fix: backfill symbol_demand for sidebar-added symbols + analyst ratings schema fix

- Add await ctx.cache.subscribe() to addSymbol mutation so symbols
  added via the sidebar get registered in symbol_demand and yfinance
  jobs are queued immediately
- Backfill PEP, WYNN, STZ, CELH into symbol_demand + adapter_queue
- Upgrade yahoo-finance2 3.15.3 -> 3.15.4 and pass validateResult:false
  to quoteSummary() to handle Yahoo schema drift
- Add error detail logging for analyst ratings schema failures
- Update .gitignore with common ignores
This commit is contained in:
Investor Flow Build
2026-07-23 18:02:24 -04:00
parent 5ef2b2f060
commit e262187c3c
204 changed files with 25014 additions and 2934 deletions
+194
View File
@@ -0,0 +1,194 @@
# Investor Flow — Functional Design (as built)
**Audit date:** 2026-07-18
**Source of truth for status:** this file + code under `app/`. Older slice SPECs, HANDOFF.md, and Automaton WORKFLOW.md are historical.
**Product intent:** Beginner-first multi-tenant educational research terminal for conviction investing (not day trading). Legal posture: educational publisher (ADR-0007).
---
## Status legend
| Status | Meaning |
|--------|---------|
| **Live** | Backend + UI end-to-end usable for the intended job |
| **Backend-ready** | Engine and/or tRPC exist with tests; little or no product UI |
| **UI-local** | Screen exists but state is browser-only (localStorage / Zustand), not multi-device / multi-tenant |
| **Stub** | Surface or procedure exists but incomplete (empty timeline, thin universe, etc.) |
| **Missing** | Designed in domain model; not implemented as a product path |
---
## 1. User jobs and surfaces
### 1.1 Identity, access, onboarding
| Capability | Status | Notes |
|------------|--------|-------|
| Email/password signup | **Live** | New users get `status=pending_approval`; no session until admin approves |
| Login + session cookie | **Live** | Non-active users rejected at login and `protectedProcedure` |
| TOTP 2FA + backup codes | **Live** | Settings page |
| GitHub / Google OAuth | **Backend-ready** | Router complete; env-dependent; OAuth users default `active` |
| Onboarding wizard | **Live** | Complexity, risk, drawdown, starter watchlist, optional portfolio; queues SEC fetch |
| Admin approve/reject | **Live** | `/admin/users` |
| Settings / profile | **Live** | `/settings` |
### 1.2 Research workbench (desktop)
| Surface | Route | Status | Data path |
|---------|-------|--------|-----------|
| Symbol overview | `/` | **Live** | `market.snapshot`, company meta, X feed, collapsible filings + options |
| Charts | `/chart-lab` | **Live** | candles, indicators, institutional buy markers |
| Institutional dashboard | `/institutional` | **Live** | dashboard rollup, flow, insider stream, filings; Q-o-Q / M-o-M |
| Filings | `/filings` | **Live** | EDGAR index + 13F + Form 4 detail |
| Options DD | `/options-dd` | **Live** | chain + greeks (read-only teaching surface) |
| Market outlook | `/market-outlook` | **Live (Phase 1–3 core)** | Beginner language. Condition strip; stronger/weaker-vs-market map; seasonality + calendar; rank snapshots; auto leadership check; truck/factory; macro notes. True ETF flow/COT still later. |
| Social / X feed | on overview | **Live** | DB-backed 30d history + admin X credentials/accounts |
| Command palette + nav | shell | **Live** | ⌘K, g-key shortcuts, theme presets |
### 1.3 Process / journal workflow
| Surface | Route | Status | Gap |
|---------|-------|--------|-----|
| Trade plan | `/trade-plan` | **UI-local** | Zustand only; not `theses` / `trades` / sizing APIs |
| Execution | `/execution` | **UI-local** | Playbook + checklist local; not `trade_executions` |
| Trade closure | `/trade-closure` | **UI-local** | localStorage |
| Daily focus / goals | `/daily-focus` | **UI-local** | localStorage goals |
| Emotion logger | on execution | **Partial** | UI + `emotionLogger` API exist; UI still primarily local path |
| Sizing math | `/risk` (calculator) | **Live** | `sizing.compute` + UI cascade (math only, ADR-0007) |
| Risk posture (M20) | `/risk` | **Live** | `risk.posture` / `risk.haltStatus` + RiskPosturePanel; halt persisted on drawdown breach |
