Watch tasks move through your broker, review failures with their payloads, rerun safely, and automate the incidents your team has already seen too many times.
Task operations loop
Flower and dashboards are good at telling you that something happened. Kanchi is shaped around what operators do next: inspect the payload, choose a recovery path, leave an audit trail, and automate the repeat case.
Detect
Kanchi keeps failed, orphaned, retrying, and running tasks visible with live broker events, worker heartbeat context, and filters that match how incidents are investigated.
Inspect
Task detail pages preserve args, kwargs, traceback, queue, worker, retry chain, rerun lineage, and progress steps in one place your team can link to.
Recover
Bulk actions open a rerun review where Kanchi checks available payloads, lets you repair inputs, skips unsafe items, and records what happened.
Automate
Use triggers, conditions, Slack or webhook notifications, retries, cooldowns, and circuit breakers to automate recovery without creating retry loops.
Basic auth, OAuth, sessions, host/origin controls, and email allowlists for teams that cannot leave dashboards open.
Manual resolve, unresolve, rerun, skipped, failed, and edited actions are stored so the next person can see what changed.
Product tour
See the surfaces operators use when queues misbehave: live task state, guarded reruns, recorded decisions, progress reporting, and automation controls.
Failed tasks, orphaned tasks, workers, and live queue events stay visible on one screen.
Inspect captured payloads, repair inputs, skip unsafe tasks, and submit a tracked rerun batch.
Every resolve, unresolve, rerun, skip, failure, and edited payload has somewhere to live.
Long-running jobs can report progress percentages, named steps, and current messages.
Build no-code task automations with conditions, cooldowns, execution limits, and circuit breakers.
Keep successful task history lean while preserving failed, retried, and orphaned history for analysis.
Why it feels different
The memorable parts of Kanchi are not the table itself, but the way task state becomes recoverable, explainable, and safe to automate.
Read the changelogKanchi stores task state, retry chains, action history, progress, and registry metadata in SQLite, Postgres, or MySQL.
Reruns can be checked, edited, skipped, and traced. Operators do not have to reconstruct payloads from memory.
Point Kanchi at RabbitMQ or Redis, keep ownership of the database, and add task operations without changing application code.
Documentation
Run it beside your workers
Start with broker visibility. Add progress instrumentation, workflow automation, auth, retention, and Prometheus metrics as the workload gets more serious.
Kanchi