Dashboard/PRISM/TODO.md
PRISM

TODO.md

PRISM TODO — Updated 2026-04-14

intel-company (Hub Module) — IN PROGRESS

Code

  • [x] 3-track pipeline implemented (Track 1 WebFetch, Track 2 Perplexity, Track 3 Gemini)
  • [x] Shared browser infrastructure (BrowserClient, tiered)
  • [x] Pipeline health log
  • [x] Schema: social fields, exec LinkedIn, competitor ticker/social
  • [x] Migration 006 applied
  • [x] Cache-first smoke test with --refresh flag
  • [ ] Wire Track 3 into activities.py (Temporal)
  • [ ] Tier 2 Playwright stealth (blocked until search audit module)

Tests

  • [ ] Unit: BrowserClient tier escalation
  • [ ] Unit: PipelineHealthLog event capture + markdown render
  • [ ] Unit: SynthesisClient mock Gemini
  • [ ] Integration: 3-track full run with VCR cassettes
  • [ ] Contract: CompanySeedOutput schema stability

UI

  • [x] Company identity card view (5-layer CompanyIntelCard — identity anchor, execs, brand portfolio, competitors, online presence)
  • [x] UIX Lego Blocks library created (vault: UIX-Lego-Blocks/, 9 components with specs + HTML demos)
  • [x] Design tokens system (frontend/lib/tokens.ts — Algolia brand enforcement, Sora 300/400/600, 12px min)
  • [ ] Pipeline health badge (HEALTHY/DEGRADED/FAILED)
  • [ ] Cache freshness indicator + refresh button
  • [ ] Discovery Call Prep card — Sales Actions tab (see below)

Spoke Modules — NOT STARTED

  • [ ] intel-social
  • [ ] intel-investor
  • [ ] intel-partner
  • [ ] intel-industry
  • [ ] intel-competitors
  • [ ] intel-queries

Discovery Call Prep — 2026-04-15

PRISM is a prospect enrichment platform, not a sales execution platform. It feeds INTO the AE's sales process — it doesn't run it. MEDDPICC fields like Decision Criteria, Decision Process, and Paper Process are internal to the prospect and only discoverable through direct conversation. PRISM should not pretend to fill those in.

What PRISM contributes (short term — this sprint)

  • [ ] Discovery Call Prep card for Sales Actions tab — focuses on 3 things PRISM actually knows:
  • Pain points: top 3 search gaps from audit with evidence + screenshots
  • ROI ammunition: pre-built business case levers (conversion lift, AOV, bounce)
  • Competitive landscape: who they're fighting, what search tech each uses, Golden Angle
  • [ ] Suggested discovery questions — NOT to fill MEDDPICC, but to help AE learn what PRISM can't see (budget owner, buying process, timeline)

Future considerations

  • [ ] Enrich executive role badges with company-size context ("likely signs independently" vs "needs escalation" based on employee count + title seniority)
  • [ ] Champion candidate strength signal — from LinkedIn activity (intel-social) + public digital transformation advocacy (intel-investor quotes)
  • [ ] Procurement complexity estimate — rough Low/Medium/High based on company size + regulated industry flag — helps AE calibrate timeline expectations

Open Architectural Decisions

  • Decision 1: merge.py MergedField vs flat suffix
  • Decision 5: audit-browser scope
  • Decision 8: eval golden fixtures
  • Spoke build order

Environment Notes

  • Docker: must be running before any test/run
  • Venv: rebuilt 2026-04-14 at current path (was broken — pointed to old Google Drive) Use: .venv/bin/python3
  • Alembic: through migration 006
  • Both prospects seeded: nike.com + orientaltrading.com