-3.5-turbo')),
+                "model": model_name,
                 "messages": messages,
                 "temperature": kwargs.get("temperature", api_config.get('default_temperature', 0.9)),
                 "top_p": kwargs.get("top_p", api_config.get('default_top_p', 1.0)),
@@ -98,6 +99,10 @@ class OpenAICompatibleBot:
                 "presence_penalty": kwargs.get("presence_penalty", api_config.get('default_presence_penalty', 0.0)),
                 "stream": stream
             }
+            # GPT-5 / GPT-5.5 / o1 series only accept default temperature/top_p and reject penalty params
+            if model_name in ("gpt-5", "gpt-5-mini", "gpt-5-nano", "gpt-5.5", "o1", "o1-mini"):
+                for key in ("temperature", "top_p", "frequency_penalty", "presence_penalty"):
+                    request_params.pop(key, None)
             
             # Add max_tokens if specified
             if kwargs.get("max_tokens"):

plugins/cow_cli/cow_cli.py

@@ -1056,6 +1056,38 @@ class CowCliPlugin(Plugin):
             logger.warning(f"[CowCli] /memory dream sync failed: {e}")
             return f"❌ 记忆蒸馏失败: {e}"
 
+    @staticmethod
+    def _resolve_active_embedding():
+        """
+        Resolve (provider_label, model, dim) from the LATEST config, not the
+        possibly-stale provider instance cached on a running agent. Used by
+        /memory status and rebuild-index hints so they reflect what a rebuild
+        will actually run as after the user changes embedding_provider.
+        Returns (label, model, dim) where any field may be None when unknown.
+        """
+        from agent.memory.embedding import EMBEDDING_VENDORS
+        from config import conf
+
+        provider_key = (conf().get("embedding_provider") or "").strip().lower()
+        cfg_model = (conf().get("embedding_model") or "").strip()
+        try:
+            cfg_dim = int(conf().get("embedding_dimensions") or 0)
+        except (TypeError, ValueError):
+            cfg_dim = 0
+
+        if not provider_key:
+            # Legacy auto path: openai -> linkai, both default to text-embedding-3-small (1536).
+            if (conf().get("open_ai_api_key") or "").strip():
+                return "openai (legacy)", "text-embedding-3-small", 1536
+            if (conf().get("linkai_api_key") or "").strip():
+                return "linkai (legacy)", "text-embedding-3-small", 1536
+            return "(legacy)", None, None
+
+        meta = EMBEDDING_VENDORS.get(provider_key) or {}
+        model = cfg_model or meta.get("default_model")
+        dim = cfg_dim if cfg_dim > 0 else meta.get("default_dimensions")
+        return provider_key, model, dim
+
     def _memory_status(self) -> str:
         """Show current memory index status."""
         from agent.memory.embedding import detect_index_dim
@@ -1078,15 +1110,14 @@ class CowCliPlugin(Plugin):
         lines.append(f"  Chunks  : {chunks} (embedded: {embedded})")
         lines.append("")
 
-        # Active provider (from running config + provider instance).
+        # Resolve from the latest config so users see what /memory rebuild-index
+        # will actually run as — not what the cached agent was initialized with.
+        cfg_provider, cfg_model, cfg_dim = self._resolve_active_embedding()
         provider_obj = memory_manager.embedding_provider
-        cfg_provider = (conf().get("embedding_provider") or "").strip().lower() or "(legacy)"
-        if provider_obj is not None:
-            cfg_model = getattr(provider_obj, "model", "?")
-            cfg_dim = getattr(provider_obj, "_dimensions", None) or "?"
+        if cfg_model:
             lines.append(f"  Provider : {cfg_provider}")
             lines.append(f"  Model    : {cfg_model}")
-            lines.append(f"  Dim      : {cfg_dim}")
+            lines.append(f"  Dim      : {cfg_dim if cfg_dim else '?'}")
         else:
             lines.append("  Provider : (未初始化, keyword-only)")
 
