Skip to content

0xprathamesh/agent-execution-engine

Repository files navigation

Agent Execution Engine

Job queue + worker backend. API enqueues jobs; a separate worker process consumes from Redis, runs processors, and syncs state to Postgres. Stuck-job recovery runs inside the worker.

Stack: Node, TypeScript, Express, Prisma (Postgres), BullMQ (Redis).


Run

cp .env.example .env   # set DATABASE_URL, REDIS_* etc.
npm install
npx prisma migrate dev

npm run start    # API (port 3000)
npm run worker   # Worker + stuck-job recovery (separate process)

API: POST /api/v1/job/job — body: { type, payload }. Types: AGENT_TASK, WEBHOOK, GENERIC.
Get job: GET /api/v1/job/job/:id
Docs: http://localhost:3000/api-docs (Swagger UI)


Phases

Phase Scope Status
1 Basic job queue, basic worker, basic API
2 Workflow DAG engine
3 Agent brain
4 Distributed scaling
5 Observability
6 Advanced failure handling

Phase 1 includes: idempotent worker (claim via setRunningIfQueued), BullMQ retry/backoff aligned with DB, stuck RUNNING recovery scheduler, job timeout and worker concurrency via env.

About

Production-grade distributed job orchestration engine powering AI agent workflows with durable execution, retry semantics, and observability-ready architecture

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors