kyle-chalmers/typesafe-jev-incident-router
User repository on GitHub
Confidence-gated incident routing with TypeSafe Jev
- Forks
- 1
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Demos & Experiments
TypeSafe Jev Incident Router
A runnable incident-routing example that keeps exact facts, model judgment, policy and human review separate: a registry assigns known assets deterministically, and Jev answers choice, score and noul questions only when the input is messy.
- Stars
- 8
- Language
- Python
- Reading time
- 2 min
- Published by
- Community
- Author
- kyle-chalmers
- Last verified
- Oct 2026
Highlights
- Three answer types in one flow: Choice suggests a response team, Score places business impact, Noul flags blocked critical work.
- Named policy constants: Choice confidence under 0.75 or Score under 0.70 enters review, and blocked critical work at 0.70 forces high-priority review.
- The asset registry runs first: when a known asset has an authoritative owner, Jev is not asked to infer ownership at all.
- Review decisions are written idempotently to .local/review_queue.jsonl and deduplicated by incident reference.
- A calibration worksheet reports misroute rate, review rate and mean minutes to the correct responder on labeled synthetic cases.
Watch out
GitHub reports no licence. Every tracked incident is synthetic, the policy thresholds are illustrative rather than tuned, and the routing fixtures are shaped responses that must not be read as fresh model output; a live run needs a TypeSafe key and prints the concrete model version it received.
Questions about this entry
Short answers drawn from the linked project or page.
The README points first at a TypeSafe Jev use-case guide for data engineering, analytics engineering, analytics and data science; the runnable router shows one of those ideas inside an operational workflow.
The synthetic registry maps known assets to response teams, so an authoritative owner is used as a fact. Only when the registry has no match does Jev receive the incident and answer three focused questions.
Named constants in tools/questions.py: Choice confidence below 0.75 or Score confidence below 0.70 enters review, a Noul probability of 0.70 or higher for blocked critical work forces high-priority review, and an impact score of 2.0 or higher sets high priority.
Yes. Fixtures keep the workflow deterministic and do not call TypeSafe: run python -m scripts.route_incident --incident data/incidents/semantic_route.json --fixture automatic, with paths for unknown_owner, critical_risk, weak_impact and api_error.
The low_confidence fixture selects data_platform while returning Choice confidence 0.39, so the router keeps the suggestion but writes a review decision instead of assigning the incident automatically. The README stresses the fixture is a policy example, not a model response.
Given labeled synthetic outcomes, it reports misroute rate over automatic assignments, review rate over all labeled cases, and mean minutes to the correct responder over all labeled cases. The included rows are examples, not accuracy measurements.
Every tracked incident is synthetic, and the README asks teams to confirm vendor approval, retention, residency and redaction before sending real incident text to any external service, and never to put credentials or customer data in fixtures or logs.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
Facebook: Intake Router keeps diagnosis out of the answer space at the front desk.
FacebookClinicnotes
X: OpenRouter Decision Model Rankings
X: herdr + jev, an agent router
X: A group chat bot with decision room routing
More Jev resources like this
A runnable reference implementation of semantic data quality: it catches silent meaning drift that passes every deterministic BigQuery check, treats a semantic quality score as a first-class ops metric, and puts a circuit breaker at the gate.
Jev Demos & Experiments#community#github#python
A standalone proof of concept showing Jev as a structured decision layer for support-ticket routing: fixed-option routing, 1–5 urgency scoring and probabilities for sensitive data, with a deterministic confidence and safety policy deciding auto-route versus human review.
Jev Demos & Experiments#community#github#streamlit
Two Claude Code hooks powered by Jev, TypeSafe's System One model: a PreToolUse risk gate that asks three noul questions per tool call and maps the top probability to allow, ask or deny, plus a tier router that rewrites subagent calls to haiku, sonnet or opus.
Jev Demos & Experiments#community#github#claude-code
Community
spmamidi/jev-claude-code-pocPython
spmamidi/jev-claude-code-pocPythonA free Claude Code skill that puts Jev in front of the expensive model: thirteen commands sort, check, score, gate and verify with Jev, so Claude reads only the unsure slice and writes the output. Needs python3 and an OpenRouter key.
Jev Agents & MCP#community#python#claude-code
Community
charlesdove977/claude-x-jevPython
charlesdove977/claude-x-jevPythonClaude Code hooks that narrow a roster in YAML, ask Jev which handler owns a prompt, and log the band before anything is blocked.
Jev Agents & MCP#community#python#coding-agent
Community
AgriciDaniel/gatekeeperPython
AgriciDaniel/gatekeeperPythonA browser playground for S1-CED, a small decision model you give a piece of text plus typed questions — yes/no, which-of, or a 0-3 score — and it answers with a probability per option and whether it is confident enough to act or a human should check. Runs on CPU; no GPU or account needed.
Jev Demos & Experiments#community#github#python
Community
ai-yatra/s1ced-playgroundPython
ai-yatra/s1ced-playgroundPythonFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.