@@ -1105,7 +1136,6 @@ class CowCliPlugin(Plugin):
                 )
 
             index_dim = detect_index_dim(memory_manager.storage)
-            cfg_dim = getattr(provider_obj, "_dimensions", None)
             if index_dim is not None and cfg_dim and index_dim != cfg_dim:
                 warnings.append(
                     f"  ⚠️ 索引中存量向量为 {index_dim} 维，与当前配置 {cfg_dim} 维不一致；"
@@ -1129,15 +1159,27 @@ class CowCliPlugin(Plugin):
             )
 
         memory_manager = agent.memory_manager
-        if memory_manager.embedding_provider is None:
+
+        # Rebuild against the LATEST config: build a fresh provider from
+        # config.json and swap it onto memory_manager. The agent's
+        # conversation_history and other state are untouched.
+        try:
+            from bridge.agent_initializer import AgentInitializer
+            fresh_provider = AgentInitializer(bridge=None, agent_bridge=None) \
+                ._init_embedding_provider(memory_manager.config, session_id=session_id)
+        except Exception as e:
+            logger.exception("[CowCli] /memory rebuild-index: build provider failed")
+            return f"⚠️ 无法根据当前配置构造 embedding provider: {e}"
+
+        if fresh_provider is None:
             return (
                 "⚠️ 当前没有可用的 embedding provider。\n"
                 "请检查 config.json 中的 embedding 相关配置 (provider / api key)。"
             )
+        memory_manager.embedding_provider = fresh_provider
 
-        provider_obj = memory_manager.embedding_provider
-        model_label = getattr(provider_obj, "model", "?")
-        dim_label = getattr(provider_obj, "dimensions", "?")
+        model_label = getattr(fresh_provider, "model", "?")
+        dim_label = getattr(fresh_provider, "dimensions", "?")
 
         # SaaS (e_context is None): run synchronously, return final result
         if e_context is None:
@@ -1168,7 +1210,7 @@ class CowCliPlugin(Plugin):
         threading.Thread(target=_run, daemon=True).start()
         return (
             f"🔧 索引重建已启动 (model={model_label}, dim={dim_label})\n\n"
-            f"将清空现有 chunks 并重新 embed 所有记忆文件，完成后会通知你。"
+            f"将重新向量化所有记忆和知识文件，完成后会通知你。"
         )
 
     @staticmethod

skills/image-generation/scripts/generate.py

@@ -1110,7 +1110,7 @@ def main():
     # Model resolution priority:
     #   1. Explicit `model` in the call args (agent / user override)
     #   2. SKILL_IMAGE_GENERATION_MODEL env var (synced from
-    #      config["skill"]["image-generation"]["model"] at startup)
+    #      config["skills"]["image-generation"]["model"] at startup)
     #   3. None → fall back to automatic provider routing (try every
     #      provider with a configured API key in global priority order)
     model = args.get("model") or os.environ.get("SKILL_IMAGE_GENERATION_MODEL") or ""

tests/test_qianfan_provider.py

@@ -394,7 +394,7 @@ class TestQianfanVisionTool(unittest.TestCase):
             "open_ai_api_key": "",
             "linkai_api_key": "",
             "use_linkai": False,
-            "tool": {},
+            "tools": {},
         }
         if values:
             data.update(values)
@@ -424,7 +424,7 @@ class TestQianfanVisionTool(unittest.TestCase):
     def test_vision_routes_ernie_model_override_to_qianfan(self):
         fake_conf = self._fake_conf({
             "qianfan_api_key": "test-qianfan-key",
-            "tool": {"vision": {"model": "ernie-4.5-turbo-vl-32k"}},
+            "tools": {"vision": {"model": "ernie-4.5-turbo-vl-32k"}},
         })
         fake_bot = MagicMock()
         fake_bot.call_vision = MagicMock()

