Skip to content

No support for OpenAI-like API type #53

Open
@Andreea1998

Description

@Andreea1998

Describe the bug
OpenAI-like API type is not supported and error is given

To Reproduce
Steps to reproduce the behavior:

  1. I followed your setup steps from https://github.com/test-zeus-ai/testzeus-hercules
  • agents_llm_config.json configuration:
{
        "mistral": {
                "planner_agent": {
                        "model_name": "mistralai/Mistral-Small-3.1-24B-Instruct-2503",
                        "model_api_key": "XXX",
                        "model_base_url": "<local_vllm>",
            "model_api_type": "openai",
            "llm_config_params": {
                "cache_seed": null,
                "temperature": 0.0,
                "seed": 12345
            }
                },
                "nav_agent": {
                        "model_name": "mistralai/Mistral-Small-3.1-24B-Instruct-2503",
                        "model_api_key": "XXX",
                        "model_base_url": "<local_vllm>",
            "model_api_type": "openai",
            "llm_config_params": {
                "cache_seed": null,
                "temperature": 0.0,
                "seed": 12345
            }
}
}
}
  • .env coniguration:
AGENTS_LLM_CONFIG_FILE=agents_llm_config.json
AGENTS_LLM_CONFIG_FILE_REF_KEY=mistral

TOKENIZERS_PARALLELISM=false
MODE=prod
HEADLESS=false
RECORD_VIDEO=false
TAKE_SCREENSHOTS=true
BROWSER_TYPE=chromium
CAPTURE_NETWORK=true
HF_HOME=./.cache
  1. Used unchanged example from opt/input/test.feature.
  2. I used these commands

A.

make run

This error is given:

openai.BadRequestError: Error code: 400 - {'detail': '400: Open WebUI: Server Connection Error'}

Full log: make_run.log

or
B.

testzeus-hercules --input-file opt/input/test.feature --output-path opt/output --test-data-path opt/test_data  --agents-llm-config-file ./agents_llm_config.json --agents-llm-config-file-ref-key mistral

This error is given:

pydantic_core._pydantic_core.ValidationError: 1 validation error for _LLMConfig
config_list.0.openai.model
  Field required [type=missing, input_value={'api_type': 'openai'}, input_type=dict]

Full log: testzeus.log

Expected behavior
I'm using Mistral model served using vllm https://github.com/vllm-project/vllm which is OpenAI-like
Using "openai" as "model_api_type" doesn't seem to work, but I expected to.
Maybe a different api_type should be used for OpenAI-like models

Desktop

  • OS: MacOS
  • Version Sonoma 14.6

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't workinghelp wantedExtra attention is needed

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions