fix(provider): normalize 1-based streaming tool_call indexes and fallback missing ids (#9590)

astrbot/core/provider/entities.py

@@ -427,17 +427,30 @@ def to_openai_tool_calls_model(self) -> list[ToolCall]:
         """The same as to_openai_tool_calls but return pydantic model."""
         ret = []
         for idx, tool_call_arg in enumerate(self.tools_call_args):
+            # Some OpenAI-compatible gateways return tool_calls without a valid id
+            # (e.g. streaming snapshots misaligned by non-zero-based indexes).
+            # ToolCall.id is declared as str, so passing None would raise a pydantic
+            # ValidationError and fail the whole tool-calling turn.
+            raw_id = (
+                self.tools_call_ids[idx] if idx < len(self.tools_call_ids) else None
+            )
+            func_name = (
+                self.tools_call_name[idx] if idx < len(self.tools_call_name) else None
+            )
+            call_id = raw_id
+            if not isinstance(call_id, str) or not call_id:
+                call_id = f"call_{idx}"
             ret.append(
                 ToolCall(
-                    id=self.tools_call_ids[idx],
+                    id=call_id,
                     function=ToolCall.FunctionBody(
-                        name=self.tools_call_name[idx],
+                        name=func_name or "__malformed_tool_name__",
                         arguments=json.dumps(tool_call_arg),
                     ),
                     # the extra_content will not serialize if it's None when calling ToolCall.model_dump()
-                    extra_content=self.tools_call_extra_content.get(
-                        self.tools_call_ids[idx]
-                    ),
+                    # Note: look up extra_content by the raw upstream id; the fallback id
+                    # (call_{idx}) never exists in tools_call_extra_content.
+                    extra_content=self.tools_call_extra_content.get(raw_id),
                 ),
             )
         return ret

astrbot/core/provider/sources/openai_source.py

@@ -638,6 +638,11 @@ async def _query_stream(
 
         state = ChatCompletionStreamState()
 
+        # 上游 tool_call.index -> 规范化后的 0 起始下标。
+        # 必须是每次请求独立的局部状态，不能挂在 self 上，
+        # 否则并发请求会互相污染下标映射。
+        tool_call_index_map: dict[int, int] = {}
+
         async for chunk in stream:
             choice = chunk.choices[0] if chunk.choices else None
             delta = choice.delta if choice else None
@@ -651,6 +656,21 @@ async def _query_stream(
                     # Gemini and some OpenAI-compatible proxies omit this field
                     if not hasattr(tc, "index") or tc.index is None:
                         tc.index = idx
+                    else:
+                        # Some OpenAI-compatible gateways start tool_call.index at 1
+                        # (the OpenAI spec requires 0-based indexing). The SDK streaming
+                        # snapshot accumulator uses index directly as a list subscript,
+                        # so index=1 on an empty list raises IndexError and the tool_call
+                        # name/arguments get split across misaligned slots, eventually
+                        # producing a malformed tool_call with id=None/name="" that fails
+                        # downstream ToolCall pydantic validation.
+                        # Remap upstream indexes to a contiguous 0-based sequence.
+                        remapped = tool_call_index_map.setdefault(
+                            tc.index,
+                            len(tool_call_index_map),
+                        )
+                        if remapped != tc.index:
+                            tc.index = remapped
             # 跳过 delta=None 的 chunk，避免 SDK 内部 _convert_initial_chunk_into_snapshot
             # 第 747 行 choice.delta.to_dict() 抛出 NoneType 错误。
             # refs: AstrBot#6689 / openai-python#5069 / #5047