voice/dashscope/dashscope_voice.py

@@ -0,0 +1,175 @@
+# encoding:utf-8
+"""DashScope voice: qwen3-asr-flash (ASR) + qwen3-tts-flash (TTS)
+via dashscope.MultiModalConversation."""
+import datetime
+import os
+import random
+from typing import Optional
+
+import dashscope
+import requests
+from dashscope import MultiModalConversation
+
+from bridge.reply import Reply, ReplyType
+from common.log import logger
+from config import conf
+from voice import audio_convert
+from voice.voice import Voice
+
+
+DEFAULT_ASR_MODEL = "qwen3-asr-flash"
+DEFAULT_TTS_MODEL = "qwen3-tts-flash"
+DEFAULT_TTS_VOICE = "Cherry"
+MAX_DURATION_SECONDS = 300
+MAX_FILE_BYTES = 10 * 1024 * 1024
+
+
+class DashScopeVoice(Voice):
+    def __init__(self):
+        pass
+
+    def voiceToText(self, voice_file: str):
+        try:
+            voice_file = self._ensure_compatible_format(voice_file)
+
+            try:
+                size = os.path.getsize(voice_file)
+                if size > MAX_FILE_BYTES:
+                    logger.warning(
+                        f"[DashScopeVoice] audio file {size}B exceeds {MAX_FILE_BYTES}B; "
+                        f"qwen3-asr-flash may reject it"
+                    )
+            except OSError:
+                pass
+
+            api_key = conf().get("dashscope_api_key", "")
+            if not api_key:
+                logger.error("[DashScopeVoice] dashscope_api_key is not configured")
+                return Reply(ReplyType.ERROR, "未配置 DashScope API key")
+            dashscope.api_key = api_key
+
+            model = conf().get("voice_to_text_model") or DEFAULT_ASR_MODEL
+            abs_path = os.path.abspath(voice_file)
+            file_uri = f"file://{abs_path}"
+
+            messages = [
+                {"role": "user", "content": [{"audio": file_uri}]},
+            ]
+            response = MultiModalConversation.call(
+                model=model,
+                messages=messages,
+                result_format="message",
+                asr_options={"enable_itn": False, "enable_lid": True},
+            )
+
+            text = self._extract_text(response)
+            if text is None:
+                logger.error(f"[DashScopeVoice] voiceToText failed: {response}")
+                return Reply(ReplyType.ERROR, "我暂时还无法听清您的语音，请稍后再试吧~")
+
+            logger.info(f"[DashScopeVoice] voiceToText model={model} text={text}")
+            return Reply(ReplyType.TEXT, text)
+        except Exception as e:
+            logger.exception(f"[DashScopeVoice] voiceToText exception: {e}")
+            return Reply(ReplyType.ERROR, "我暂时还无法听清您的语音，请稍后再试吧~")
+
+    def textToVoice(self, text: str):
+        try:
+            api_key = conf().get("dashscope_api_key", "")
+            if not api_key:
+                logger.error("[DashScopeVoice] dashscope_api_key is not configured")
+                return Reply(ReplyType.ERROR, "未配置 DashScope API key")
+            dashscope.api_key = api_key
+
+            model = conf().get("text_to_voice_model") or DEFAULT_TTS_MODEL
+            voice = conf().get("tts_voice_id") or DEFAULT_TTS_VOICE
+            response = MultiModalConversation.call(
+                model=model,
+                api_key=api_key,
+                text=text,
+                voice=voice,
+                stream=False,
+            )
+
+            url = self._extract_audio_url(response)
+            if not url:
