Preview. These maps are our reading of public sources and haven't been checked with the projects' maintainers yet.
Jevstiller
A drop-in proxy in front of Jev that learns a small local model (bge-small encoder plus logistic head) from Jev's own answers and answers locally only when it can guarantee agreement with Jev at a set rate (for example 98%).
Release v0.5.3 on 2026-10-22; repo pushed 2026-10-24. Its Show HN (2026-09-29, 67 points) was the top-scoring tool post in the local-model group. (checked 2026-10-25)
In The Jev ecosystem
Where it stands apart
- Can it run entirely on your own device or infrastructure, without the maker's servers? partly (some features need the maker's servers) (19% of 21 others)
- Which models answer its questions? Pick the first that applies. only TypeSafe's Jev (direct or through a reseller) (10% of 21 others)
- What is the main thing it ships? a runtime, server, proxy, or desktop app for running models (24% of 21 others)
- Which question types does it handle? a subset or a different set (10% of 21 others)
- Does it check whether its probabilities can be trusted? Pick the first that applies. it calibrates or sets thresholds against your own labelled examples (19% of 21 others)
- Where do the decisions run by default? no default: you choose a local or hosted backend (14% of 21 others)
- Does it train on Jev's outputs? (TypeSafe's customer agreement forbids using output to distill or build competing models.) yes, it trains on Jev's outputs (0% of 19 others)
- Can you use it without a TypeSafe account? no, a TypeSafe API key is required (5% of 21 others)
What it shares with most of the community
- Is it actively maintained? active (62% of 21 others)
- How is the main thing licensed? permissive open source (MIT/Apache/BSD) (86% of 21 others)
- How is it mainly sustained? individual side project, no revenue (55% of 20 others)
- Do you need an account to use it? no account (86% of 21 others)
- Who is it for? developers (95% of 21 others)
- Which shared standard or format does it mainly build on? TypeSafe's System One API (state + typed questions, /v1/systemone) (81% of 21 others)
- Are its answers typed values rather than generated text? (identity question) yes, typed answers only (67% of 21 others)
Nearest neighbors
- THX-01 (HAL-X AI) 53% of 15 shared answers match
- jevals (Openlayer) 50% of 16 shared answers match
- JevBench (Benchmark Heaven) 50% of 16 shared answers match
Every answer
| Question | Answer | Source |
|---|---|---|
| Is it actively maintained? | active | source 2026-10-25 |
| How is the main thing licensed? | permissive open source (MIT/Apache/BSD) Apache-2.0 | source 2026-10-25 |
| How is it mainly sustained? | individual side project, no revenue Independent individual repo, Apache-2.0, no pricing. | source 2026-10-26 |
| Can it run entirely on your own device or infrastructure, without the maker's servers? | stands apart partly (some features need the maker's servers) Local student model; every request it can't answer goes to hosted Jev. | source 2026-10-25 |
| Do you need an account to use it? | no account Proxy; only forwards your own TypeSafe key (provider key, not an account with the maker). | source 2026-10-26 |
| Who is it for? | developers | source 2026-10-25 |
| Which shared standard or format does it mainly build on? | TypeSafe's System One API (state + typed questions, /v1/systemone) Drop-in proxy in front of Jev. | source 2026-10-26 |
| Which models answer its questions? Pick the first that applies. | stands apart only TypeSafe's Jev (direct or through a reseller) Jev is the teacher and the fallback. | source 2026-10-25 |
| What is the main thing it ships? | stands apart a runtime, server, proxy, or desktop app for running models | source 2026-10-25 |
| Which question types does it handle? | stands apart a subset or a different set Answers choice and noul locally; score and box are forwarded to Jev. | source 2026-10-25 |
| Does it check whether its probabilities can be trusted? Pick the first that applies. | stands apart it calibrates or sets thresholds against your own labelled examples Calibrates a routing threshold with a statistical bound on agreement, using Jev's answers as the labels. | source 2026-10-25 |
| Where do the decisions run by default? | stands apart no default: you choose a local or hosted backend | source 2026-10-25 |
| Does it train on Jev's outputs? (TypeSafe's customer agreement forbids using output to distill or build competing models.) | stands apart yes, it trains on Jev's outputs Its docs warn that TypeSafe's MCA s.2.3(b) restricts this and leave the decision to the user. | source 2026-10-25 |
| Can you use it without a TypeSafe account? | stands apart no, a TypeSafe API key is required | source 2026-10-25 |
| Are its answers typed values rather than generated text? (identity question) | yes, typed answers only | source 2026-10-25 |
| Does it publish head-to-head numbers against Jev? | yes, head-to-head numbers against Jev Measures agreement with Jev. | source 2026-10-25 |
Sources
This page was derived from:
- GitHub · github.com/tomerglick57/Jevstiller · read 2026-10-25
- GitHub · github.com/tomerglick57/Jevstiller/blob/main/docs/coverage.md · read 2026-10-25
- Web page · jevstiller.pages.dev/posts/the-guarantee/ · read 2026-10-25
RSS for this project · Its own release feed
Are you a maintainer? Claiming and correcting projects is coming soon.