wire-ai-component

name	wire-ai-component
description	Step-by-step workflow for building and wiring a new AI component (classifier, reranker, etc.) into the ORBIS pipeline

/wire-ai-component

When to use

Use this when introducing a new AI component (e.g. intent classifier, reranker, retrieval module) that needs to be integrated into the existing ORBIS request pipeline and metrics system.

Steps

1. Create the component module — Write the new component in its own file under the relevant package (e.g. agent/intent.py, memory/reranker.py). Define a clean class or function interface.
2. Hook into the nearest data layer — Integrate the component into the lowest-level module that owns the relevant data (e.g. hook reranker into memory/facts.py search()).
3. Wire into the app entrypoint — Import the component in app.py, instantiate it, and insert it into the pre-LLM processing path at the correct point.
4. Add metrics — Find the _METRICS dict in app.py and register any new counters, histograms, or gauges for the component.

Examples

• Intent classifier: Create agent/intent.py → hook into app.py pre-LLM path → add intent_classified_total counter to _METRICS.
• BM25 + cross-encoder reranker: Create memory/reranker.py → hook into memory/facts.py search() → wire result into app.py → add rerank_latency_seconds histogram to _METRICS.