…, E2E

Complete the FinAlly AI trading workstation on top of the existing market-data
subsystem. Coordinated via planning/TEAM_CONTRACT.md.

Backend (FastAPI):
- app/config.py, app/runtime.py (shared PriceCache/source accessors)
- app/services/portfolio.py: execute_trade (weighted avg cost, validation,
  snapshot-on-trade) + build_portfolio
- app/api/{health,portfolio,watchlist,chat}.py per the wire contract
- app/main.py lifespan: init_db -> market source -> 30s snapshots -> SSE +
  REST + static frontend serving
- app/llm/: pydantic schema, prompts, Cerebras client (LiteLLM->OpenRouter),
  deterministic LLM_MOCK; chat auto-executes trades/watchlist changes
- 132 tests pass (services/api/llm added), ruff clean

Frontend (Next.js static export):
- Terminal orchestrator, TradeBar, PositionsTable, ChatPanel; Vitest config +
  unit tests
- Fix usePriceStream to parse the batched {ticker: update} SSE payload; add
  change_percent to the PriceUpdate type

DevOps + E2E:
- Fix container DB path (absolute FINALLY_DB_PATH=/app/db/finally.db) and stop
  tmpfs from shadowing schema.sql; full docker build runs healthy
- Playwright suite in test/ — 6 E2E scenarios pass against the running app

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0177Si2U5NgcHtJ57FLSGGuj