multi-model

Multi-model orchestration coordinates cross-model AI reviews by dispatching a fixed 4-question scaffold to multiple CLIs in parallel, gathering diverse analyses, and synthesizing the results with model-attribution to reduce reliance on a single viewpoint.

Runtime CLI discovery detects installed models including Codex, Copilot, Gemini, Opencode, Pi, Aider, llm, aichat, and Goose, running a quorum check before dispatching the scaffold to the available CLIs for parallel analysis.

To run a cross-model council, supply a scaffold and let the available CLIs respond in parallel. The system dispatches the fixed 4-question scaffold, verifies a quorum, and synthesizes the parallel responses into a single attributed output.

Multi-model collaboration persists context by storing transcripts and decisions in memory-backed storage using wicked-brain:memory and jam, enabling full auditability and iterative refinement of technical decisions over time.

The fixed 4-question scaffold limits the scope of cross-model AI reviews to predefined inquiries, requiring a quorum of detected CLIs to proceed, which restricts ad-hoc questioning and depends on multiple models being installed at runtime.

Use a multi-model council when technical decisions require diverse perspectives to avoid single-viewpoint bias, applying parallel dispatch across multiple CLIs to surface contrasting analyses and synthesize a comprehensive, model-attributed conclusion.