fix(skill_manager): use patchable SKILLS_DIR in _find_skill

gateway/platforms/api_server.py

@@ -7,6 +7,7 @@
 - GET  /v1/responses/{response_id} — Retrieve a stored response
 - DELETE /v1/responses/{response_id} — Delete a stored response
 - GET  /v1/models                  — lists hermes-agent as an available model
+- GET  /v1/sessions                — lists sessions from the shared DB (source/limit/offset)
 - GET  /health                     — health check
 
 Any OpenAI-compatible frontend (Open WebUI, LobeChat, LibreChat,
@@ -400,6 +401,7 @@ def _create_agent(
         session_id: Optional[str] = None,
         stream_delta_callback=None,
         tool_progress_callback=None,
+        model_override: Optional[str] = None,
     ) -> Any:
         """
         Create an AIAgent instance using the gateway's runtime config.
@@ -408,13 +410,21 @@ def _create_agent(
         base_url, etc. from config.yaml / env vars.  Toolsets are resolved
         from config.yaml platform_toolsets.api_server (same as all other
         gateway platforms), falling back to the hermes-api-server default.
+
+        If *model_override* is provided and is not "hermes-agent", it is used
+        as the model instead of the value from config.yaml.  This lets Open
+        WebUI (or any OpenAI-compatible frontend) select the underlying LLM
+        via the ``model`` field in the chat completions request.
         """
         from run_agent import AIAgent
         from gateway.run import _resolve_runtime_agent_kwargs, _resolve_gateway_model, _load_gateway_config
         from hermes_cli.tools_config import _get_platform_tools
 
         runtime_kwargs = _resolve_runtime_agent_kwargs()
-        model = _resolve_gateway_model()
+        if model_override and model_override not in ("hermes-agent",):
+            model = model_override
+        else:
+            model = _resolve_gateway_model()
 
         user_config = _load_gateway_config()
         enabled_toolsets = sorted(_get_platform_tools(user_config, "api_server"))
@@ -446,25 +456,71 @@ async def _handle_health(self, request: "web.Request") -> "web.Response":
         return web.json_response({"status": "ok", "platform": "hermes-agent"})
 
     async def _handle_models(self, request: "web.Request") -> "web.Response":
-        """GET /v1/models — return hermes-agent as an available model."""
+        """GET /v1/models — return hermes-agent plus per-provider model list."""
         auth_err = self._check_auth(request)
         if auth_err:
             return auth_err
 
-        return web.json_response({
-            "object": "list",
-            "data": [
-                {
-                    "id": "hermes-agent",
-                    "object": "model",
-                    "created": int(time.time()),
-                    "owned_by": "hermes",
-                    "permission": [],
-                    "root": "hermes-agent",
-                    "parent": None,
-                }
-            ],
-        })
+        now = int(time.time())
+
+        # Always include the default hermes-agent entry (uses config.yaml model)
+        models = [
+            {
+                "id": "hermes-agent",
+                "object": "model",
+                "created": now,
+                "owned_by": "hermes",
+                "permission": [],
+                "root": "hermes-agent",
+                "parent": None,
+                "description": "Default model from config.yaml",
+            }
+        ]
+
+        # Add per-provider models based on which API keys are configured
+        try:
+            from hermes_cli.models import _PROVIDER_MODELS
+            import os
+
+            _provider_env: list[tuple[str, str]] = [
+                ("anthropic",    "ANTHROPIC_API_KEY"),
+                ("openrouter",   "OPENROUTER_API_KEY"),
+                ("nous",         "NOUS_API_KEY"),
+                ("deepseek",     "DEEPSEEK_API_KEY"),
+                ("zai",          "GLM_API_KEY"),
+                ("kimi-coding",  "KIMI_API_KEY"),
+                ("minimax",      "MINIMAX_API_KEY"),
+                ("opencode-zen", "OPENCODE_ZEN_API_KEY"),
+                ("opencode-go",  "OPENCODE_GO_API_KEY"),
+            ]
+
+            seen: set[str] = {"hermes-agent"}
+            for provider, env_var in _provider_env:
+                if not os.getenv(env_var):
+                    continue
+                for model_id in _PROVIDER_MODELS.get(provider, []):
+                    if model_id in seen:
+                        continue
+                    seen.add(model_id)
+                    # Normalise to provider/model format for anthropic native
+                    display_id = (
+                        f"anthropic/{model_id}"
+                        if provider == "anthropic" and "/" not in model_id
+                        else model_id
+                    )
+                    models.append({
+                        "id": display_id,
+                        "object": "model",
+                        "created": now,
+                        "owned_by": provider,
+                        "permission": [],
+                        "root": display_id,
+                        "parent": None,
+                    })
+        except Exception:
+            pass  # Fall back to hermes-agent only
+
+        return web.json_response({"object": "list", "data": models})
 
     async def _handle_chat_completions(self, request: "web.Request") -> "web.Response":
         """POST /v1/chat/completions — OpenAI Chat Completions format."""
@@ -571,6 +627,7 @@ def _on_tool_progress(name, preview, args):
                 stream_delta_callback=_on_delta,
                 tool_progress_callback=_on_tool_progress,
                 agent_ref=agent_ref,
+                model_override=model_name,
             ))
 
             return await self._write_sse_chat_completion(
@@ -585,6 +642,7 @@ async def _compute_completion():
                 conversation_history=history,
                 ephemeral_system_prompt=system_prompt,
                 session_id=session_id,
+                model_override=model_name,
             )
 
         idempotency_key = request.headers.get("Idempotency-Key")
@@ -1180,6 +1238,50 @@ async def _handle_run_job(self, request: "web.Request") -> "web.Response":
         except Exception as e:
             return web.json_response({"error": str(e)}, status=500)
 