| Derisking suggestions | (no screen) | **Backend-ready** | `derisking.suggest` |
| Thesis monitor | (no screen) | **Stub** | `thesisMonitor.assess` / `timeline` (empty events / empty timeline) |
### 1.4 Discovery & strategy lab
| Capability | Status | Notes |
|------------|--------|-------|
| Filter screener (M15a) | **Backend-ready** | `screener.filter` over watchlist quotes only; **no UI**; not in frontend `api` client |
| Strategy screener (M15b) | **Backend-ready** | `screener.strategy`; thin universe; no UI |
| Strategy authoring | **Backend-ready** | `strategies.list/create`; no Strategy Lab UI |
| Backtest | **Backend-ready** | `backtest.run` / `evaluateLatest`; no equity-curve UI |
| Sector confirmation cross-link | **Stub** | `sectorCrosslink.confirm` with empty universe in practice |
| Screener saved filters | **Schema only** | tables exist; incomplete product loop |
### 1.5 Options convexity sleeve (M17)
| Capability | Status | Notes |
|------------|--------|-------|
| Options DD (read-only) | **Live** | M3 teaching panel |
| Unlock ladder + payoff API | **Backend-ready** | `optionsConvexity.unlock` / `getPayoff` |
| Option legs book + risk contribution | **Live (MVP)** | `portfolio.optionLegs` / add / remove; Risk posture premium-at-risk + CSP cash reserve |
| Portfolio sleeve workflow (M17 full) | **Missing** | Unlock UX + sleeve risk budget not enforced in product path |
### 1.6 Macro (M18)
| Capability | Status | Notes |
|------------|--------|-------|
| FRED series / key admin | **Live** | Admin X Accounts page also manages FRED key |
| Truck sales + manufacturing PMI charts | **Live** | Market routes |
| Sector rotation heatmap-style panel | **Live** | Relative-strength style rotation, not full designed phase library UI |
| Regime classify / history | **Backend-ready** | `macro.regimeClassify`, `regimeHistory` |
| Portfolio-impact commentary | **Live** (partial) | `macro.commentary` via Market Outlook; **2 unit tests failing** on wording assertions |
| Full economic calendar UI | **Stub** | `macro.calendar` |
### 1.7 Alerts & mobile
| Capability | Status | Notes |
|------------|--------|-------|
| Alert list / ack / unacked count | **Live** | Used on mobile + rotation alert chip |
| Rotation signal → alert create | **Live** | `market.rotationCheckForAlert` |
| Full hybrid AlertEngine (Form4, thesis, cluster, halt…) | **Partial** | Engine types/tests exist; not all event producers wired continuously |
| SSE push | **Missing** | Design called for SSE; poll/list only |
| Mobile companion `/mobile` | **Live (thin)** | Auth, holdings glance, unacked alerts — not full M19 job A vision |
### 1.8 Admin / operator
| Capability | Status |
|------------|--------|
| Users, sessions, reset password | **Live** |
| Pending approval queue | **Live** |
| Adapter queue health, pause/resume, retry, schedules, error stacks | **Live** |
| SEC queue fetch + lint holders/insiders + data quality | **Live** |
| X credentials, accounts, prune | **Live** |
| FRED key | **Live** |
| Audit log | **Live** |
| Server restart control | **Live** |
| GDPR export | **Backend-ready** (admin API) |
### 1.9 Reports
| Capability | Status |
|------------|--------|
| HTML research note generator | **Backend-ready** | `reports.generate` scopes include symbol/watchlist/portfolio/rotation/sizing/risk — **no UI**, not in frontend client |
---
## 2. Functional architecture (intended jobs)
```
Job A — Monitor (mobile + alerts)
alerts.list/ack · portfolio.holdings · rotation chip
Job B — Research (desktop workbench)
watchlist → active symbol → market + SEC + institutional + options + social
market outlook (macro + rotation)
Job C — Process (planned investor loop)
thesis + confluence + sizing → plan → execute → emotion → close → unlock
[mostly UI-local today; engines exist but not productized]
Job D — Discover (planned)
filter/strategy screener → open in workbench
[API only]
```
---
## 3. Design principles still in force
Unchanged from domain model / ADRs:
1. **Primary Rule (ADR-0007)** — education, not advice; Primary-Rule lint test on curated strings.
