Documentation
Install
The Python package (distribution fastbi-io) is not yet published on PyPI; private install via INSTALL-PRIVATE.md. One package, all tiers, imported as import fastbi as fbi. Then run fastbi login --provider anthropic (or openai) to authenticate; it saves your model provider's API key.
Deterministic engine (all tiers)
Profiling, measures, evidence (facts, z-scores, aggregates) run locally in every tier, including Free.
Measures manifest
Every dashboard carries a measures manifest: the derived measures — Revenue, Return Rate, AOV, Units per Order — each with its pandas expression, format, and confidence.
AI insights (Pro)
Pro unlocks AI insights served from our backend over locally-computed facts.
Local-LLM mode (Max)
Max swaps the endpoint: set FASTBI_BASE_URL to any Anthropic-Messages-API-compatible server (Ollama ≥ 0.14, LM Studio). Minimum context window 32K; a mid-size instruct model is recommended. Your data stays on your machine for narration.