Practical walkthrough: Jev for model routing and code-review triage
kristiyanstoyanovAI · reddit · 2026-09-22
A 24-minute walkthrough of Jev, TypeSafe's bounded-decision model, with two open examples: (1) an open-source model router (github.com/krisitown/jev-router) where Jev picks between local Qwen and hosted Sonnet based on request context, with a control UI exposing the actual routing request, response and destination to make the decision/generation boundary inspectable; (2) a Python/Strands code-review workflow where agents inspect public GitHub PRs and propose findings, a judge (LLM vs Jev compared) assesses severity and relevance, with final inclusion rules kept in Python and results output as an HTML report. Covers the noul, choice and score request types with API examples. Framed as an integration tutorial rather than a benchmark, focused on division of responsibility: LLMs investigate and propose, specialized models answer bounded questions, application code decides.
More from coding & agent
- mcp-server-scraper: Free Firecrawl Alternative MCP Server for Web Scraping — modelcontextprotocol · 2026-09-22
- APIHub Launches Pay-Per-Call API Marketplace for AI Agents With USDC — modelcontextprotocol · 2026-09-22
- AI Coding Shifted the Bottleneck to Reviewing: Do You Read the Plan or the Diff? — Live-Performer7535 · 2026-09-22
- Ops agent that learned contract redlines from screen recordings hits $140M ARR in 90 days — thisdudelikesAI · 2026-09-22
- GitHub takes down Bend programming language repo; team appeals for reinstatement — TheMoonMidas · 2026-09-22
- 4B Qwen tuned with SFT+RL produces 81% faster Postgres query plans for $1,200 — JeremyCMorgan · 2026-09-22