2. **Analyst Voice (ADR-0005)** — Alfred/Druckenmiller blend where LLM copy is used.
3. **P6 plain English / P7 teach mechanics** — UI redesign token system supports this.
4. **Local-first multi-tenant (ADR-0001)** + shared cache dedupe (ADR-0004).
5. **LLM data provenance (ADR-0006)** — sensitive data stays on local-class providers; full gateway module is thin/incomplete vs original design.
6. **Anti-gamification** — no confetti / outcome celebration.
---
## 4. Explicit non-goals (still)
- Brokerage execution or custody
- Investment adviser recommendations
- Mobile-first full research workbench
- Undefined-risk options strategies for beginners
- Training on user portfolio/thesis content via external LLMs
---
## 5. Gap priority (product value)
| P | Gap | Why it matters |
|---|-----|----------------|
| P0 | Fix failing MacroRegime commentary tests | **Done 2026-07-18** — suite green (504 pass / 1 skip) |
| P0 | Expose SizingEngine + RiskEngine on tRPC + M20 UI | **Done 2026-07-18** — `/risk` + `sizing`/`risk` routers |
| P1 | Persist trade plan / thesis / execution / closure server-side | Today process screens are single-browser toys |
| P1 | Frontend client coverage for backend-only routers | APIs unreachable from SPA client |
| P1 | Wire emotion logger UI to `emotionLogger` API | Schema + API exist; dual paths confuse |
| P2 | Strategy Lab + backtest UI | Engines exist; no learning surface |
| P2 | Screener UI (filter + strategy) | Discovery job unfinished |
| P2 | Derisking + thesis monitor product surfaces | Process depth |
| P2 | Options convexity sleeve UI | M17 incomplete |
| P3 | Reports UI | Export pedagogy |
| P3 | Schema cleanup (duplicate `strategies` DDL) | Maintainability |
| P3 | SSE alerts / continuous event producers | Ops polish |
| P3 | Full LLM Gateway as designed | Provenance enforcement completeness |
---
## 6. Related docs
- Tech design (modules, stack, API map): [`TECH_DESIGN.md`](./TECH_DESIGN.md)
- Domain glossary: [`../CONTEXT.md`](../CONTEXT.md)
- ADRs: [`adr/`](./adr/)
- Living wiki page: Obsidian `investor-flow.md` (synced with this audit)
+268
View File
@@ -0,0 +1,268 @@
# Investor Flow — Technical Design (as built)
**Audit date:** 2026-07-18
**Companion:** [`FUNCTIONAL_DESIGN.md`](./FUNCTIONAL_DESIGN.md)
**Canonical code roots:** `app/src` (Next frontend), `app/server/src` (Node backend)
---
## 1. Stack
| Layer | Actual technology | Notes vs older docs |
|-------|-------------------|---------------------|
| Frontend | Next.js **16.2**, React **19**, Tailwind **v4**, Recharts, Zustand, Radix UI | Not Bun SPA |
| Backend runtime | **Node ≥22** (dev on Node 26), native TS via `--experimental-strip-types` | DESIGN said Bun; runtime glue is Node + `node:sqlite` |
| API | **tRPC v11** HTTP, path `/api/trpc/*` | Frontend uses hand-rolled fetch client in `app/src/lib/trpc.ts` (not `@trpc/client`) |