+                logger.error(f"[DashScopeVoice] textToVoice failed: {response}")
+                return Reply(ReplyType.ERROR, "语音合成失败")
+
+            local_path = self._download_audio(url)
+            if not local_path:
+                return Reply(ReplyType.ERROR, "语音合成失败")
+
+            logger.info(f"[DashScopeVoice] textToVoice model={model} voice={voice} file={local_path}")
+            return Reply(ReplyType.VOICE, local_path)
+        except Exception as e:
+            logger.exception(f"[DashScopeVoice] textToVoice exception: {e}")
+            return Reply(ReplyType.ERROR, "语音合成失败")
+
+    @staticmethod
+    def _extract_audio_url(response) -> Optional[str]:
+        try:
+            if getattr(response, "status_code", 200) != 200:
+                return None
+            audio = response.output.get("audio") if response.output else None
+            if isinstance(audio, dict):
+                return audio.get("url") or None
+            return getattr(audio, "url", None)
+        except Exception:
+            return None
+
+    @staticmethod
+    def _download_audio(url: str) -> Optional[str]:
+        try:
+            tmp_dir = os.path.join(os.getcwd(), "tmp")
+            os.makedirs(tmp_dir, exist_ok=True)
+            ts = datetime.datetime.now().strftime("%Y%m%d%H%M%S")
+            ext = os.path.splitext(url.split("?", 1)[0])[1].lower() or ".wav"
+            if ext not in (".mp3", ".wav", ".m4a", ".aac", ".opus"):
+                ext = ".wav"
+            dst = os.path.join(tmp_dir, f"dashscope_tts_{ts}_{random.randint(0, 9999)}{ext}")
+            resp = requests.get(url, timeout=60)
+            resp.raise_for_status()
+            with open(dst, "wb") as f:
+                f.write(resp.content)
+            return dst
+        except Exception as e:
+            logger.error(f"[DashScopeVoice] download audio failed: {e}")
+            return None
+
+    @staticmethod
+    def _ensure_compatible_format(voice_file: str) -> str:
+        # qwen3-asr-flash doesn't accept AMR/SILK; mp3/wav/m4a/aac/opus pass through.
+        lower = voice_file.lower()
+        if lower.endswith(".amr") or lower.endswith(".silk") or lower.endswith(".slk"):
+            try:
+                mp3_file = os.path.splitext(voice_file)[0] + ".mp3"
+                audio_convert.any_to_mp3(voice_file, mp3_file)
+                return mp3_file
+            except Exception as e:
+                logger.warning(f"[DashScopeVoice] mp3 convert failed: {e}")
+        return voice_file
+
+    @staticmethod
+    def _extract_text(response) -> Optional[str]:
+        try:
+            if getattr(response, "status_code", 200) != 200:
+                return None
+            choices = response.output.get("choices") or []
+            if not choices:
+                return None
+            content = choices[0].get("message", {}).get("content")
+            if isinstance(content, str):
+                return content.strip() or None
+            if isinstance(content, list):
+                parts = []
+                for item in content:
+                    if isinstance(item, dict) and "text" in item:
+                        parts.append(item["text"])
+                    elif isinstance(item, str):
+                        parts.append(item)
+                text = "".join(parts).strip()
+                return text or None
+            return None
+        except Exception:
+            return None

