1.4 KiB
1.4 KiB
Research Agent Spawning
Overview
System for spawning isolated sub-agents to handle research tasks asynchronously
Status
🔄 In Progress — Designing spawn and report workflows
Goals
- Spawn sub-agents for independent research tasks
- Parallel execution for multiple research queries
- Automatic summarization and reporting back
- Resource management (don't overwhelm system)
Architecture Ideas
Option 1: Cron-Based Research
- Spawn agents via cron jobs
- Results delivered via Discord
- Good for: scheduled research, news monitoring
Option 2: On-Demand Spawning
- User triggers: "/research [topic]"
- Spawns isolated session
- Reports back when complete
- Good for: deep dives, ad-hoc questions
Option 3: Persistent Research Channel
- Dedicated Discord channel
- All research requests go there
- Agent monitors and spawns workers
- Good for: continuous research, collaborative
Current Exploration
Testing in #projects channel (1468257895152881796)
Questions to Answer
- How many concurrent agents?
- How to handle long-running research?
- Where to store results? (Obsidian? SQLite?)
- How to prevent spawn loops?
- Cost tracking per agent?
Related
- Multi-User Agent Architecture — multi-user considerations
- Memory System — where research results might go
Created: 2026-02-25 Next: Test spawning mechanisms