| DB | SQLite file (`app/server/data/investor-flow.db`) | Single file multi-tenant |
| Market data | `yahoo-finance2` **v3** class API | |
| SEC | EdgarAdapter + SecFetchAdapter + SecLintAdapter + `secDataFetcher` | Filer CIK from accession prefix (critical fix) |
| Social | XCookieAdapter (encrypted ct0/auth_token), RedditAdapter | |
| Macro | FredAdapter | Key stored admin-side |
| Auth | Sessions + TOTP + OAuth (GitHub/Google) | Password hash via local crypto helpers |
| Tests | `node --test --experimental-strip-types` | Backend ~503 tests; **500 pass, 2 fail, 1 skip** as of audit |
| Deploy | `docker-compose.yml` (frontend + backend + volume) | Present; operator-run |
---
## 2. Process topology
```
Browser (Next :3000)
│ fetch /api/trpc/<proc> (credentials include)
▼
Next rewrites / proxy → Backend (:3001)
│
├─ appRouter (tRPC)
├─ CacheRepository + AdapterQueue
├─ Source adapters (yfinance, sec-*, x, reddit, fred)
└─ SQLite
```
- Dev helpers: `restart-servers.sh`, admin `serverRestart`.
- Queue drain loop + 30s schedule loop inside backend process.
---
## 3. Multi-tenant data tiers (still valid)
| Tier | Content | Isolation |
|------|---------|-----------|
| A | quotes, candles, options, filings, institution_filings, insider_tx, sector_map, macro | Shared, no owner |
| B | threads, x_cookie_posts, reddit_posts, adapter_queue*, queue_* | Shared infrastructure |
| C | watchlists, portfolio_holdings, trades, strategies, alerts, theses, emotion_logs, reports, saved filters | `owner_id` / `user_id` |
| D | users, sessions, admin_audit, x_credentials (singleton), llm_* | System |
**Demand set:** `symbol_demand` refcount drives which symbols the queue refreshes.
**Stale-while-revalidate:** CacheRepository returns cached rows and schedules refresh when past TTL.
### 3.1 Rate-limit-first data plane (ADR-0009)
All vendors (Yahoo, X, FRED, SEC, Reddit) have short rate limits. The system is designed so stress yields **stale/static UI**, not stampede:
| Rule | Mechanism |
|------|-----------|
| Request path never stampede | Serve SQLite / `kv_cache` / static fallback first; short timeout if live is unavoidable |
| One outbound owner | `AdapterQueue` only (background drain); UI schedules via `CacheRepository.get` |
| Steady throttle | Per-source min-interval (`sourceRatePolicy.DEFAULT_SOURCE_MIN_INTERVAL_MS`) |
| 429 cool-down | Source-wide pause 2→5→15→30→60 min; skip all jobs for that source; no schedule flood |
| Demand-bounded work | Only `symbol_demand` symbols get scheduled refresh |
| Observability | `queue.health().sourceCooldowns` |
**Anti-patterns (do not reintroduce):** N parallel Yahoo charts on a click; live `quoteSummary` without cache on every panel open; treating 429 as a 2s job retry that keeps hammering the same edge.
Slow-changing composition (ETF top holdings) uses `kv_cache` + `etfHoldingsFallback.ts`. Live upgrade is best-effort when the source is not cooling down.
