How focuslead/ai-council-framework works
The repository presents a research-backed methodology for multi-AI deliberation, not an implemented runtime: the tree contains documentation and examples but no source code. The graph follows the README’s documented user-to-answer protocol and roles; implementation details beyond that documentation are not inferred.
Main entry points: README.md Synthesize consensus
Components 14
Council orchestration
- Project manager
README.md - Depth mode
README.md - Council members
README.md
Deliberation protocol
- Distribute prompt
README.md - Collect responses
README.md - Synthesize consensus
README.md - Structured debate
README.md - Consensus scoring
README.md
Validation
- Fresh Eyes validator
README.md - Web search
- Verification
README.md
Decision output
- Final recommendation
README.md - Human decision
README.md
Other components
- Human user
Connections 18
- Human user → Project manager: submits question
- Human user → Depth mode: selects depth
- Depth mode → Project manager: guides rigor
- Project manager → Distribute prompt: orchestrates
- Distribute prompt → Council members: sends prompt
- Council members → Collect responses: provide responses
- Collect responses → Project manager: returns responses
- Project manager → Synthesize consensus: synthesizes
- Synthesize consensus → Consensus scoring: scores claims
- Synthesize consensus → Structured debate: surfaces disagreement
- Structured debate → Council members: requests evidence
- Structured debate → Synthesize consensus: informs revision
- Synthesize consensus → Fresh Eyes validator: sends answer
- Fresh Eyes validator → Verification: validates constructively
- Verification → Web search: checks claims
- Verification → Final recommendation: informs recommendation
- Consensus scoring → Final recommendation: provides scores
- Final recommendation → Human decision: presents split decision