Prompt template issue in Gemini 2.0 and 1.5 (Its an issue with Gemini API not fast agent)

[usamaJ17]

I am facing an error on gemini-2.0-flash , gemini-2.0-flash-lite and gemini-1.5-flash , basically all models I tried below flash 2.5.
this is the error :
{'error': {'code': 400, 'message': '* GenerateContentRequest.contents[23].parts: contents.parts must not be empty.\n', 'status': 'INVALID_ARGUMENT'}}
I. think its due to some prompt template, as in langchain this issue is because of a prompt template.

[evalstate]

Which provider is this with?

[usamaJ17]

hi @evalstate
this is with google LLM apis.. If I use. flash2.5 or pro 2.5 , it works, but If i try. to use a lower model, like flash2.0, then suddenly durring working, it throw this error..

[usamaJ17]

And after I updated fast agent version, flash 2.5 pro is not. working as well

[usamaJ17]

This is. the. error trace:

Traceback (most recent call last):
File "/Users/my_user/fast-agent/main.py", line 30, in
asyncio.run(main())
~~~~~~~~~~~^^^^^^^^
File "/usr/local/Cellar/python@3.13/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/asyncio/runners.py", line 195, in run
return runner.run(main)
~~~~~~~~~~^^^^^^
File "/usr/local/Cellar/python@3.13/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/asyncio/runners.py", line 118, in run
return self._loop.run_until_complete(task)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^
File "/usr/local/Cellar/python@3.13/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/asyncio/base_events.py", line 725, in run_until_complete
return future.result()
~~~~~~~~~~~~~^^
File "/Users/my_user/fast-agent/main.py", line 26, in main
await agent.interactive()
File "/Users/my_user/fast-agent/.venv/lib/python3.13/site-packages/mcp_agent/core/agent_app.py", line 267, in interactive
return await prompt.prompt_loop(
^^^^^^^^^^^^^^^^^^^^^^^^^
...<5 lines>...
)
^
File "/Users/my_user/fast-agent/.venv/lib/python3.13/site-packages/mcp_agent/core/interactive_prompt.py", line 182, in prompt_loop
result = await send_func(user_input, agent)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/my_user/fast-agent/.venv/lib/python3.13/site-packages/mcp_agent/core/agent_app.py", line 264, in send_wrapper
return await self.send(message, agent_name)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/my_user/fast-agent/.venv/lib/python3.13/site-packages/mcp_agent/core/agent_app.py", line 92, in send
return await self._agent(agent_name).send(message)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/my_user/fast-agent/.venv/lib/python3.13/site-packages/mcp_agent/agents/base_agent.py", line 225, in send
response = await self.generate([prompt], None)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/my_user/fast-agent/.venv/lib/python3.13/site-packages/mcp_agent/agents/base_agent.py", line 596, in generate
return await self._llm.generate(multipart_messages, request_params)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/my_user/fast-agent/.venv/lib/python3.13/site-packages/mcp_agent/llm/augmented_llm.py", line 212, in generate
assistant_response: PromptMessageMultipart = await self._apply_prompt_provider_specific(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
multipart_messages, request_params
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/Users/my_user/fast-agent/.venv/lib/python3.13/site-packages/mcp_agent/llm/providers/augmented_llm_google_native.py", line 419, in _apply_prompt_provider_specific
responses = await self._google_completion(request_params=request_params)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/my_user/fast-agent/.venv/lib/python3.13/site-packages/mcp_agent/llm/providers/augmented_llm_google_native.py", line 389, in _google_completion
new_multipart_messages = self._converter.convert_from_google_content_list(
new_google_messages
)
File "/Users/my_user/fast-agent/.venv/lib/python3.13/site-packages/mcp_agent/llm/providers/google_converter.py", line 334, in convert_from_google_content_list
return [self._convert_from_google_content(content) for content in contents]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^
File "/Users/my_user/fast-agent/.venv/lib/python3.13/site-packages/mcp_agent/llm/providers/google_converter.py", line 346, in _convert_from_google_content
for part in content.parts:
^^^^^^^^^^^^^
TypeError: 'NoneType' object is not iterable

[evalstate]

If you are able, can you try merging this PR: #215
The other option is to switch to the OpenAI compatible endpoint using googleoai as the provider (and update the API key/config file to be GOOGLEOAI_API_KEY).

[usamaJ17]

hi @evalstate
I cant merge this PR, as I don’t have the required permission. I tried again with gemini flash 2.0 and again got the same error.. even though I am now on updated version.
still this error

File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/llm/providers/google_converter.py", line 346, in _convert_from_google_content
for part in content.parts:
TypeError: 'NoneType' object is not iterable

i would be happy to merge this PR and try again.. but I don’t have required permissions
Edit : even with this PR (I updated that PR code to my local) getting same error

[usamaJ17]

This is the complete error log: (got that again. sometimes it works sometime not)

Traceback (most recent call last):
File "/Users/usama.jalal/bucket-mcp/fast-agent/main.py", line 30, in
asyncio.run(main())
File "/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/asyncio/runners.py", line 190, in run
return runner.run(main)
^^^^^^^^^^^^^^^^
File "/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/asyncio/runners.py", line 118, in run
return self._loop.run_until_complete(task)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/asyncio/base_events.py", line 654, in run_until_complete
return future.result()
^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/main.py", line 26, in main
await agent.interactive()
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/core/agent_app.py", line 285, in interactive
return await prompt.prompt_loop(
^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/core/interactive_prompt.py", line 209, in prompt_loop
result = await send_func(user_input, agent)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/core/agent_app.py", line 277, in send_wrapper
result = await self.send(message, agent_name)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/core/agent_app.py", line 94, in send
return await self._agent(agent_name).send(message)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/agents/base_agent.py", line 226, in send
response = await self.generate([prompt], None)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/agents/base_agent.py", line 597, in generate
