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?
No comments yet.