---
## 4. Backend module map
```
app/server/src/
index.ts HTTP server entry
trpc/router.ts ~2k LOC — all routers
trpc/context.ts session, cookies, db, cache, queue injection
db/schema.sql DDL (note: duplicate strategies blocks — debt)
db/*Repository.ts watchlist, portfolio, emotion logs
cache/CacheRepository.ts content-addressed cache API
queue/AdapterQueue.ts schedule, pause, retry, source cool-downs (ADR-0009)
queue/sourceRatePolicy.ts 429 detect, cool-down ladders, min-intervals
adapters/ YFinance, Options, Edgar, SecFetch, SecLint, X, Reddit
analysis/ indicators, rotation, seasonality, etfHoldingsFallback, tickerContext
admin/ operator functions + CLI
auth/ totp, oauth, backup codes
alerts/AlertEngine.ts
risk/RiskEngine.ts + haltCircuitBreaker.ts ← pure; NOT on tRPC
sizing/SizingEngine.ts + convictionUnlock + twoAxisMatrix ← pure; NOT on tRPC
strategy/BacktestEngine.ts
screener/UniverseEvaluator.ts + SectorCrosslink.ts
options/ConvexityGate.ts
derisking/DeriskingEngine.ts
thesis/ThesisMonitor.ts
macro/FredAdapter.ts + MacroRegime.ts
reports/ReportRunner.ts
onboarding/starter.ts
services/secDataFetcher.ts
x/backfill.ts
```
### 4.1 tRPC surface (mounted)
| Router | Procedures (summary) |
|--------|----------------------|
| `auth` | signup, login, logout, me, enable2fa, confirm2fa, oauthStart, oauthCallback |
| `onboarding` | starter, complete |
| `market` | snapshot, candles, indicators, truckSales, manufacturingPmi, rotation, rotationCheckForAlert, sectorHoldings |
| `dashboard` | rollup |
| `admin` | users, sessions, passwords, queue*, lint, data quality, X creds/accounts/prune, FRED key, pending/approve/reject, audit, serverRestart, gdprExport |
| `alerts` | list, acknowledge, acknowledgeAll, unackedCount |
| `institutional` | flow, insiderStream, ownershipHistory, buyEvents |
| `edgar` | filings_index, company_facts, filer_cik_meta, full_text_search, form13f_holdings, form4_tx |
| `watchlists` | list, addSymbol, removeSymbol |
| `portfolio` | holdings, addHolding, removeHolding |
| `options` | chain, greeks |
| `reports` | generate |
| `screener` | filter, strategy |
| `strategies` | list, create |
| `backtest` | run, evaluateLatest |
| `sectorCrosslink` | confirm |
| `optionsConvexity` | unlock, getPayoff |
| `derisking` | suggest |
| `macro` | series, calendar, regimeClassify, commentary, regimeHistory |
| `thesisMonitor` | assess, timeline |
| `x` | feed, cashtag_search, timeline, accountsForSymbol |
| `reddit` | subreddit, search |
| `emotionLogger` | add, getByTrade, delete |
### 4.2 Sizing + risk (mounted 2026-07-18)
| Router | Procedures |
|--------|------------|
| `sizing` | `compute` — pure `sizePosition` + portfolio/unlock/halt context |
| `risk` | `posture` — pure `assessRisk` + portfolio load + halt persist; `haltStatus` |
Still deferred: journal create path that throws `HaltedError` when plans are server-persisted (`persist-trade-plan-execution-loop`).
---
## 5. Frontend architecture
```
app/src/
app/ Next App Router pages
components/ Panels + ui kit + layout
stores/ Zustand (several persist to localStorage)
lib/trpc.ts Partial API client (subset of routers)
lib/chart-theme.ts CSS-variable themed Recharts
lib/strings.ts Primary-rule sensitive copy
```
### 5.1 Routes
| Path | Role |
|------|------|
| `/` | Workbench overview |
| `/chart-lab` | Charts |
| `/institutional` | Institutional + filings |
| `/filings` | Filings only |
| `/options-dd` | Options DD |
| `/market-outlook` | Macro / rotation / notes |
| `/trade-plan`, `/execution`, `/trade-closure`, `/daily-focus` | Process (local) |
| `/settings` | Auth + onboarding |
| `/mobile` | Thin companion |
| `/admin`, `/users`, `/queue`, `/audit-logs`, `/x-accounts` | Admin |
### 5.2 Client coverage gap
`lib/trpc.ts` exposes: market, edgar, auth, onboarding, watchlists, options, portfolio, dashboard, institutional, emotionLogger, alerts, x, admin.
