pi-jev
Speed-focused Pi extension combining a local coding model with Jev judgments for file retrieval and cached requests.
Overview
pi-jev (madeye) is a speed-focused Pi extension that pairs a local or cheap coding model (Qwen, DeepSeek and others) with hosted TypeSafe Jev judgments. Jev-assisted file retrieval reduces the context sent to the generator by returning focused excerpts, while exact-result caching and an outage cooldown limit repeated network overhead. The generator still writes code and uses Pi's tools.
The README is explicit that it does not accelerate model inference: it targets end-to-end latency, every uncached Jev call adds latency, and reading small files directly is often cheaper, with positive and negative measurements kept in VALIDATION.md. Without a key, jev_search still works with local retrieval while Jev-based speed, skills and tools options stay disabled. TYPESAFE_BASE_URL can point at a self-hosted Jev-compatible server. It requires Node 22+ and Pi 0.85.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Is this file excerpt relevant to the current task?(inferred) | yes · no |
How it works
- Agent issues a search or needs files.
- Retrieve candidate excerpts locally.
- Ask Jev which are relevant; cache exact results.
- Send focused excerpts to the generator model.
Strengths
- Less context for the generator.
- Caching and cooldown limit network overhead.
- Works partly without a key.
Caveats
- Uncached Jev calls add latency.
- Pinned to Pi 0.85 series.
Getting started
pi install git:github.com/madeye/pi-jev@v0.1.0 and export TYPESAFE_API_KEY.At a glance
- Type
- Project
- Status
- beta
- Language
- TypeScript
- Runtime
- Pi extension
- Has code
- yes
- Has tests
- yes
Use cases and integrations
How curated lists describe it
Jev-assisted file retrieval and request caching for faster Pi workflows.
hellogumbo/awesome-jev
Frequently asked questions
Does it make models faster?
No, it targets end-to-end latency by sending less context.
Can I use a self-hosted Jev-compatible server?
Yes, set TYPESAFE_BASE_URL.
Does it work without a key?
jev_search still works with local retrieval only.
Related projects
PiJev: Pi-based coding agent where Jev shortlists skills, ranks code excerpts and triages tool failures while your model writes code.
ProjectprototypeTypeScriptCLI and MCP server giving coding agents Jev-powered semantic search, focused reads and list picking in about a second for about a cent.
BenchmarkbetaPythonDeterministic code-graph context for coding agents on Postgres, AGE and pgvector, with optional Jev or local re-selection of symbols.
MCP serverbetaPythonClaude Code plugin that adds Jev as an independent judge for review findings, debugging hypotheses, design options and search hits.
PluginbetaRetrieval layer for Codex that finds evidence locally, optionally ranks sanitized excerpts with Jev, and returns bounded, cited evidence.
ProjectbetaPipe grep or file lists into Jev to get every candidate scored, sorted and cut at a calibrated threshold, so agents read only what matters.
ProjectbetaTypeScript