voice/factory.py

@@ -58,4 +58,12 @@ def create_voice(voice_type):
         from voice.minimax.minimax_voice import MinimaxVoice
 
         return MinimaxVoice()
+    elif voice_type == "dashscope":
+        from voice.dashscope.dashscope_voice import DashScopeVoice
+
+        return DashScopeVoice()
+    elif voice_type == "zhipu" or voice_type == "zhipuai":
+        from voice.zhipuai.zhipuai_voice import ZhipuAIVoice
+
+        return ZhipuAIVoice()
     raise RuntimeError

voice/linkai/linkai_voice.py

@@ -1,16 +1,18 @@
-"""
-google voice service
-"""
+"""LinkAI voice: Whisper ASR + multi-vendor TTS (OpenAI / Doubao / Baidu)
+proxied via https://docs.link-ai.tech/platform/api/voice-speech."""
+import datetime
+import os
 import random
+
 import requests
-from voice import audio_convert
+
 from bridge.reply import Reply, ReplyType
+from common import const
 from common.log import logger
 from config import conf
+from voice import audio_convert
 from voice.voice import Voice
-from common import const
-import os
-import datetime
+
 
 class LinkAIVoice(Voice):
     def __init__(self):
@@ -21,8 +23,7 @@ class LinkAIVoice(Voice):
         try:
             url = conf().get("linkai_api_base", "https://api.link-ai.tech") + "/v1/audio/transcriptions"
             headers = {"Authorization": "Bearer " + conf().get("linkai_api_key")}
-            model = None
-            if not conf().get("text_to_voice") or conf().get("voice_to_text") == "openai":
+            # Pin whisper-1: gateway ignores any other ASR model id.
             model = const.WHISPER_1
             if voice_file.endswith(".amr"):
                 try:
@@ -30,54 +31,59 @@ class LinkAIVoice(Voice):
                     audio_convert.any_to_mp3(voice_file, mp3_file)
                     voice_file = mp3_file
                 except Exception as e:
-                    logger.warn(f"[LinkVoice] amr file transfer failed, directly send amr voice file: {format(e)}")
-            file = open(voice_file, "rb")
-            file_body = {
-                "file": file
-            }
-            data = {
-                "model": model
-            }
-            res = requests.post(url, files=file_body, headers=headers, data=data, timeout=(5, 60))
-            if res.status_code == 200:
-                text = res.json().get("text")
-            else:
-                res_json = res.json()
-                logger.error(f"[LinkVoice] voiceToText error, status_code={res.status_code}, msg={res_json.get('message')}")
+                    logger.warning(f"[LinkVoice] amr file transfer failed, directly send amr voice file: {e}")
+            with open(voice_file, "rb") as file:
+                res = requests.post(
+                    url,
+                    files={"file": file},
+                    headers=headers,
+                    data={"model": model},
+                    timeout=(5, 60),
+                )
+            if res.status_code != 200:
+                msg = ""
+                try:
+                    msg = res.json().get("message", "")
+                except Exception:
+                    pass
+                logger.error(f"[LinkVoice] voiceToText error, status_code={res.status_code}, msg={msg}")
                 return None
-            reply = Reply(ReplyType.TEXT, text)
+            text = res.json().get("text")
             logger.info(f"[LinkVoice] voiceToText success, text={text}, file name={voice_file}")
+            return Reply(ReplyType.TEXT, text)
         except Exception as e:
             logger.error(e)
             return None
-        return reply
 
     def textToVoice(self, text):
         try:
             url = conf().get("linkai_api_base", "https://api.link-ai.tech") + "/v1/audio/speech"
             headers = {"Authorization": "Bearer " + conf().get("linkai_api_key")}
-            model = const.TTS_1
-            if not conf().get("text_to_voice") or conf().get("text_to_voice") in ["openai", const.TTS_1, const.TTS_1_HD]:
-                model = conf().get("text_to_voice_model") or const.TTS_1
+            # Gateway routes by `model` (tts-1 / doubao / baidu) + `voice` from
+            # that engine's catalog. `app_code` is optional workspace override.
             data = {
-                "model": model,
                 "input": text,
                 "voice": conf().get("tts_voice_id"),
-                "app_code": conf().get("linkai_app_code")
+                "app_code": conf().get("linkai_app_code"),
             }
+            model = conf().get("text_to_voice_model")
+            if model:
+                data["model"] = model
             res = requests.post(url, headers=headers, json=data, timeout=(5, 120))
-            if res.status_code == 200:
+            if res.status_code != 200:
+                msg = ""
+                try:
+                    msg = res.json().get("message", "")
+                except Exception:
+                    pass
+                logger.error(f"[LinkVoice] textToVoice error, status_code={res.status_code}, msg={msg}")
+                return None
             tmp_file_name = "tmp/" + datetime.datetime.now().strftime('%Y%m%d%H%M%S') + str(random.randint(0, 1000)) + ".mp3"
+            os.makedirs(os.path.dirname(tmp_file_name), exist_ok=True)
             with open(tmp_file_name, 'wb') as f:
                 f.write(res.content)
-                reply = Reply(ReplyType.VOICE, tmp_file_name)
-                logger.info(f"[LinkVoice] textToVoice success, input={text}, model={model}, voice_id={data.get('voice')}")
-                return reply
-            else:
-                res_json = res.json()
-                logger.error(f"[LinkVoice] textToVoice error, status_code={res.status_code}, msg={res_json.get('message')}")
-                return None
+            logger.info(f"[LinkVoice] textToVoice success, input={text}, voice_id={data.get('voice')}")
+            return Reply(ReplyType.VOICE, tmp_file_name)
         except Exception as e:
             logger.error(e)
-            # reply = Reply(ReplyType.ERROR, "遇到了一点小问题，请稍后再问我吧")
             return None