**Not exposed in client (but on server):** screener, strategies, backtest, sectorCrosslink, optionsConvexity, derisking, macro (except commentary aliased under `market.commentary`), reports, thesisMonitor, reddit, some admin helpers (e.g. gdprExport).
---
## 6. Caching & adapters
| Source kind | Role | Typical freshness |
|-------------|------|-------------------|
| yfinance | quote, candles, sector, options | minutes / EOD |
| sec / sec-fetch | filings, 13F, Form 4 bulk | daily schedule default |
| sec-lint-* | gap detection / backfill | on demand admin |
| x | timelines / cashtags | demand + prune |
| reddit | subreddit/search | on demand |
| fred / macro | series + commentary inputs | slower |
| llm | summaries (tables present) | sparse product use |
**AdapterQueue features (built):** pause/resume, per-job errors with stacks, retry job/source, clear done, schedules, startup in_flight→pending recovery.
---
## 7. AuthZ model
| Gate | Behavior |
|------|----------|
| `publicProcedure` | No session required (many market/edgar/watchlist reads still public — intentional for local demo; tighten later if multi-tenant hardens) |
| `protectedProcedure` | Session + `users.status === 'active'` |
| `adminProcedure` | `is_admin` flag |
Watchlist list currently falls back to `userId ?? 'anonymous'` — convenient for local dev, weak isolation if exposed beyond localhost.
---
## 8. Testing
| Suite | Command | Audit result |
|-------|---------|--------------|
| Backend | `cd app/server && npm test` | **504 pass / 0 fail / 1 skip** (after P0 2026-07-18) |
| Frontend lite | Primary-Rule lint + options pure tests | Present |
Known red: `MacroRegime` commentary tests (`generateMacroCommentary` short/long-term + no macro-trade recommendation).
---
## 9. Schema debt
1. **`strategies` defined twice** in `schema.sql` (legacy `regime_gate/setup/risk_policy` vs `components/unlocked`). SQLite keeps first-created shape depending on migration history — dangerous drift.
2. Parallel filter tables: `screener_filters` vs `saved_filters`.
3. Process tables (`trades`, `trade_executions`, `theses`, `emotion_logs`) exist ahead of UI persistence.
---
## 10. LLM / provenance (as built vs design)
| Designed | As built |
|----------|----------|
| Full LLMGateway with classifyPayload gate | Tables `llm_providers`, `llm_dispatch_audit`, `llm_summaries`; no complete gateway module under `src/` |
| Ornith default `is_local=true` | Env-driven URL in docker-compose (`LLM_PROVIDER_URL`) |
| Sensitive thesis data never leaves host | Thesis content mostly not yet flowing through LLM product paths |
ADR-0006/0008 still govern intent; implementation is incomplete.
---
## 11. Deployment
```bash
# Local dev (typical)
cd app/server && npm run dev # :3001
cd app && npm run dev # :3000
# Compose
docker compose up --build
```
Secrets: `.env.example` documents session secret, SEC operator email, X cookie fields, Reddit, FRED, LLM URL.
---
## 12. Documentation map (after this audit)
| Doc | Role |
|-----|------|
| `docs/FUNCTIONAL_DESIGN.md` | What users can do / pending |
| `docs/TECH_DESIGN.md` | This file — how it is built |
| `CONTEXT.md` | Ubiquitous language (glossary only) |
| `docs/adr/*` | Durable decisions |
| Obsidian `investor-flow.md` | Session orientation wiki |
| `.automaton/tasks/*` | Work items; many `complete/` slices are historical |
Deprecated as status sources: root `HANDOFF.md` (2026-06-30 orchestrator snapshot), Automaton WORKFLOW “21 tasks” pool list, slice DECOMPOSITION checkbox state.
