Access Required
Policy Tracker HQ
Build dashboard · Restricted access
Email
Password
Invalid credentials
Build HQ
Dashboard.
Sessions
03
completed to date
Pipeline
Live
crawl → score → brief
Phase 1
25%
engine in progress
Next session
Mon
Remon first PR + DNS
Build Status Phase 1 — The Engine
Repo + scaffold — Next.js 15, TypeScript, Tailwind
Supabase schema — 7 tables, RLS, migrations
Ingestion pipeline — crawl, score, store
Brief generation — Sonnet, bilingual EN/中文
Vercel deploy — app live
Domain — policy-tracker.dev, DNS pending
Dashboard UI — not started
DingTalk delivery — not started
Auth + login — not started
Conversational agent — not started
Exports — PPTX, XLSX, PDF
July 2026
Mon
Tue
Wed
Thu
Fri
Sat
Sun
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Next
Session 04 — Remon first PR + DNS
remon/brief-generation-test · connect policy-tracker.dev to Vercel
Session history
Build Log
Session 01 Jul 1 Gabi
Full infrastructure setup and first working pipeline
Built
  • GitHub repo + Next.js 15 (App Router, TypeScript, Tailwind)
  • CLAUDE.md + DECISIONS.md from day one
  • Supabase — 7 tables, RLS, org-scoped roles
  • Pipeline: crawl (Tavily) → score (Haiku) → store
  • First Vercel deploy live
Fixed
  • Next.js resolved to v16 — pinned to 15.5.19
  • ES module hoisting — lazy Anthropic client init
  • server-only guard broke scripts — doc comment
Learned
  • Service-role client for all pipeline writes, never browser client
  • Zod schema for structured AI output
  • Lazy client init pattern for API clients in Next.js
crawl → score → store — reusable template for any AI ingestion project
NextBrief generation (Sonnet, bilingual EN/中文) + Remon onboarding
Jul 1
Session 02 Jul 2 Gabi + Remon
Brief generation built · Remon onboarding partial
Built
  • lib/pipeline/briefs.ts — bilingual Sonnet, Zod, prompt caching
  • sources insert in store.ts — traceability enforced in code
  • feature/brief-generation — security reviewed, merged to dev
BlockerCorporate VPN blocked Claude.ai OAuth (403) — resolved in S03
Learned
  • Security review required on all pipeline PRs before merge
  • sources table is the gatekeeper — no source URL, no brief
bilingual EN/中文 generation in one prompt — portable to any content pipeline
NextResolve auth → Prompt 1 + 2 → first PR
Jul 2
Session 03 Jul 3 Gabi + Remon
Remon authenticated — Block 2 complete
Fixed
  • Corporate VPN — switched to personal Mac
  • GitHub auth via VS Code extension (automatic on Mac)
Remon completed
  • Repo cloned on personal Mac
  • Claude Code authenticated
  • Prompt 1 — understood the project
  • Prompt 2 — understood the pipeline architecture
Learned
  • Work laptop (corporate VPN) = wrong environment for dev
  • CLAUDE.md gives Claude Code full project context automatically
clone → Prompt 1 → Prompt 2 → standard onboarding for any Claude Code project
NextPrompt 3 — remon/brief-generation-test + DNS for policy-tracker.dev
Jul 3
Mon
Session 04 Mon Jul 7 Remon
First PR · DNS · remon/brief-generation-test
Full scope
Roadmap
Phase 1 The Engine — In progress
The first usable product — the full monitoring, scoring, and briefing engine, live on public data. Starting coverage EU + US.
Repo + scaffold
Next.js 15, Supabase, Vercel, CLAUDE.md
Ingestion pipeline
Crawl (Tavily) → Score (Haiku) → Store
Brief generation
Sonnet, bilingual EN/中文, 5 sections, source-traced
Domain + China access
policy-tracker.dev — DNS pending
Auth + login
Magic link, admin/member roles, RLS
Dashboard UI
Filterable by region, tier, urgency, type, status
Policy detail view
Brief sections, lifecycle timeline, sources
DingTalk delivery
Webhook alerts for tier 1 + 2 items
Email delivery
Daily digest or immediate alert via Resend
Conversational querying
Natural language queries over the tracked landscape
Exports
PPTX deck, XLSX, PDF — on demand
n8n scheduling
Automated pipeline runs, multiple daily
Coverage is a dial, not a phase. EU + US at launch. Each new market requires a source list and OPEX confirmation before adding.
Phase 2 Depth & Access — Not started
Higher-complexity layers — new source modalities and a new security model. Begins only when Phase 1 is complete and deployed.
Social-media monitoring
Official accounts only — new source modality
Legal Summary section
Higher-rigor, AI-drafted, human-verified, source-linked
Platform Responsibility
Per-policy obligations for platforms
Merchant Responsibility
Per-policy obligations for merchants
Recommended Action
Actionable guidance, AI-drafted, human-verified
Role-based access
Field-level RLS — new security model
Deeper querying
Full agent across the whole tracked landscape
Broader market coverage
Markets beyond EU + US, by agreement
Phase trigger: a new source modality (social) or a new security model (field-level RLS). Adding markets alone does not trigger a new phase.