shabs-ch/automotive-reskill
Add directional disclaimer to Screen 1
Week 7: architecture and README finalised
Week 7: future.md updated
Week 7: architecture updated - skill extraction method clarified as generative IE
Week 7: add corpus health monitoring and drift monitoring sections to architecture
Week 7: retrieval method decision made explicit in architecture and decisions
Week 7: architecture doc, UI improvements, learnings and future updated
Week 7: README links to architecture doc
Week 7: architecture document added
Week 7: README complete - full project documentation
Day 30: error handling - user-friendly messages for API errors, timing fix
Day 30: Screen 4 UI redesign, gap analyzer deduplication fix, Screen 5 reverted
Week 6: context handoff updated - full project status
Day 25: add YAML knowledge files to repo for HF deployment
Day 25: add missing dependencies to requirements.txt
Day 25: ChromaDB Cloud migration - role matcher uses cloud in production
Day 24: learnings, decisions, future backlog, feedback round 1 updated
Day 24: learnings, decisions, future backlog updated post integration testing
Day 24: role matcher - always show min 2 families when second scores above 0.45
Day 24: course quality weighting - tier1 providers, BG eligibility balanced, month duration cap
Day 24: UI improvements - Screen 2 compact layout, Screen 6 tabs, course links
Day 24: integration fixes - source field, BG dedup, SDV boost, domain filter both
Day 23: fix automotive domain filter - no cross-domain fallback, all automotive bridge actions populated
Day 23: fix roadmap parse failure - increase max_tokens, user-friendly error message
Day 23: course corpus updated - 108 courses, metadata updated, level field standardised
Day 23: fix roadmap double-click, TypeError in course formatting, Screen 6 complete
Day 22-23: BG checker via agent - Screen 6 complete, observations noted
Day 22-23: BG checker via agent - Screen 6 complete, observations noted
Day 22: course corpus enriched - 97 courses, 23 BG-eligible confirmed
Day 22: BG checker simplified - no rules engine, guidance only
Day 21: clean up app.py - remove redundant imports and client after agent refactor
Day 21: decisions log updated
Day 21: agent orchestrator layer - fixed sequence pipeline with state management
Fix gitignore formatting - separate merged lines
Week 4-5: eval baseline 87/87, course corpus 84 courses, prompt improvements, temperature=0
Add canonical readiness level definitions to decisions log
Day 20: mid-project review decisions finalised
Day 19: decisions, learnings and future ideas updated
Day 19: roadmap generator complete - RAG grounded in course corpus
Day 18: course corpus - 84 unique courses across 4 role families
Add corpus health check script - all distributions balanced
Days 16-17: eval framework learnings added
Days 16-17: decisions log consolidated, future.md updated
Day 16-17: eval framework 15 cases, classification 87%, gap analysis 87%
Day 14-15: fix top 6 user feedback issues - labels, captions, dropdown, rationale
Day 14: user feedback round 1 documented
Week 3: document ESCO not used in current pipeline
Week 3: engineering learnings documented
Week 3: update corpus inventory and future ideas
Week 3: add learnings.md reference to context handoff