Docs - How It Works
How AGI-HIVE Works
End-to-end flow: ingest code, scan, generate tasks, route fixes, verify, and learn from every action.
Pipeline
1) Ingest: Forge pulls a repo (GitHub URL) or accepts files. It builds a memory map of critical files.
2) Scan: Sentinel scans code and dependencies; Hunger detects gaps and proposes tasks.
3) Map: Worldview + Forge memory map highlight critical paths, dependencies, circular links, and issue hotspots.
4) Plan: Hunger ranks quick wins and high-impact work; Sentinel issues sit side-by-side for action.
5) Route: P-Axis classifies the task and picks the cheapest capable model. You see the routing reasoning.
6) Apply: Patch preview is generated; you verify, run risk checks, and apply. Evidence logs automatically.
7) Learn: Cortex ingests patterns; Narrator explains evidence chains; dashboards surface learnings.
Why It Helps
- Faster triage: Sentinel findings + Hunger tasks in one place.
- Safer execution: P-Axis routing, risk checks, and evidence logging are built in.
- Continuous learning: Cortex patterns and Narrator explanations improve every run.
- Human control: You approve patches, pick tasks, and can audit every decision trail.