fix: non-streaming gemini tool calling (#20207)

Author: AstraBert

llama-index-integrations/llms/llama-index-llms-google-genai/llama_index/llms/google_genai/base.py

@@ -392,9 +392,14 @@ def gen() -> ChatResponseGen:
                         content += content_delta
                 llama_resp = chat_from_gemini_response(r)
                 if content:
-                    llama_resp.message.blocks.append(TextBlock(text=content))
+                    llama_resp.message.blocks = [TextBlock(text=content)]
                 if thoughts:
-                    llama_resp.message.blocks.append(ThinkingBlock(content=thoughts))
+                    if llama_resp.message.blocks:
+                        llama_resp.message.blocks.append(
+                            ThinkingBlock(content=thoughts)
+                        )
+                    else:
+                        llama_resp.message.blocks = [ThinkingBlock(content=thoughts)]
                     yield llama_resp
 
             if self.use_file_api:
@@ -445,15 +450,17 @@ async def gen() -> ChatResponseAsyncGen:
                                 else:
                                     content += content_delta
                             llama_resp = chat_from_gemini_response(r)
-                            llama_resp.delta = content_delta
                             if content:
-                                llama_resp.message.blocks.append(
-                                    TextBlock(text=content)
-                                )
+                                llama_resp.message.blocks = [TextBlock(text=content)]
                             if thoughts:
-                                llama_resp.message.blocks.append(
-                                    ThinkingBlock(content=thoughts)
-                                )
+                                if llama_resp.message.blocks:
+                                    llama_resp.message.blocks.append(
+                                        ThinkingBlock(content=thoughts)
+                                    )
+                                else:
+                                    llama_resp.message.blocks = [
+                                        ThinkingBlock(content=thoughts)
+                                    ]
                             yield llama_resp
 
             if self.use_file_api:

llama-index-integrations/llms/llama-index-llms-google-genai/llama_index/llms/google_genai/utils.py

@@ -191,7 +191,7 @@ def chat_from_gemini_response(
                     )
                 content_blocks.append(
                     ToolCallBlock(
-                        tool_call_id=part.function_call.id or "",
+                        tool_call_id=part.function_call.name or "",
                         tool_name=part.function_call.name or "",
                         tool_kwargs=part.function_call.args or {},
                     )
@@ -200,7 +200,6 @@ def chat_from_gemini_response(
                 # follow the same pattern as for transforming a chatmessage into a gemini message: if it's a function response, package it alone and return it
                 additional_kwargs["tool_call_id"] = part.function_response.id
                 role = ROLES_FROM_GEMINI[top_candidate.content.role]
-                print("RESPONSE", json.dumps(part.function_response.response))
                 return ChatResponse(
                     message=ChatMessage(
                         role=role, content=json.dumps(part.function_response.response)

llama-index-integrations/llms/llama-index-llms-google-genai/pyproject.toml

@@ -27,7 +27,7 @@ dev = [
 
 [project]
 name = "llama-index-llms-google-genai"
-version = "0.7.0"
+version = "0.7.1"
 description = "llama-index llms google genai integration"
 authors = [{name = "Your Name", email = "[email protected]"}]
 requires-python = ">=3.9,<4.0"