jev-apps
Two small Python apps: Jev triages failed pipeline tasks into review, retry, page or ticket, and routes support messages to agents.
Overview
This repository contains two small Python apps built on Jev. The pipeline failure triage app makes one Jev call per failed data-pipeline task to answer two questions together: the root cause as a Choice and the impact on downstream data as a Score from 0 to 2. Code turns the answers into one action: human review if the cause is unknown or confidence is below 0.6, a cause-specific retry if the cause can clear by itself (network, late upstream data, resources) and attempts remain, paging on-call if expected impact is at least 1.5, or otherwise a ticket for the owning team.
Failures are JSON objects with pipeline, task, attempt and error log plus any context fields. Samples include eight hand-written failures and ten real dbt failures, and actions are a dry run meant to be replaced with scheduler, ticketing and paging APIs such as Airflow, Jira and PagerDuty. The second app routes customer messages to specialist agents. A blog post describes triaging pipeline failures in about 100 ms with Jev.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | What is the root cause of this pipeline failure? | data_quality · network · late upstream data · resources · unknown · ... |
| Score | What is the impact on downstream data? | 0 · 1 · 2 |
| Choice | Which specialist agent should handle this customer message?(inferred) | — |
How it works
- Receive a failed task as JSON
- One Jev call returns cause and impact
- Code maps answers to review, retry, page or ticket
- Dispatch (dry run by default)
Strengths
- Clear, editable thresholds and owners at the top of the script
- Tested on real dbt failures
- Human review when cause is unknown or uncertain
Caveats
- Dispatch is a dry run; integrations must be added
- Small sample set
Getting started
`export TYPESAFE_API_KEY=...` then `uv run triage.py samples/*.json samples/dbt/*.json`.At a glance
- Type
- Project
- Status
- prototype
- Language
- Python
- Runtime
- CLI scripts
- Has code
- yes
Use cases and integrations
How curated lists describe it
Pipeline failure triage and a customer support router built on TypeSafe AI.
wh000wh000/awesome-jev-live
Frequently asked questions
What does Jev decide for pipeline failures?
The root cause and the downstream impact, which code turns into review, retry, page or ticket.
When does a human review?
When the cause is unknown or Jev's confidence in it is below 0.6.
Does it page anyone for real?
No, dispatch is a dry run until you wire in your own APIs.
Related projects
Data-incident router and use-case guide: registry facts first, then Jev Choice, Score and Noul with confidence-gated human review.
ProjectprototypePythonR&D repo with three Jev pipelines (ticket triage, alert triage, deploy-risk gate) built from four reusable patterns.
ProjectresearchPythonClaude Code skill that hands bulk 'read a lot to decide a little' judgments over files, logs and tickets to Jev and returns a shortlist.
Agent skillbetaJavaScriptTool that lets code and agents ask Jev typed questions via reusable jevels, mapping answers to act, mark or fall_back.
ProjectprototypeTypeScriptWeb app that sends identical ticket-routing and alert-triage questions to OpenAI Decisions and Jev side by side.
BenchmarkprototypePythonSelf-hosted Stacks blockchain data service whose Slack gate and fault-triage paths reportedly run on Jev decisions.
ServiceproductionTypeScript