Forge Intelligence

Handbook

forge-intelligence

Private Forge product for problem-class intelligence: general + specific packs that reduce token footprints by routing through deterministic ladders before bounded LLM acceleration.

Updated

What it is

  • General intelligence — catalog + reasoning ladder (deterministic → cheap model → break-in → escalate).
  • Specific intelligence — per-domain packs (packs/<domain>/) with policies, lmeta flows, and fixtures.
  • Session dual-wiki — machine JSON (M) + generated human report (H) + freeze gate per intelligence session.

What it is not

  • Not a second LCDL runtime (consumes forge-lcdl tasks and FlowExecutor).
  • Not the ForgeRun system of record (Lenses).
  • Not the Platform doc-hydration "intelligence layer".

Handbook

Markdown source lives in docs/. The private handbook deploys from forge-intelligence-website (abbr fiw).

Quick start

pip install -e ".[dev]"
pytest

Session dual-wiki

PYTHONPATH=src python3 -m forge_intelligence.wiki.freeze_gate sessions/<session_id>

See docs/DUAL-WIKI.md.