LLM Wiki — Project Knowledge Base
This wiki is the primary knowledge source for AI tools (GitHub Copilot, future LLMs) working in this project. Read this first — before scanning docs/, specs/, or activity reports.
It is published at docs.invest.justingoufo.com/llm-wiki/ but intentionally absent from the public navigation sidebar. It is an AI-operational artifact, not public content.
When to Use This Wiki
| You need to… | Go to |
|---|---|
| Understand why a technical decision was made | 01_architecture/key-decisions.md |
| Understand data ownership and persistence contracts | 02_data-models/ |
| Follow end-to-end operational workflows | 03_workflows/ |
| Execute deployment and maintenance runbooks | 04_operations/ |
| Find a code pattern (retry, error handling, security, testing) | 05_common-patterns/ |
| Learn from incidents and performance gaps | 06_lessons-learned/ |
| Find how to perform a specific developer task | index-by-task.md |
| Understand the overall architecture | 01_architecture/ |
Section Index
| Section | Status | Purpose |
|---|---|---|
| 01_architecture/ | ✅ V1 — Full content | Architecture decisions, key choices, incidents log |
| 02_data-models/ | ✅ V2 — Core content | Entity schemas, data contracts, persistence layer |
| 03_workflows/ | ✅ V2 — Core content | Import pipeline, alert system, GitHub issue bridge |
| 04_operations/ | ✅ V2 — Core content | Deployment, environment setup, log management |
| 05_common-patterns/ | ✅ V1 — Full content | Retry strategies, error handling, security, testing |
| 06_lessons-learned/ | ✅ V2 — Core content | Incidents, optimizations, anti-patterns |
Context
The project previously had no structured knowledge base. AI sessions spent tokens scanning 70+ bilingual activity reports and specs, wasting ~40% of context budget on redundant content. This wiki provides curated, AI-readable context in 2–3 reads instead of 20+.
Details
Current scope: all 6 sections are now documented with operational content. V1 sections (01_architecture/, 05_common-patterns/) remain the historical foundation; V2 sections (02_data-models/, 03_workflows/, 04_operations/, 06_lessons-learned/) now cover the previously missing workflows and runbooks.
Latest update (September 2026): The AI workflow now includes a manual /speckit.session-review closing review. It uses Chronicle aggregates first, compares at most five relevant activity reports, and records process and token-saving recommendations without permanent session instrumentation.
Maintenance rule: Wiki entries are updated in the same commit/PR as the code change they document. Entries are archived (moved to _archive/) rather than deleted. Quarterly review cycle.
Language: English only. This wiki is an AI-operational artifact — the "draft in French first" documentation workflow does not apply here (constitution exception approved 2026-05-05).
Links
- index-by-task.md — task-oriented cross-reference
- 01_architecture/key-decisions.md — full decision log
- 05_common-patterns/ — code patterns with pseudocode
Examples
AI session start: The AI reads docs/llm-wiki/README.md first, then opens the relevant section (01_architecture/ or 05_common-patterns/) based on the question type. This replaces scanning all of docs/ or specs/.
Maintenance Procedure
Archiving outdated entries:
1. Move the outdated .md file to the section's _archive/ subdirectory.
2. Update the entry's status to archived before moving.
3. Add a note in the section index.md that the entry was archived and why.
4. Do not delete — archived entries preserve audit history.
Quarterly review checklist:
- [ ] Compare each current entry against live src/ code — confirm no contradictions
- [ ] Check last_updated dates — flag entries older than 6 months for review
- [ ] Verify all internal links resolve
- [ ] Update SC-010 tracking: confirm zero inconsistencies after review
Adding a new entry: Follow the standard template (title, status, last_updated frontmatter; ## Context / ## Details / ## Links body sections). Add a row to index-by-task.md if the entry maps to a developer task.