voice/minimax/minimax_voice.py

@@ -1,8 +1,7 @@
 # encoding:utf-8
-"""
-MiniMax TTS voice service
-"""
+"""MiniMax TTS via /v1/t2a_v2 (SSE stream, hex-encoded mp3 chunks)."""
 import datetime
+import json
 import random
 import requests
 
@@ -12,24 +11,12 @@ from config import conf
 from voice.voice import Voice
 
 
-MINIMAX_TTS_VOICES = [
-    "English_Graceful_Lady",
-    "English_Insightful_Speaker",
-    "English_radiant_girl",
-    "English_Persuasive_Man",
-    "English_Lucky_Robot",
-    "English_expressive_narrator",
-    "Chinese_Warm_Woman",
-    "Chinese_Gentle_Man",
-]
-
-
 class MinimaxVoice(Voice):
     def __init__(self):
         self.api_key = conf().get("minimax_api_key")
-        self.api_base = conf().get("minimax_api_base") or "https://api.minimax.io"
-        # Strip trailing /v1 if present so we can always append /v1/t2a_v2
-        self.api_base = self.api_base.rstrip("/")
+        # Mainland endpoint matches `sk-api-0-...` keys; override via
+        # `minimax_api_base` for international (api.minimax.io) workspaces.
+        self.api_base = (conf().get("minimax_api_base") or "https://api.minimaxi.com").rstrip("/")
         if self.api_base.endswith("/v1"):
             self.api_base = self.api_base[:-3]
 
@@ -68,12 +55,14 @@ class MinimaxVoice(Voice):
             response = requests.post(url, headers=headers, json=payload, stream=True, timeout=60)
             response.raise_for_status()
 
-            # Parse SSE stream and collect hex-encoded audio chunks
+            # MiniMax returns HTTP 200 even on errors; capture base_resp for diagnostics.
             audio_chunks = []
-            buffer = ""
+            last_base_resp = None
+            event_count = 0
             for raw in response.iter_lines():
                 if not raw:
                     continue
+                event_count += 1
                 line = raw.decode("utf-8") if isinstance(raw, bytes) else raw
                 if not line.startswith("data:"):
                     continue
@@ -81,16 +70,31 @@ class MinimaxVoice(Voice):
                 if not json_str or json_str == "[DONE]":
                     continue
                 try:
-                    import json
                     event_data = json.loads(json_str)
-                    audio_hex = event_data.get("data", {}).get("audio")
-                    if audio_hex:
-                        audio_chunks.append(bytes.fromhex(audio_hex))
                 except Exception:
                     continue
+                base_resp = event_data.get("base_resp") or {}
+                if base_resp:
+                    last_base_resp = base_resp
+                audio_hex = (event_data.get("data") or {}).get("audio")
+                if audio_hex:
+                    try:
+                        audio_chunks.append(bytes.fromhex(audio_hex))
+                    except Exception as e:
+                        logger.warning(f"[MINIMAX] skip bad audio hex chunk: {e}")
 
             if not audio_chunks:
-                logger.error("[MINIMAX] TTS returned no audio data")
+                ct = response.headers.get("Content-Type", "")
+                if last_base_resp and last_base_resp.get("status_code") not in (None, 0):
+                    logger.error(
+                        f"[MINIMAX] TTS failed: status_code={last_base_resp.get('status_code')}, "
+                        f"status_msg={last_base_resp.get('status_msg')}, model={model}, voice_id={voice_id}"
+                    )
+                else:
+                    logger.error(
+                        f"[MINIMAX] TTS returned no audio data, model={model}, voice_id={voice_id}, "
+                        f"url={url}, http={response.status_code}, content_type={ct!r}, events={event_count}"
+                    )
                 return Reply(ReplyType.ERROR, "语音合成失败，未获取到音频数据")
 
