How snowby666/forge works
Forge is an autonomous hackathon system: it discovers and scores events, researches competitors, judges and sponsors, plans a project, builds and verifies it, then prepares and submits a demo and pitch. The Commander coordinates the agent workflow and human approval checkpoints; infrastructure agents preserve learning and monitor runs. A separate web dashboard manages runs and exposes live status. The README documents capabilities beyond the sampled excerpts, so unsampled stages are represented at their documented directory paths; uncertain internal call wiring is omitted.
Main entry points: commander.py Commander · routes Dashboard API · server.ts Browser automation · hackathon_scout.py Scout
Components 25
Hackathon workflow
- Commander
agents/python/orchestrator/commander.py - Scout
agents/python/intelligence/hackathon_scout.py - Human checkpoints
agents/python/orchestrator/commander.py - Demo and submission
agents/python/submission/submission_pipeline.py
Specialist agents
- Project intelligence
agents/python/intelligence - Strategy and planning
agents/python/strategy - Parallel project build
config/agents_config.py - Quality verification
agents/python/verify/ux_auditor.py - Project polish
agents/python/polish/polish_agents.py
Learning and operations
- Persistent memory
agents/python/infra/memory_keeper.py - Outcome learning
agents/python/infra/outcome_tracker.py - Knowledge updates
agents/python/infra/knowledge_updater.py - Run monitoring
agents/python/infra/monitor_and_calendar.py - Run state
config/redis_client.py
Web dashboard
- Dashboard API
forge_web/routes - Dashboard services
forge_web/services.py - Dashboard UI
web/src/app/(dashboard)
External services
- Browser automation
agents/browser/server.ts - Hackathon platforms
- ElectronHub LLM
- Submission platforms
- Google Calendar
- Web search
Other components
- Hackathon organizer
- Human participant
Connections 30
- Hackathon organizer → Scout: requests discovery
- Scout → Browser automation: scrapes listings
- Browser automation → Hackathon platforms: reads event data
- Scout → Commander: registers opportunities
- Commander → Project intelligence: dispatches research
- Commander → Strategy and planning: dispatches planning
- Project intelligence → ElectronHub LLM: analyzes evidence
- Strategy and planning → ElectronHub LLM: generates plans
- Commander → Human checkpoints: coordinates approvals
- Human checkpoints → Human participant: requests decisions
- Human participant → Human checkpoints: approves stages
- Commander → Parallel project build: dispatches build
- Commander → Quality verification: dispatches audits
- Commander → Project polish: dispatches polish
- Commander → Demo and submission: dispatches submission
- Quality verification → Browser automation: requests screenshots
- Demo and submission → Browser automation: records demo
- Demo and submission → Submission platforms: submits entry
- Commander → Run state: reads and writes state
- Project intelligence → Persistent memory: learns from prior runs
- Outcome learning → Persistent memory: feeds results
- Knowledge updates → Web search: researches trends
- Knowledge updates → Strategy and planning: updates design knowledge
- Run monitoring → Google Calendar: schedules checkpoints
- Dashboard UI → Dashboard API: calls dashboard API
- Dashboard API → Dashboard services: uses services
- Dashboard services → Run state: reads and writes state
- Dashboard API → Commander: launches runs
- Dashboard API → Human participant: serves approvals
- Commander → Run monitoring: tracks run activity