feat: expose /internal/models/raw for authoritative model metadata

Lets callers see Lingma's raw config/queryModels response, so the
official per-key displayName/description is discoverable without
reverse-engineering the VSIX. Falls back to the pool's pick() unless
a specific instance is requested.

Made-with: Cursor

app/main.py

@@ -633,6 +633,36 @@ async def internal_auto_login_status():
     return {"ok": True, "instances": out}
 
 
+@app.get("/internal/models/raw", dependencies=[Depends(auth_guard)])
+async def internal_models_raw(instance: str | None = None):
+    """Return the raw `config/queryModels` response from Lingma.
+
+    This is the authoritative source for per-key displayName, description,
+    capability flags, etc. We only ever extract `key` + `displayName` for
+    OpenAI compatibility, but clients may want to inspect everything.
+    """
+    p = _require_pool()
+    target = None
+    if instance:
+        for inst in p.instances:
+            if inst.name == instance:
+                target = inst
+                break
+        if target is None:
+            raise HTTPException(status_code=404, detail={"error": f"instance {instance} not found"})
+    else:
+        target = p.pick()
+    await _ensure_instance_logged_in(target)
+    raw = await target.client.query_models()
+    name_map = build_model_name_map(raw if isinstance(raw, dict) else {})
+    return {
+        "instance": target.name,
+        "raw": raw,
+        "extracted_name_map": name_map,
+        "exposed_keys": flatten_model_keys(raw if isinstance(raw, dict) else {}),
+    }
+
+
 @app.get("/internal/stats", dependencies=[Depends(auth_guard)])
 async def internal_stats():
     p = _require_pool()