+    async def _handle_sessions(self, request: "web.Request") -> "web.Response":
+        """GET /v1/sessions — list sessions from the shared DB."""
+        auth_err = self._check_auth(request)
+        if auth_err:
+            return auth_err
+
+        source_param = request.rel_url.query.get("source") or None
+        try:
+            limit = min(int(request.rel_url.query.get("limit", 50)), 200)
+        except (ValueError, TypeError):
+            limit = 50
+        try:
+            offset = max(int(request.rel_url.query.get("offset", 0)), 0)
+        except (ValueError, TypeError):
+            offset = 0
+
+        db = self._ensure_session_db()
+        if db is None:
+            return web.json_response({"object": "list", "data": [], "count": 0})
+
+        try:
+            sessions = db.list_sessions_rich(
+                source=source_param,
+                exclude_sources=["tool"],
+                limit=limit,
+                offset=offset,
+            )
+        except Exception as e:
+            logger.warning("list_sessions_rich failed: %s", e)
+            return web.json_response({"error": str(e)}, status=500)
+
+        data = [
+            {
+                "id": s.get("id"),
+                "title": s.get("title"),
+                "preview": s.get("preview"),
+                "last_active": s.get("last_active"),
+                "source": s.get("source"),
+                "message_count": s.get("message_count"),
+            }
+            for s in sessions
+        ]
+        return web.json_response({"object": "list", "data": data, "count": len(data)})
+
     # ------------------------------------------------------------------
     # Output extraction helper
     # ------------------------------------------------------------------
@@ -1245,6 +1347,7 @@ async def _run_agent(
         stream_delta_callback=None,
         tool_progress_callback=None,
         agent_ref: Optional[list] = None,
+        model_override: Optional[str] = None,
     ) -> tuple:
         """
         Create an agent and run a conversation in a thread executor.
@@ -1265,6 +1368,7 @@ def _run():
                 session_id=session_id,
                 stream_delta_callback=stream_delta_callback,
                 tool_progress_callback=tool_progress_callback,
+                model_override=model_override,
             )
             if agent_ref is not None:
                 agent_ref[0] = agent
@@ -1298,6 +1402,7 @@ async def connect(self) -> bool:
             self._app.router.add_get("/health", self._handle_health)
             self._app.router.add_get("/v1/health", self._handle_health)
             self._app.router.add_get("/v1/models", self._handle_models)
+            self._app.router.add_get("/v1/sessions", self._handle_sessions)
             self._app.router.add_post("/v1/chat/completions", self._handle_chat_completions)
             self._app.router.add_post("/v1/responses", self._handle_responses)
             self._app.router.add_get("/v1/responses/{response_id}", self._handle_get_response)

hermes_cli/commands.py

@@ -128,6 +128,8 @@ class CommandDef:
     CommandDef("commands", "Browse all commands and skills (paginated)", "Info",
                gateway_only=True, args_hint="[page]"),
     CommandDef("help", "Show available commands", "Info"),
+    CommandDef("keys", "Show keyboard shortcuts", "Info",
+               cli_only=True, aliases=("shortcuts",)),
     CommandDef("usage", "Show token usage for the current session", "Info"),
     CommandDef("insights", "Show usage insights and analytics", "Info",
                args_hint="[days]"),

hermes_cli/main.py

@@ -3983,6 +3983,18 @@ def main():
         metavar="SESSION_NAME",
         help="Resume a session by name, or the most recent if no name given"
     )
+    parser.add_argument(
+        "-m", "--model",
+        default=None,
+        metavar="MODEL",
+        help="Model to use for this session (e.g. anthropic/claude-sonnet-4-6)"
+    )
+    parser.add_argument(
+        "--provider",
+        default=None,
+        metavar="PROVIDER",
+        help="Inference provider (e.g. anthropic, openrouter)"
+    )
     parser.add_argument(
         "--worktree", "-w",
         action="store_true",
@@ -5375,10 +5387,11 @@ def cmd_acp(args):
     if (args.resume or args.continue_last) and args.command is None:
         args.command = "chat"
         args.query = None
-        args.model = None
-        args.provider = None
-        args.toolsets = None
-        args.verbose = False
+        # model and provider already set from root parser — don't stomp them
+        if not hasattr(args, "toolsets"):
+            args.toolsets = None
+        if not hasattr(args, "verbose"):
+            args.verbose = False
         if not hasattr(args, "worktree"):
             args.worktree = False
         cmd_chat(args)
@@ -5387,10 +5400,11 @@ def cmd_acp(args):
     # Default to chat if no command specified
     if args.command is None:
         args.query = None
-        args.model = None
-        args.provider = None
-        args.toolsets = None
-        args.verbose = False
+        # model and provider already set from root parser — don't stomp them
+        if not hasattr(args, "toolsets"):
+            args.toolsets = None
+        if not hasattr(args, "verbose"):
+            args.verbose = False
         args.resume = None
         args.continue_last = None
         if not hasattr(args, "worktree"):