Execute consolidated expansion plan: metadata, discovery vault, and governance UX

[simonperegrinvs]

Summary

This PR executes the consolidated plan in task-by-task commits with quality gates after each task.

Task 1

• Baseline implementation + node description backfill for new research nodes (P47-P61).
• Added consolidated plan and research artifacts.

Task 2

• Enforced node description quality in ingestion and reporting:
  • node add --description --require-description
  • node update --description-file
  • ref import --require-node-description
  • report brief now includes node descriptions.

Task 3

• Added AI-assisted free-source discovery pipeline and vault tooling:
  • discover run (OpenAlex + Semantic Scholar + CrossRef + arXiv)
  • default vault base path: ~/data/modularpsi
  • run artifacts + audit logs under vault runs
  • snapshot safety commands: snapshot create/list/rollback
• Added discovery unit tests and utility layer for dedupe + mapping.

Task 4

• Added node metadata completeness indicator + missing-description warning in UI node panel.

Task 5

• Added scalable graph visualization controls:
  • family filtering (root-child cluster lens)
  • evidence metadata filters (effect, study type, replication status)
  • edge trust threshold slider
  • focus mode with configurable k-hop depth
  • recent-change overlay and recent-only filter
• Added node hover description previews and recent highlighting in graph nodes.

Task 6

• Strengthened discovery governance:
  • strict validation mode for references in discovery runs
  • hard guardrails (max-new-refs, max-new-nodes, max-trust-delta)
  • richer run artifacts (hypothesis-rank.json, evidence-brief.md)
• Added scheduled CI workflow .github/workflows/discovery-daily.yml for daily discovery dry-runs with artifact upload.

Task 7

• Added backend env gate import and .env support:
  • discover run non-dry apply path now requires MPSI_AGENTIC_ENABLED=1
  • CLI auto-loads .env / .env.local
  • new env helpers in src/cli/env.ts
• Added env-driven orchestration script:
  • npm run agentic:daily (snapshot + discover + rank + brief)
  • .env.example includes MPSI_AGENTIC_ENABLED=1 and default agentic settings
• Updated workflow/docs:
  • discovery daily workflow now runs agentic:daily with explicit env
  • README/CLI docs updated for env-driven flow.

Quality Gates

Executed after each task and again at the end:

• npm run test ✅
• npm run build ✅
• npm run lint ✅ (pre-existing warnings only, no lint errors)

Commit Structure

1. Task 1: baseline implementation + node description backfill
2. Task 2: enforce node descriptions in ingestion and reporting
3. Task 3: add vault-based discovery pipeline and snapshot commands
4. Task 4: add node metadata completeness and missing-description warning
5. Task 5: add scalable graph view controls (family/evidence filters, focus, recency, edge threshold)
6. Task 6: add strict discovery guardrails and scheduled dry-run workflow
7. Task 7: add env-driven backend gate and agentic daily runner

Notes

• Discovery dry-run validated with vault output at ~/data/modularpsi.
• No paid or restricted data sources are used.