             audio_data = b"".join(audio_chunks)

voice/openai/openai_voice.py

@@ -31,7 +31,8 @@ class OpenaiVoice(Voice):
                 "file": file,
             }
             data = {
-                "model": "whisper-1",
+                # Override via `voice_to_text_model` (e.g. fall back to whisper-1).
+                "model": conf().get("voice_to_text_model") or "gpt-4o-mini-transcribe",
             }
             response = requests.post(url, headers=headers, files=files, data=data)
             response_data = response.json()

voice/zhipuai/zhipuai_voice.py

@@ -0,0 +1,173 @@
+# encoding:utf-8
+"""ZhipuAI voice: glm-asr-2512 (ASR) + glm-tts (TTS) via BigModel REST API."""
+import datetime
+import os
+import random
+
+import requests
+
+from bridge.reply import Reply, ReplyType
+from common.log import logger
+from config import conf
+from voice import audio_convert
+from voice.voice import Voice
+
+
+DEFAULT_ASR_MODEL = "glm-asr-2512"
+DEFAULT_TTS_MODEL = "glm-tts"
+DEFAULT_TTS_VOICE = "tongtong"
+DEFAULT_API_BASE = "https://open.bigmodel.cn/api/paas/v4"
+MAX_FILE_BYTES = 25 * 1024 * 1024
+REQUEST_TIMEOUT = (5, 60)
+
+
+class ZhipuAIVoice(Voice):
+    def __init__(self):
+        pass
+
+    def voiceToText(self, voice_file: str):
+        try:
+            voice_file = self._ensure_compatible_format(voice_file)
+
+            try:
+                size = os.path.getsize(voice_file)
+                if size > MAX_FILE_BYTES:
+                    logger.warning(
+                        f"[ZhipuAIVoice] audio file {size}B exceeds {MAX_FILE_BYTES}B; "
+                        f"glm-asr-2512 may reject it"
+                    )
+            except OSError:
+                pass
+
+            api_key = conf().get("zhipu_ai_api_key", "")
+            if not api_key:
+                logger.error("[ZhipuAIVoice] zhipu_ai_api_key is not configured")
+                return Reply(ReplyType.ERROR, "未配置 ZhipuAI API key")
+
+            api_base = (conf().get("zhipu_ai_api_base") or DEFAULT_API_BASE).rstrip("/")
+            url = f"{api_base}/audio/transcriptions"
+            model = conf().get("voice_to_text_model") or DEFAULT_ASR_MODEL
+
+            with open(voice_file, "rb") as f:
+                files = {"file": (os.path.basename(voice_file), f)}
+                data = {"model": model, "stream": "false"}
+                headers = {"Authorization": f"Bearer {api_key}"}
+                response = requests.post(
+                    url, headers=headers, files=files, data=data, timeout=REQUEST_TIMEOUT
+                )
+
+            if response.status_code != 200:
+                logger.error(
+                    f"[ZhipuAIVoice] voiceToText failed: status={response.status_code} "
+                    f"body={response.text[:500]}"
+                )
+                return Reply(ReplyType.ERROR, "我暂时还无法听清您的语音，请稍后再试吧~")
+
+            payload = response.json()
+            text = (payload.get("text") or "").strip()
+            if not text:
+                logger.error(f"[ZhipuAIVoice] voiceToText empty text: {payload}")
+                return Reply(ReplyType.ERROR, "我暂时还无法听清您的语音，请稍后再试吧~")
+
+            logger.info(f"[ZhipuAIVoice] voiceToText model={model} text={text}")
+            return Reply(ReplyType.TEXT, text)
+        except Exception as e:
+            logger.exception(f"[ZhipuAIVoice] voiceToText exception: {e}")
+            return Reply(ReplyType.ERROR, "我暂时还无法听清您的语音，请稍后再试吧~")
+
+    def textToVoice(self, text: str):
+        try:
+            api_key = conf().get("zhipu_ai_api_key", "")
+            if not api_key:
+                logger.error("[ZhipuAIVoice] zhipu_ai_api_key is not configured")
+                return Reply(ReplyType.ERROR, "未配置 ZhipuAI API key")
