build complete market data backend

[zuoang25]

@claude
Read all the documents in the planning directory, then build the complete Market Data backend in the backend/directory. this should include:

1. Massive API interface
2. unified market data interface
3. market data simulator
   include full unit tests

[claude]

Claude finished @zuoang25's task in 1m 24s —— View job

---

Market Data Backend — Already Complete ✅

• Read planning/PLAN.md and planning/MARKET_DATA_SUMMARY.md
• Understand the request (Massive API interface, unified market data interface, simulator, full unit tests)
• Inspect current state of backend/app/market/ and backend/tests/market/
• Manually verify correctness by reading through all source and test files
• Determine if any work remains
• Final summary

Finding

The requested Market Data backend already exists in full on this branch, at backend/app/market/:

Requirement	Implementation
Unified market data interface	interface.py — MarketDataSource ABC (start/stop/add_ticker/remove_ticker/get_tickers)
Market data simulator	simulator.py — GBMSimulator (Geometric Brownian Motion with Cholesky-correlated sector moves + random shock events) and SimulatorDataSource
Massive API interface	massive_client.py — MassiveDataSource, REST polling client for Polygon.io via the massive package
Source selection	factory.py — create_market_data_source() picks Massive vs. simulator based on MASSIVE_API_KEY
Supporting pieces	models.py (PriceUpdate), cache.py (thread-safe PriceCache with version counter), seed_prices.py (seed prices + per-ticker vol/drift + correlation groups), stream.py (SSE router for /api/stream/prices)
Full unit tests	backend/tests/market/ — 6 modules (test_models.py, test_cache.py, test_simulator.py, test_simulator_source.py, test_factory.py, test_massive.py), 73 tests total per planning/MARKET_DATA_SUMMARY.md

I read through every module and every test file line-by-line. The code is well-structured (strategy pattern for the two data sources, single-writer/multi-reader PriceCache, GBM math checked out, correlation/Cholesky logic sound, Massive polling has proper error handling and timestamp conversion, SSE endpoint does version-based change detection and disconnect handling). This matches the architecture described in planning/PLAN.md §6 exactly.

git log shows these files were already committed on this branch (part of 1110df9), and git status is clean — there is nothing new to implement or commit.

Limitation: I was not able to actually execute the test suite (uv run pytest / pip install / python -m venv all required approval that isn't available in this non-interactive session) to confirm the 73 tests pass at runtime. I verified correctness by careful manual code review instead. If you'd like, cd backend && uv sync --extra dev && uv run pytest -v will run the full suite locally.

No code changes were made — the branch already satisfies this issue's requirements, so there's nothing to open a PR for.