Labels: hard enhancement gssoc-2026
Connect every module together into a single real-time loop: Perception Bus → Processing → Intelligence Core → Trust Scorer → WebSocket broadcast. This is the core runtime that makes Execra live.
What you'll code:
- Create
core/pipeline.py
- Implement
ExecraPipeline class:
__init__(domain, mode) — initialises all subsystems: PerceptionBus, ContextEngine, IntelligenceCore, GuidanceDispatcher, ActionLoggerasync def run() — the main async loop: consumes frames from the perception bus queues, runs processing, calls intelligence core, dispatches guidance via WebSocket- Implement backpressure: if the processing queue is full, drop the oldest frame (never block perception)
- Implement guidance deduplication: do not send the same instruction twice in a row
- Target latency: ≤500ms from frame arrival to WebSocket broadcast
async def stop() — cleanly shuts down all subsystems
- Integrate pipeline startup into
main.py and api/main.py startup event
- Write end-to-end integration tests with mocked perception and LLM
- Benchmark latency and document results
Skills needed: Python · asyncio · system design · performance engineering · integration
Labels:
hardenhancementgssoc-2026Connect every module together into a single real-time loop: Perception Bus → Processing → Intelligence Core → Trust Scorer → WebSocket broadcast. This is the core runtime that makes Execra live.
What you'll code:
core/pipeline.pyExecraPipelineclass:__init__(domain, mode)— initialises all subsystems:PerceptionBus,ContextEngine,IntelligenceCore,GuidanceDispatcher,ActionLoggerasync def run()— the main async loop: consumes frames from the perception bus queues, runs processing, calls intelligence core, dispatches guidance via WebSocketasync def stop()— cleanly shuts down all subsystemsmain.pyandapi/main.pystartup eventSkills needed: Python · asyncio · system design · performance engineering · integration