Error when using Hugginface Inference API with Phidata tools

[nagendragillalu]

I am trying to execute a demo example using Hugging face Inference API using phidata tools
though I change to any model I get following error

[/usr/local/lib/python3.10/dist-packages/phi/agent/agent.py](https://localhost:8080/#) in print_response(self, message, messages, stream, markdown, show_message, show_reasoning, show_full_reasoning, console, **kwargs)
   2810                     live_log.update(Group(*panels))
   2811 
-> 2812                 for resp in self.run(message=message, messages=messages, stream=True, **kwargs):
   2813                     if isinstance(resp, RunResponse) and isinstance(resp.content, str):
   2814                         if resp.event == RunEvent.run_response:

[/usr/local/lib/python3.10/dist-packages/phi/agent/agent.py](https://localhost:8080/#) in _run(self, message, stream, audio, images, videos, messages, stream_intermediate_steps, **kwargs)
   1804         if self.stream:
   1805             model_response = ModelResponse(content="")
-> 1806             for model_response_chunk in self.model.response_stream(messages=messages_for_model):
   1807                 if model_response_chunk.event == ModelResponseEvent.assistant_response.value:
   1808                     if model_response_chunk.content is not None and model_response.content is not None:

[/usr/local/lib/python3.10/dist-packages/phi/model/huggingface/hf.py](https://localhost:8080/#) in response_stream(self, messages)
    734 
    735         if stream_data.response_tool_calls is not None:
--> 736             _tool_calls = self._build_tool_calls(stream_data.response_tool_calls)
    737             if len(_tool_calls) > 0:
    738                 assistant_message.tool_calls = _tool_calls

[/usr/local/lib/python3.10/dist-packages/phi/model/huggingface/hf.py](https://localhost:8080/#) in _build_tool_calls(self, tool_calls_data)
    824         for _tool_call in tool_calls_data:
    825             _index = _tool_call.index
--> 826             _tool_call_id = _tool_call.id
    827             _tool_call_type = _tool_call.type
    828             _function_name = _tool_call.function.name if _tool_call.function else None

AttributeError: 'str' object has no attribute 'id'

This is my sample agent

from phi.agent import Agent
from phi.model.huggingface import HuggingFaceChat
from phi.tools.duckduckgo import DuckDuckGo
import os 

web_agent = Agent(
    name="Web Agent",
    model=HuggingFaceChat(
        id="meta-llama/Meta-Llama-3-8B-Instruct",
        max_tokens=4096,
    ),
    tools=[DuckDuckGo()],
    instructions=["Always include sources"],
    show_tool_calls=True,
    markdown=True,
)
web_agent.print_response("Whats happening in France?", stream=True)

[ysolanky]

Hello @nagendragillalu I was able to replicate this issue with HuggingFaceChat. Our HuggingFaceChat model class requires an update and I am on it. Sharing updates soon!

[omkarjoshi]

@ysolanky : any timeline when this will be fixed ?

[ysolanky]

@omkarjoshi I will have a PR out by the end of this week. Apologies for the delay

[Abhishek-kumar-samsung]

Hii @ysolanky when i am running llm locally using TGI wrapper and passing it as argument in OpenAIChat like this
model=OpenAIChat(
base_url="llm url",
api_key="-"
)

i am getting similar error.

_tool_call_id = _tool_call.id
^^^^^^^^^^^^^
AttributeError: ‘str’ object has no attribute ‘id’

[dirkbrnd]

@Abhishek-kumar-samsung we haven't forgotten about this. We have a WIP fix that we plan to get out soon.