kb/methodology/INDEX.md

Methodology — Pointers to all design docs

All design / decision docs in one navigable place.

Foundation docs (in 70-docs/)

Doc	What it covers
01-requirements.md	Product vision, 3 user personas, buyer journey, 10 success criteria, 11 non-goals, constraints
02-architecture.md	Stack picks, system diagram, schema groupings, repo layout, c-readiness commitments
03-eval-plan.md	3-pipeline gold Q&A construction, grader, metrics, run cadence
04-failure-modes.md	16 named failure modes (F-01..F-16), detection + mitigation per mode
05-needs-analysis-flow.md	Fact-find question graph, bilingual prompts, termination logic
decisions.md	Append-only log of 17+ technical decisions with alternatives + reasoning
tech-stack-rationale.md	22-row stack pick table + selection rubric + cost envelope
scorecard-methodology.md	62-field schema → 24 scored fields → 6 sub-scores → A-F grade
ROADMAP.md	v1 vertical slice → v2 platform plan
information_source_map.md	Corpus catalog (auto-generated by rag/source_map.py)

KB sub-indexes (each regenerable via python -m rag.build_kb)

Path	Content
kb/INDEX.md	Master index
kb/AUDIT_TRAIL.md	10-stage data lineage + decision-to-artifact map
kb/policies/	One sheet per extracted policy
kb/research/	Corpus acquisition, URL verification, insurer universe
kb/calculations/	Scorecard results, eval results, extraction audit
kb/reviews/	Per-insurer reputation sheets + leaderboard
kb/premiums/	Illustrative pricing samples + scaling factors
kb/security/	Upload gates + hallucination defense
kb/eval/	Gold Q&A composition + run history

How to navigate

• New visitor → start at kb/INDEX.md
• Auditor → start at kb/AUDIT_TRAIL.md
• Engineer onboarding → start at 70-docs/02-architecture.md
• BFSI compliance review → kb/security/ + 70-docs/04-failure-modes.md + logs/hallucinations.jsonl
• Buyer-side curiosity → any kb/policies/<id>.md ends with a "What the bot will and won't say" section.