+
+            api_base = (conf().get("zhipu_ai_api_base") or DEFAULT_API_BASE).rstrip("/")
+            url = f"{api_base}/audio/speech"
+            model = conf().get("text_to_voice_model") or DEFAULT_TTS_MODEL
+            voice_id = conf().get("tts_voice_id") or DEFAULT_TTS_VOICE
+
+            payload = {
+                "model": model,
+                "input": text,
+                "voice": voice_id,
+                "response_format": "wav",
+                "speed": 1.0,
+                "volume": 1.0,
+            }
+            headers = {
+                "Authorization": f"Bearer {api_key}",
+                "Content-Type": "application/json",
+            }
+            response = requests.post(
+                url, headers=headers, json=payload, timeout=REQUEST_TIMEOUT
+            )
+
+            if response.status_code != 200:
+                logger.error(
+                    f"[ZhipuAIVoice] textToVoice failed: status={response.status_code} "
+                    f"body={response.text[:500]} model={model} voice={voice_id}"
+                )
+                return Reply(ReplyType.ERROR, "语音合成失败，请稍后再试")
+
+            # Some errors come back as JSON / SSE with HTTP 200.
+            ct = response.headers.get("Content-Type", "")
+            if "application/json" in ct or "text/event-stream" in ct:
+                try:
+                    err = response.json()
+                except Exception:
+                    err = {"raw": response.text[:500]}
+                logger.error(
+                    f"[ZhipuAIVoice] textToVoice unexpected text response "
+                    f"(content_type={ct}): {err}"
+                )
+                return Reply(ReplyType.ERROR, "语音合成失败，请稍后再试")
+
+            audio_bytes = response.content
+            ext = self._sniff_audio_ext(audio_bytes) or "wav"
+
+            file_name = (
+                "tmp/" + datetime.datetime.now().strftime("%Y%m%d%H%M%S")
+                + str(random.randint(0, 1000)) + "." + ext
+            )
+            os.makedirs(os.path.dirname(file_name), exist_ok=True)
+            with open(file_name, "wb") as f:
+                f.write(audio_bytes)
+            logger.info(
+                f"[ZhipuAIVoice] textToVoice model={model} voice={voice_id} "
+                f"file={file_name} bytes={len(audio_bytes)} ext={ext}"
+            )
+            return Reply(ReplyType.VOICE, file_name)
+        except Exception as e:
+            logger.exception(f"[ZhipuAIVoice] textToVoice exception: {e}")
+            return Reply(ReplyType.ERROR, "语音合成失败，请稍后再试")
+
+    @staticmethod
+    def _sniff_audio_ext(data: bytes) -> str:
+        """Detect audio container by magic bytes; returns '' on unknown."""
+        if len(data) < 12:
+            return ""
+        head = data[:12]
+        if head[:4] == b"RIFF" and head[8:12] == b"WAVE":
+            return "wav"
+        if head[:3] == b"ID3" or head[:2] == b"\xff\xfb" or head[:2] == b"\xff\xf3" or head[:2] == b"\xff\xf2":
+            return "mp3"
+        if head[:4] == b"OggS":
+            return "ogg"
+        if head[:4] == b"fLaC":
+            return "flac"
+        return ""
+
+    @staticmethod
+    def _ensure_compatible_format(voice_file: str) -> str:
+        # glm-asr-2512 only accepts .wav / .mp3
+        lower = voice_file.lower()
+        if lower.endswith(".mp3") or lower.endswith(".wav"):
+            return voice_file
+        try:
+            mp3_file = os.path.splitext(voice_file)[0] + ".mp3"
+            audio_convert.any_to_mp3(voice_file, mp3_file)
+            return mp3_file
+        except Exception as e:
+            logger.warning(f"[ZhipuAIVoice] mp3 convert failed: {e}")
+            return voice_file