pinky-ai

Update date : 2026-06-01 08:26

Cortex AI wrapper for the pinky suite — semantic cache, model routing, token compression, and cost intelligence.

---

• explanation/design.md — architecture, backend abstraction, Layer 4 cost model. Not for: feature lists (→ features.md), decisions (→ adr/)
• explanation/features.md — feature list by status (stable / planned). Not for: design rationale (→ design.md)
• reference/generated/api/ai.md — PinkyAI class + get_ai factory. Not for: hand-written content — edit docstrings in src/
• reference/generated/api/cost.md — CostIntelligence class. Not for: hand-written content — edit docstrings in src/
• adr/0001-semantic-cache.md — semantic cache (Cortex has no native equivalent)
• adr/0002-backend-abstraction.md — single PinkyAI interface across cortex/ollama/groq/task
• adr/0003-model-sovereignty.md — Mistral EU-first; Meta/Llama excluded for minor data
• adr/0004-cortex-vs-eai.md — Cortex (0 WH) for sessions; EAI for serverless batch