ai-decision-council

ai-decision-council

ai-decision-council is a plug-and-play Python SDK for multi-model council workflows:

1. Collect first-pass answers from multiple models.
2. Run peer ranking across those answers.
3. Synthesize one final decision-style response.

Default behavior uses a curated set of 5 models (configurable from 2 to 20).

Install

pip install ai-decision-council

5-minute startup

# Use your OpenRouter API key here.
export LLM_COUNCIL_API_KEY="..."
ai-decision-council doctor
ai-decision-council run --prompt "Explain CAP theorem in simple terms"

5-minute API integration

pip install "ai-decision-council[api]"
export LLM_COUNCIL_API_KEY="..."
export LLM_COUNCIL_REFERENCE_API_TOKEN="replace-with-long-random-token"
ai-decision-council api serve

Then call POST /v1/conversations on http://localhost:8001.

Python integration

from ai_decision_council import Council

council = Council.from_env()
result = council.run_sync("Summarize vector databases in 5 bullets")
print(result.final_response)

CLI

ai-decision-council doctor
ai-decision-council models --defaults
ai-decision-council models --defaults --count 7
ai-decision-council init
ai-decision-council init --api fastapi
ai-decision-council api bootstrap --path .
ai-decision-council api openapi --output ./openapi.json
ai-decision-council api sdk --output-dir ./sdk
ai-decision-council run --prompt "Your prompt" --json

Legacy alias remains supported:

llm-council ...

Core configuration

• LLM_COUNCIL_API_KEY (preferred; your OpenRouter API key)
• LLM_COUNCIL_MODELS (explicit model list)
• LLM_COUNCIL_MODEL_COUNT (default 5)
• LLM_COUNCIL_CHAIRMAN_MODEL
• LLM_COUNCIL_TITLE_MODEL
• LLM_COUNCIL_API_URL
• LLM_COUNCIL_REFERENCE_API_TOKEN (required for packaged API auth)

Current provider scope is OpenRouter-first. Fallback key support: OPENROUTER_API_KEY.

Documentation

• docs/quickstart.md
• docs/python-integration.md
• docs/cli-usage.md
• docs/framework-snippets.md
• docs/troubleshooting.md
• docs/migration.md
• docs/openapi.json