This document outlines the tasks for Developer A, focusing on the backend infrastructure, agent development, and core logic.
- Initialize repository.
- Set up Python virtual environment (
venv). - Install dependencies (
pip install uagents). - Create Agentverse account and generate API keys.
- Set up Supabase project (Database & Auth).
- Set up Redis instance (Upstash or local).
- Configure Google Calendar + Gmail OAuth credentials.
- Implement
Context Sentinelagent usinguAgents. - Integrate Google Calendar API to poll for changes.
- Implement
ContextChangeEventemission logic. - Register agent on Agentverse with
mailbox=True. - Implement Chat Protocol for ASI:One discovery.
- Test: Verify ASI:One discoverability.
- Implement
Disruption Detectoragent.- Receive
ContextChangeEvent. - Classify severity using rules + LLM reasoning.
- Emit
DisruptionEvent.
- Receive
- Register
Disruption Detectoron Agentverse. - Implement
Scheduler Kernelagent foundation.- Receive
DisruptionEvent. - Implement placeholder scheduling (simple swap to buffer).
- Receive
[CHECKPOINT 1 - Phase 1 Complete]
- Test Integration: Verify the flow: Calendar change → Sentinel → Detector → Kernel.
- Sync: Merge all backend agent code to
main. - Verify: Confirm 3 agents are live and discoverable on Agentverse.
- Implement Task Buffer: Redis hash table with composite keys.
- Implement LTS (Long-Term Scheduler): Daily planning pull logic.
- Implement MTS (Medium-Term Scheduler): Swap-in/swap-out logic triggered by
DisruptionEvent. - Implement priority queue data structures using
heapq.
- Implement STS (Short-Term Scheduler): MLFQ priority queues (P0–P3).
- Implement
Energy Monitoragent.- Time-of-day heuristics + task velocity tracking.
- Emit
EnergyLevel.
- Register
Energy Monitoron Agentverse. - Wire
energy_levelinto STS scheduling constraints.
- Implement
Profiler Agent.- Track task completions.
- Compute
peak_hours,estimation_bias,automation_comfort.
- Register
Profiler Agenton Agentverse. - Full Pipeline Integration:
- Sentinel → Detector → Kernel (queries Profiler + Energy Monitor) → GhostWorker → Kernel update.
[CHECKPOINT 2 - Phase 2 Complete]
- Test Integration: Verify full pipeline from disruption to rescheduling with real data.
- Sync: Merge backend intelligence updates to
main. - Verify: All 6 agents registered on Agentverse. Profiler providing real data to Kernel.
- Implement Fetch.ai Payment Protocol on
GhostWorker.- Implement buyer/seller flow.
- Test with demo wallet.
- Create detailed
README.mdfiles for all 6 agents (keyword-rich for ASI:One discoverability). - Test: Verify all agents respond correctly via ASI:One Chat.
- Handle rapid successive disruptions.
- Handle concurrent swaps.
- Handle energy model boundaries.
- Handle sparse data scenarios for Profiler.
- (Stretch) Implement Gmail API integration for
Context Sentinel(new email detection).
- Support automated demo flow scripting (set up scenario data).
- Fix backend bugs found during rehearsal.
- Assist with recording backup demo video (agent logs/terminal).
[CHECKPOINT 3 - Phase 3 Complete]
- Verification: Payment Protocol works (micro-transactions confirmed). ASI:One discovery works.
- Sync: Code freeze on major features. Only bug fixes allowed.
- (Hours 30-32) Final backend bug fixes. Ensure high availability of agents.
- (Hours 32-34) Generate architecture diagrams and agent address tables for documentation.
- (Hours 34-36) Final demo support. Submit to prize tracks.