return await self._llm.generate(multipart_messages, request_params)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/llm/augmented_llm.py", line 223, in generate
assistant_response: PromptMessageMultipart = await self._apply_prompt_provider_specific(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/llm/providers/augmented_llm_google_native.py", line 453, in _apply_prompt_provider_specific
responses = await self._google_completion(request_params=request_params)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/llm/providers/augmented_llm_google_native.py", line 423, in _google_completion
new_multipart_messages = self._converter.convert_from_google_content_list(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/llm/providers/google_converter.py", line 334, in convert_from_google_content_list
return [self._convert_from_google_content(content) for content in contents]
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/llm/providers/google_converter.py", line 334, in
return [self._convert_from_google_content(content) for content in contents]
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/llm/providers/google_converter.py", line 340, in _convert_from_google_content
if content.role == "model" and any(part.function_call for part in content.parts):
^^^^^^^^^^^^
AttributeError: 'NoneType' object has no attribute 'role'

[usamaJ17]

@evalstate
any update in this?? this error is very teasing.. and I am getting this multiple times..

[usamaJ17]

@monotykamary @evalstate
even after merging PR, still face error.. I have observed that error is from this line
// google_converter.py (around line 474)
if content.role == "model" and any(part.function_call for part in content.parts):

if there is a tool call.. and tool returns some data.. content is None.. and in that case using content.role results in :
AttributeError: 'NoneType' object has no attribute 'role'

[evalstate]

Thanks @usamaJ17 I've been wanting a reproducible test case before changing (need to keep track of all the provider specific tweaks!) and that info helps.

Does this happen with both the new native and legacy oai providers?

[usamaJ17]

hi @evalstate , what I have observed, I am getting this on gemini API. I tried 2.5 flash lite and 2.5 flash as well.. getting that error on both.
also what I observe, I face this error while using filesystem tools, I am using This Filesysem MCP
i tried to go down into code to fix these.. but I ddnt get a very clear idea of what can be the issue,, I tried to filter None from content array, that doesnot throw error, but it also doesnot gave me proper message on console.,. (and I was shown input in console, but without any message) .
that main issue is, when i checked logs MALFORMED_FUNCTION_CALL is the cause of final API failure, and in contents array in gemini API response, last index was None. it was supposed to be an object, but in response it was None,. and that why we got the error when we tried to get attribute of none, removing None from array also ddnt helped,. it ddnt throw error, but just stoped the execution..

[usamaJ17]

@evalstate , also I get this as well

Traceback (most recent call last):
File "/Users/usama.jalal/bucket-mcp/fast-agent/main.py", line 30, in
asyncio.run(main())
File "/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/asyncio/runners.py", line 190, in run
return runner.run(main)
^^^^^^^^^^^^^^^^
File "/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/asyncio/runners.py", line 118, in run
return self._loop.run_until_complete(task)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/asyncio/base_events.py", line 654, in run_until_complete
return future.result()
^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/main.py", line 26, in main
await agent.interactive()
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/core/agent_app.py", line 285, in interactive
return await prompt.prompt_loop(
^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/core/interactive_prompt.py", line 209, in prompt_loop
result = await send_func(user_input, agent)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/core/agent_app.py", line 277, in send_wrapper
result = await self.send(message, agent_name)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/core/agent_app.py", line 94, in send
return await self._agent(agent_name).send(message)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/agents/base_agent.py", line 226, in send
response = await self.generate([prompt], None)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/agents/base_agent.py", line 597, in generate
return await self._llm.generate(multipart_messages, request_params)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/llm/augmented_llm.py", line 223, in generate
assistant_response: PromptMessageMultipart = await self._apply_prompt_provider_specific(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/llm/providers/augmented_llm_google_native.py", line 430, in _apply_prompt_provider_specific
responses = await self._google_completion(request_params=request_params)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/llm/providers/augmented_llm_google_native.py", line 400, in _google_completion
new_multipart_messages = self._converter.convert_from_google_content_list(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/llm/providers/google_converter.py", line 465, in convert_from_google_content_list
return [self._convert_from_google_content(content) for content in contents]
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/llm/providers/google_converter.py", line 465, in
return [self._convert_from_google_content(content) for content in contents]
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/usama.jalal/bucket-mcp/fast-agent/.venv/lib/python3.11/site-packages/mcp_agent/llm/providers/google_converter.py", line 474, in _convert_from_google_content
if content.role == "model" and any(part.function_call for part in content.parts):
^^^^^^^^^^^^
AttributeError: 'NoneType' object has no attribute 'role'

[usamaJ17]

@evalstate , I was trying to fix that, but whats seems like, its not in our framework, its in google genai .
you can review this issue : googleapis/python-aiplatform#4472 , as its mention in this as well, that they have migrated to google Gen AI SDK, instead of Vertex Ai SDK, BUT, this framework is using google GEN AI SDK , so the actual issue is I think This : googleapis/python-genai#430 , because what I observered, it almost always occurs while writing final output to a file, and that output is a bit long.. not too much, but like around 150 lines of text.