jev-bigquery-cloudrun
Demo that classifies support tickets in BigQuery via a remote function backed by a Cloud Run adapter calling Jev.
Overview
Jev + BigQuery + Cloud Run classifies 30 fictional support tickets by owning team, urgency and whether a workflow is blocked. A private Cloud Run adapter calls TypeSafe's Jev API, and a BigQuery remote function saves the answers so they can be queried with ordinary SQL. It is aimed at data teams who want typed decisions available inside their warehouse.
Each result includes the team, route confidence, urgency and blocking probabilities, the model and a rubric version; closed tickets also get a resolution-support score, with zero by rule for a missing resolution and null for open tickets. The adapter exposes a direct tickets endpoint and a BigQuery-shaped endpoint. Scoring is run once and stored, so example queries for team summaries, urgent or blocked tickets, review queues and resolution checks make no further Jev calls. The README notes that the 0.80 cutoffs in the queries are examples, not validated thresholds, and that scoring incurs inference charges.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which team owns this ticket? | — |
| Yes / no | Is this ticket urgent at submission? | yes · no |
| Yes / no | Is a workflow blocked? | yes · no |
| Score | How well does the resolution address the ticket? | — |
How it works
- Deploy the Cloud Run adapter and register the BigQuery remote function
- Run the scoring SQL once to store answers
- Query the enriched_tickets view with ordinary SQL
- Apply example cutoffs for urgent and blocked queues
Strengths
- Decisions available in plain SQL
- Score once, query many times without new calls
- Results carry model and rubric version
Caveats
- Example thresholds not validated
- Requires a billed Google Cloud project
Getting started
Follow docs/setup.md to deploy, then run `bq query ... < sql/03_score_tickets.sql` and `sql/04_examples.sql`.At a glance
- Type
- Project
- Status
- prototype
- Language
- Python, SQL
- Runtime
- Cloud Run, BigQuery remote function
- Has code
- yes
Use cases and integrations
How curated lists describe it
Classify support tickets in BigQuery with Jev and Cloud Run
logicrw/awesome-jev-projects
Frequently asked questions
What does Jev decide?
Ticket team, urgency, blocked status and resolution support.
Do queries call Jev each time?
No. Scores are stored once and example queries make no further calls.
Are the 0.80 cutoffs recommended?
No, they are examples, not validated thresholds.
Related projects
Ruby gem that stores Jev Noul, Choice and Score answers about a record's text as ordinary ActiveRecord columns you can query.
LibrarybetaRubyOpen-source team agent platform for Slack, GitHub and more that asks Jev once per session which agent runtime and model to use.
ServicebetaTypeScriptGitHub Action that triages issues with Jev, applying label, spam, needs-info and duplicate tags only above set thresholds.
ProjectbetaTypeScriptSelf-hosted Next.js API and GitHub Actions workflow that has Jev pick one label for each new issue or pull request.
ServicebetaTypeScriptProbot GitHub app that classifies, scores, checks completeness and dedupes issues and PRs with Jev, routing uncertain ones to humans.
ProjectbetaTypeScriptOne YAML task to classify, route, score or verify text with rules, Jev, NanoJev or an OpenAI-compatible model, with cost records.
CLIbetaPython