| https://trading.snagra.com It’s an autonomous trading system that does more than signal generation. It runs an end-to-end loop: - market/news scanning
- opportunity discovery
- risk checks (position concentration + stop logic)
- execution through Alpaca
- rationale logging for each decision
- daily/weekly strategy evolution The part I think HN might find interesting: the orchestration page The orchestration page is where you can see the system as a pipeline, not a black box. It shows: - scheduled jobs (pre-market, open, intraday monitor, close, weekly review)
- which stage is running and what each stage does
- escalation flow (scout → deeper orchestrator)
- status + outputs of each run
- links between analysis, action, and logged rationale I built this because most “AI trading bots” show entries/exits, but not the decision process. Core design choices - Adaptive strategy layer (not locked to one style) - Two-stage orchestration: fast scout + deeper decision pass
- Deterministic guardrails for risk/mechanical actions - Public audit trail in activity feed (analysis + rationale + evolution logs) - Suggestion workflow: people can submit ideas; system reviews before adoption Tech stack (current) - OpenClaw agent orchestration
- GPT-5.3 Codex for active model workflows
- Alpaca for broker/data execution
- Python scripts for screening/intel/risk/execution
- FastAPI + React site for visibility What I’d love feedback on
1. Orchestration UX: what would make the pipeline easier to inspect/debug?
2. Failure mode design: what should be surfaced more clearly?
3. Governance: how would you structure safe community suggestions for a system like this? |