typesafe-poc
Portuguese-language Go proof of concept that tests whether Jev's confidence is calibrated enough to auto-route support tickets.
Overview
This proof of concept, documented in Portuguese, evaluates TypeSafe's typed classification API on one use case: automatic support-ticket routing with calibrated confidence. The question it asks is whether the model's confidence is well calibrated enough that automatic routing decisions rarely need human review.
The pipeline makes three decisions in parallel: ticket urgency (yes/no probability), responsible department (choice among six) and customer frustration (five-level rubric). It includes an evaluation tool with a dry-run mode that simulates API responses, a full evaluation against a small synthetic dataset, and an HTTP server with POST /triage. The README is clear that the metrics validate the pipeline architecture, not production viability, and that fine-tuning and real ticket-queue integration are out of scope.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Is this ticket urgent? | yes · no |
| Choice | Which of six departments should handle this ticket? | — |
| Score | How frustrated is the customer (five levels)? | — |
How it works
- Submit a ticket JSON to POST /triage.
- Jev answers urgency, department and frustration in parallel.
- The server returns a routing decision with confidence scores.
- An evaluation tool measures calibration on a synthetic dataset.
Strengths
- Dry-run mode works without credentials.
- Explicitly measures calibration for auto-routing.
- Clear statement of scope and limits.
Caveats
- Small synthetic dataset.
- Not validated for production.
- Documentation in Portuguese.
Getting started
make eval-dry (no key) or export TYPESAFE_API_KEY and run make eval / make runAt a glance
- Type
- Demo
- Status
- prototype
- Language
- Go
- Runtime
- CLI + HTTP server
- Has code
- yes
Use cases and integrations
How curated lists describe it
Prova de Conceito do Jev, modelo da typesafe.ai.
hellogumbo/awesome-jev
Frequently asked questions
What does the POC test?
Whether Jev's confidence is calibrated enough to auto-route support tickets without frequent review.
Can I run it without a key?
Yes, make eval-dry simulates API responses.
Is it production-ready?
No, the README says the metrics only validate the architecture.
Related projects
GitHub Action that triages issues with Jev, applying label, spam, needs-info and duplicate tags only above set thresholds.
ProjectbetaTypeScriptOne YAML task to classify, route, score or verify text with rules, Jev, NanoJev or an OpenAI-compatible model, with cost records.
CLIbetaPythonMultiplayer GitHub triage dashboard where Jev answers kind, severity, urgency, duplicate and next step per issue.
AppbetaCLI, MCP server and skill for issue triage where Jev predicts issue kind, affected module and whether information is sufficient.
MCP serverprototypePythonPython sandbox turning Jev issue-category and reproduction judgments into testable routing proposals, with offline replay.
ProjectprototypePythonExperimental lab for semantic business gates: each gate is an independent Jev Noul that code maps to NO, REVIEW or YES by threshold.
Projectprototype