fix: native gem

Author: sagitchu

README.md

@@ -158,6 +158,13 @@ gemini:
       default_temperature: 0.2
       top_p: 0.8
       max_output_tokens: 1024
+
+    # Native Gem (Use Gems created on gemini.google.com)
+    - id: "my-native-gem"
+      model: "gemini-2.0-flash"
+      is_native: true  # This flag passes the 'id' to Google's servers
+      default_temperature: 1.0 # Parameters can still be overridden locally
+      max_output_tokens: 8192
 ```
 
 Example request (Chat Completions):

README.zh.md

@@ -158,6 +158,13 @@ gemini:
       default_temperature: 0.2
       top_p: 0.8
       max_output_tokens: 1024
+
+    # 原生 Gem (使用在 gemini.google.com 上创建的 Gems)
+    - id: "my-native-gem"
+      model: "gemini-2.0-flash"
+      is_native: true  # 此标志会将 'id' 传递给 Google 服务器
+      default_temperature: 1.0 # 仍然可以在本地覆盖参数
+      max_output_tokens: 8192
 ```
 
 请求示例（Chat Completions）：

app/server/chat.py

@@ -709,6 +709,8 @@ async def create_chat_completion(
         if gem.system_prompt:
             _inject_gem_system_prompt(request.messages, gem.system_prompt)
 
+    native_gem_id = gem_id if (gem and gem.is_native) else None
+
     try:
         model = Model.from_name(actual_model)
     except ValueError as exc:
@@ -733,6 +735,7 @@ async def create_chat_completion(
         model,
         request.messages,
         gem_id=gem_id,
+        native_gem_id=native_gem_id,
     )
 
     # Keep track of whether we are using a reused session for fallback logic
@@ -767,7 +770,7 @@ async def create_chat_completion(
         try:
             client = await pool.acquire()
             logger.info(f"[DEBUG_GEM] Initializing new chat session with model={model} using client: {client.id}")
-            session = client.start_chat(model=model)
+            session = client.start_chat(model=model, gem=native_gem_id)
             messages_to_send = _prepare_messages_for_model(
                 request.messages, request.tools, request.tool_choice, extra_instructions
             )
@@ -845,7 +848,7 @@ async def create_chat_completion(
                     f"Fallback attempt {i+1}/{max_retries}: Switching to client {retry_client.id}"
                 )
 
-                retry_session = retry_client.start_chat(model=model)
+                retry_session = retry_client.start_chat(model=model, gem=native_gem_id)
                 response = await _send_with_split(
                     retry_session, fallback_model_input, files=fallback_files
                 )
@@ -982,6 +985,8 @@ async def create_response(
     if gem:
         _apply_gem_overrides(request_data, gem)
 
+    native_gem_id = gem_id if (gem and gem.is_native) else None
+
     structured_requirement = _build_structured_requirement(request_data.response_format)
     if structured_requirement and request_data.stream:
         logger.debug(
@@ -1064,6 +1069,7 @@ async def create_response(
         model,
         messages,
         gem_id=gem_id,
+        native_gem_id=native_gem_id,
     )
 
     async def _build_payload(
@@ -1098,7 +1104,7 @@ async def _build_payload(
         try:
             client = await pool.acquire()
             logger.info(f"Initializing new session for responses with client: {client.id}")
-            session = client.start_chat(model=model)
+            session = client.start_chat(model=model, gem=native_gem_id)
             payload_messages = messages
             model_input, files = await _build_payload(payload_messages, _reuse_session=False)
         except ValueError as e:
@@ -1185,7 +1191,7 @@ async def _build_payload(
                     f"Fallback attempt {i+1}/{max_retries}: Switching to client {retry_client.id}"
                 )
 
-                retry_session = retry_client.start_chat(model=model)
+                retry_session = retry_client.start_chat(model=model, gem=native_gem_id)
                 model_output = await _send_with_split(
                     retry_session, fallback_model_input, files=fallback_files
                 )
@@ -1432,6 +1438,7 @@ async def _find_reusable_session(
     model: Model,
     messages: list[Message],
     gem_id: str | None = None,
+    native_gem_id: str | None = None,
 ) -> tuple[ChatSession | None, GeminiClientWrapper | None, list[Message]]:
     """Find an existing chat session that matches the *longest* prefix of
     ``messages`` **whose last element is an assistant/system reply**.
@@ -1463,7 +1470,7 @@ async def _find_reusable_session(
             try:
                 if conv := db.find(model.model_name, search_history, gem_id=gem_id):
                     client = await pool.acquire(conv.client_id)
-                    session = client.start_chat(metadata=conv.metadata, model=model)
+                    session = client.start_chat(metadata=conv.metadata, model=model, gem=native_gem_id)
                     remain = messages[search_end:]
                     return session, client, remain
             except Exception as e:

app/utils/config.py

@@ -55,6 +55,10 @@ class GemDefinition(BaseModel):
 
     id: str = Field(..., description="Unique identifier for the gem")
     model: str = Field(..., description="Model name to use when this gem is selected")
+    is_native: bool = Field(
+        default=False,
+        description="If True, this gem ID is passed upstream to Google's servers. If False, it is a local configuration preset.",
+    )
     system_prompt: Optional[str] = Field(
         default=None,
         description="Optional system prompt injected before user messages",