wr-plutus — Session 6: Dashboard Launch + Live Data Pipeline

Objective

Deploy the Trading War Room dashboard as a live React app at trading.boas.dev, wire live data pipelines (Morning Brief Aggregator, Alpaca credentials to VPS), and evaluate TradingView MCP for agent integration. Cross-room coordination with Atlas (Hades credential extraction) and Nexus (credential status verification).

Deliverables

Session Output

Deliverable	Status	Reference
React dashboard (7 pages + 7 demo JSONs)	Live	Task 004
CF Pages project + trading.boas.dev DNS	Active	Task 003
CF Tunnel route api.trading.boas.dev	Routed	Task 003
VPS credential deployment (5 env vars)	Applied	Task 006
Morning Brief Aggregator workflow	Live	HKZ72BpQEhKfKXiI
Deployment spec	Committed	trading-boas-dev-deployment-spec.md
TradingView MCP evaluation	Declined	No Pine scripts

Key Decisions

Architecture Choices

Decision 01

trading.boas.dev as dedicated subdomain

Authority: Plutus

Clean separation from pantheon.boas.dev. Independent CF Pages project, own deploy pipeline. Matches the FastAPI proxy spec.

Decision 02

Single app with demo data fallback

Authority: Plutus

No auth layer. fetchWithFallback pattern tries API/R2 first, falls back to bundled demo JSON. Public visitors see realistic placeholder data. Live data activates when API is available.

Decision 03

TradingView MCP declined

Authority: Plutus

Deep research revealed: ToS violation for non-display usage, fragile CDP dependency (889 open issues), 500MB+ RAM for Desktop app, only one agent can use it. Unique value (Pine indicator extraction) irrelevant without custom Pine scripts. Webhooks remain primary integration.

Decision 04

Morning Brief as n8n workflow

Authority: Plutus → Trading-spoke

n8n handles orchestration, error per branch, and webhook trigger. Already has credential pipeline. 14 nodes, fan-out parallel design matching proven market-data-pipeline pattern.

Decision 05

Cross-room credential chain

Authority: Plutus → Atlas/Hades → Proteus

Hades extracts from vault, writes to staging file. Proteus SSHes to VPS, applies to Docker env. No credential values in LLM context at any point.

Decision 06

VPS FastAPI proxy deferred

Authority: Plutus

Proxy code exists at apps/api/main.py but needs completion (history endpoint, computed fields). Tunnel route is ready — deploy when proxy is finished. Dashboard works with demo data meanwhile.

Team

Agent Contributions

Apollo

Dashboard UI & Design System

Built full React dashboard from HTML mockups — 7 pages, 7 demo JSON fixtures
Design system with CSS custom properties matching mockup tokens exactly
Build clean: tsc + vite, 19.5KB CSS, 678KB JS
TradingView widgets on /market with async loading + skeleton placeholders
Recharts equity curve on /analytics with teal gradient fill

Trading-spoke

n8n Workflow Builder

Built Morning Brief Aggregator workflow — 14 nodes, fan-out parallel
7 watchlist symbols with Alpaca bars + TA calculations (RSI, EMA, SMA, MACD, Bollinger)
FRED macro data (10Y, 2Y, VIX, Fed Funds)
Account snapshot with flag rules for overbought/oversold/trend/volatility
Resolved VPS env var blocker via cross-room Hades chain

Proteus

Infrastructure & Deployment

Created CF Pages project, wired trading.boas.dev CNAME
Added api.trading.boas.dev tunnel route to trading-signals tunnel
Applied 5 env vars to VPS n8n container from Hades staging file
Deployed React dist/ to CF Pages
All three tunnel ingress rules confirmed healthy

Discoveries

What We Learned

Discovery 01

Credentials were deposited 2 days before this session (2026-04-13). The ROADMAP showed them as the critical blocker, but they were already done. Lesson: always verify blocker status before assuming.

Discovery 02

Deploy script referenced by Hades doesn’t exist. scripts/deploy-trading-war-room-env.sh was listed but the scripts/ directory is empty. Credential deploy was done manually via staging file + SSH.

Discovery 03

Proteus can’t deploy trading credentials — they’re wired to the agent-plutus nono profile, not agent-proteus. Required cross-room chain: Hades extracts → file → Proteus applies.

Discovery 04

Wrangler deploy needs CF API token — not available in war room sessions (launched via cc-warroom, not cc-nono). Proteus handles all CF deploys since they have the token.

Discovery 05

TradingView MCP landscape is fragmented. The Jackson fork (CDP to Desktop) is the best option but still ToS-violating. The ertugrul59 fork (Selenium + stolen cookies) is worse. None are production-grade for a systematic trading desk.

Unfinished

Carrying Items

FastAPI proxy completion — Add /api/history endpoint, computed fields on /positions and /account, CORS update, env var alignment. Then deploy to VPS port 8100.

Owner: Plutus

R2 pipeline activation — Market Data Pipeline and Trade Log Pipeline have R2 writes stubbed. Need R2 credentials wired to n8n.

Owner: Trading-spoke

Fleet skills install on 4 agent repos — start, end, chronicle, handoff, state, ground, skill-plan.

Owner: Atlas / Plutus

ntfy endpoint fix — ECONNREFUSED from VPS. ntfy runs on K20N0S localhost:2586. Needs public URL or VPS-side deployment.

Owner: Proteus

Wire Morning Brief to dashboard — Workflow returns live JSON but dashboard still shows demo data. Need R2 write from Morning Brief + dashboard R2 read.

Owner: Trading-spoke + Plutus

ROADMAP.md update — Last updated Session 4. Session 5 and 6 completions not reflected.

Owner: Plutus

Next Session

Resume Instructions

Complete FastAPI proxy (apps/api/main.py) — add /history, computed fields, CORS
Deploy FastAPI to VPS (Proteus) — systemd service on port 8100
Wire R2 credentials for pipeline writes
Connect Morning Brief output to dashboard (R2 write → dashboard R2 read)
Update ROADMAP.md with Session 5+6 completions

DO NOT RE-INVESTIGATE: TradingView MCP (declined) • Credential availability (all live) • CF Pages setup (done) • Tunnel routing (done)

Dashboard Launch +Live Data Pipeline

Dashboard Launch +
Live Data Pipeline