@@ -0,0 +1,79 @@
# ADR-0009: Rate-limit-first data plane
Date: 2026-07-19
Status: Accepted
## Context
Every external vendor we use has a short rate limit:
| Source | Typical limit / failure mode |
|--------|------------------------------|
| Yahoo Finance (`yahoo-finance2`) | Edge 429 / "Too Many Requests" under concurrent chart+quote+summary |
| X (bird CLI / cookie session) | HTTP 429 on search / timeline |
| FRED | API key quota; burst-sensitive |
| SEC EDGAR | Fair-access pacing (~10 req/s official guidance) |
| Reddit | OAuth / public endpoint throttles |
The product already had **shared cache + queue dedupe** (ADR-0004) and **stale-while-revalidate**, but request handlers still opened **live vendor calls** (ETF top holdings charts, condition VIX, peers) and the drain loop treated 429 like a normal error with multi-second job backoff. Result: thrash → empty UI panels → worse rate limits.
## Decision
Design the data plane around rate limits as a first-class constraint:
### 1. Request path never stampede
tRPC handlers **prefer local state**:
1. SQLite / `kv_cache` / typed tables (quotes, candles, …)
2. Stale-while-revalidate via `CacheRepository.get` (schedule background refresh)
3. **Static / curated fallbacks** for slow-changing composition (e.g. ETF top holdings)
4. Live vendor only when nothing local exists, with a **short timeout** and graceful empty/stale result
UI clicks must not fan out N charts or N quoteSummary calls.
### 2. One shared queue owns outbound pacing
All background refreshes go through `AdapterQueue`:
- **Per-source min-interval** between fetches (steady-state throttle)
- **Source-wide cool-down** on rate-limit signals (2 → 5 → 15 → 30 → 60 minutes escalating)
- While cool-down is active: **skip all jobs for that source**; do not enqueue schedule floods
- Job exponential backoff remains for *ordinary* failures only
- Rate-limit hits **do not burn** `MAX_ATTEMPTS` into permanent `failed` without a long cool-down first
### 3. Demand set bounds work
Only symbols in `symbol_demand` (watchlist ∪ holdings) get scheduled yfinance/sec/x refresh. Breadth of interest, not user count, drives cost (ADR-0001 / CONTEXT demand set).
### 4. Stale is better than empty
Showing yesterday’s holdings weights or a 10-minute-old quote with a “cached” affordance beats a blank panel that hammers the vendor. Education product (ADR-0007) does not require millisecond freshness for composition / macro context.
### 5. Observability
Queue health exposes active **source cool-downs** so operators can see “Yahoo paused 4m” instead of a pile of failed jobs.
## Implementation map
| Piece | Location |
|-------|----------|
| Policy helpers (detect 429, ladders) | `app/server/src/queue/sourceRatePolicy.ts` |
| Cool-down + drain skip | `app/server/src/queue/AdapterQueue.ts` |
| ETF holdings cache + static fallback | `market.sectorHoldings` + `etfHoldingsFallback.ts` |
| FRED series write-through | `kv_cache` in condition path |
| Shared market cache | ADR-0004, `CacheRepository` |
## Consequences
- **Positive:** Under vendor stress the UI keeps serving cache/static; queue self-throttles instead of amplifying 429s; operators see cool-downs.
- **Positive:** New features have a clear rule: “cache first, queue refresh, static if needed.”
- **Trade-off:** After a cool-down, data may be minutes-to-hours stale until the next successful drain.
- **Trade-off:** Static ETF weights drift until the next successful Yahoo refresh (acceptable for education peek panels).
- **Follow-ups:** Route remaining live Yahoo calls in `market.condition` / peer resolution through the same cache-or-queue pattern; surface cool-downs in admin queue UI.
## Rejected alternatives
- **Retry harder on 429:** Makes thrash worse.
- **Per-user fetch with no shared cool-down:** Multiplies load; violates ADR-0004.
- **Block UI until live succeeds:** Timeouts and empty states; bad UX for a terminal-style education app.