multi_agent_tool

Environment Variables (Local LLM)

This project loads .env automatically via python-dotenv.

Set the local LLM variables below to run against a local OpenAI-compatible server:

Variable	Required	Default	Description
LLAMA_SERVER_BASE_URL	Yes (for local mode)	none	Base URL of your local server, e.g. http://127.0.0.1:8080/v1.
LLAMA_SERVER_MODEL_ID	Yes (for local mode)	none	Model name/id sent as model in chat completion requests.
LLAMA_SERVER_API_KEY	No	empty	Optional bearer token for local server auth.
LLAMA_SERVER_TIMEOUT	No	120	Request timeout in seconds. Must be > 0.
LLAMA_SERVER_FOLLOWUP_PROMPT	No	Please provide your response.	Prompt appended when the previous message role is assistant or tool.
LLAMA_SERVER_CONTINUE_PROMPT	No	Continue.	Prompt used when the model stops with finish_reason=length.
LLAMA_SERVER_MAX_CONTINUATIONS	No	0	Number of continuation retries when response is truncated. Must be >= 0.

Optional App Variables

Variable	Required	Default	Description
MAF_WORKFLOW_MODE	No	sequential	One of sequential, group_chat, concurrent.
MAF_GROUPCHAT_MAX_ROUNDS	No	number of agents	Max rounds for group_chat mode. Must be integer > 0.
AGENTS_TABLE_PATH	No	agents.csv	Path to agent configuration CSV.

Quick Start

1. Copy .env.example to .env.
2. Fill in local server values.
3. Run:

python run_test_agent.py