Roadmap
The roadmap is grouped by horizon. Each item is a short title and a one-line “why”.
v0.2 — quality polish (next 1-2 months)
Section titled “v0.2 — quality polish (next 1-2 months)”- GitHub Actions CI — pytest + ruff + mypy on every push, badge in README. Catches regressions before they hit master.
- 100% type-hint coverage — bring the codebase from ~52% to fully typed,
gated by
mypy --strict. Makes contributions safer and editors smarter. - Pre-commit hooks — ruff, black, mypy locally. Same gate as CI, just earlier in the loop.
- Module-size cleanup — split
cli/backtest.py(currently 689 lines) anddb/repository.py(588 lines) into focused modules. The 800-line per-file guideline exists for a reason. - Sentiment agent model bump — move from
claude-3-haiku-20240307to the current Haiku release. Cheaper, faster, better classification. - Sample fixtures +
--use-fixtures— letcents researchrun end-to- end without API keys. The cleanest possible demo improvement. - Graceful agent degradation — orchestrator returns partial results if NewsAPI or FRED is unavailable, instead of failing the whole run.
- Coverage badge — surface test coverage publicly so contributions know the bar.
v1.0 — productization (longer horizon)
Section titled “v1.0 — productization (longer horizon)”- Web API (Cloudflare Worker) —
POST /research/{symbol}returning the sameAgentResultJSON. Same shape, different transport. - Live Cloudflare Worker demo — fixture-backed live demo on this site, same shape as the static export. Removes the “cool, but does it run?” friction.
- Slack/webhook alerts —
CENTS_WEBHOOK_URLis already plumbed through the alert pipeline; needs a payload formatter and rate-limit niceties. - Agent performance profiler —
cents metrics --by-agentshowing accuracy + cost + latency per agent. Forces every agent to earn its weight. - Broker paper-trading dashboard — open positions, P&L by thesis, realized accuracy. Closes the loop between research and outcomes.
Out of scope (deliberate)
Section titled “Out of scope (deliberate)”- Real-time streaming data. Cents is a research and review tool, not a trading terminal.
- Auto-execution from the CLI. The broker integration is paper-only by design; live trading stays opt-in and manual.
- Hosted multi-user product. Local-first, single-user is a feature.
Suggestions and counter-proposals are welcome on GitHub issues.
Not financial advice. Cents is an educational and research
tool for tracking your own investment theses. Outputs are model-generated and
may be inaccurate. You are solely responsible for your own investment
decisions.