hrishikeshchandi/adaptive-rl-task-orchestrator

How hrishikeshchandi/adaptive-rl-task-orchestrator works

This is a distributed task orchestrator: clients submit resource-requirement and priority data, the API stores tasks, workers register and pull capacity-fitting tasks, and the scheduler scores candidates using duration estimates and Q-learning weights. Workers report outcomes, which feed task history and the background learner’s Q-table updates. Recovery and heartbeat services provide retries and dead-worker handling; the README also documents Prometheus/Grafana monitoring and a web dashboard. Unsampled subsystems are included where documented, but their internal links are limited to README evidence.

Main entry points: task.py Task API · main.py FastAPI App · connection.py Tasks and History · workers.py Worker API

Components 20

API Control Plane

FastAPI App main.py
Task API routers/task.py
Worker API routers/workers.py
Learning API routers/learning.py
Request and Task Models models/model.py

Adaptive Scheduling

Duration Estimator scheduling/estimator.py
Q-Learning Agent scheduling/rl_agent.py
Background Learner scheduling/learner.py

Workers and Recovery

Worker Runtime workers/worker.py
Heartbeat Monitor workers/heartbeat.py
Task Recovery recovery/recovery.py

Persistent State

MongoDB Access db/connection.py
Tasks and History db/connection.py
Worker Records db/connection.py
Q-Table db/connection.py

Monitoring and UI

Prometheus Metrics core/metrics.py
Prometheus Scraper prometheus.yml
Grafana Dashboards grafana/dashboards/dashboard.yml
Web Dashboard frontend/index.html

Other components

Client

Connections 33

Client → Task API: submits tasks README.md
Client → Worker API: registers workers README.md
FastAPI App → Task API: routes requests main.py
FastAPI App → Worker API: routes requests main.py
FastAPI App → Learning API: routes requests main.py
FastAPI App → MongoDB Access: initializes main.py
FastAPI App → Background Learner: starts main.py
Task API → Request and Task Models: validates data routers/task.py
Worker API → Request and Task Models: validates data routers/workers.py
Task API → Tasks and History: reads and writes routers/task.py
Task API → Worker Records: claims capacity routers/task.py
Task API → Duration Estimator: estimates duration routers/task.py
Task API → Q-Learning Agent: scores candidates routers/task.py
Task API → Prometheus Metrics: updates counters routers/task.py
Worker API → Worker Records: registers and updates routers/workers.py
Worker Runtime → Worker API: registers and heartbeats workers/worker.py
Worker Runtime → Task API: pulls and reports workers/worker.py
Heartbeat Monitor → Worker Records: checks worker status workers/heartbeat.py
Task Recovery → Tasks and History: resets timed-out tasks recovery/recovery.py
Learning API → Tasks and History: records outcomes routers/learning.py
Learning API → Duration Estimator: estimates duration routers/learning.py
Learning API → Q-Table: reads Q metrics routers/learning.py
Duration Estimator → Tasks and History: reads history scheduling/estimator.py
Background Learner → Tasks and History: reads training history scheduling/learner.py
Background Learner → Q-Learning Agent: updates policy scheduling/learner.py
Q-Learning Agent → Q-Table: loads and saves scheduling/rl_agent.py
MongoDB Access → Tasks and History: provides collection access db/connection.py
MongoDB Access → Worker Records: provides collection access db/connection.py
MongoDB Access → Q-Table: provides collection access db/connection.py
FastAPI App → Prometheus Metrics: exposes metrics main.py
Prometheus Scraper → FastAPI App: scrapes metrics README.md
Grafana Dashboards → Prometheus Scraper: visualizes metrics README.md
Web Dashboard → FastAPI App: uses dashboard APIs README.md

Drawn by GitDiagram on November 15, 2026 from the repository's file tree, README and source files. Read as Markdown ·