# mmastrac/djev

> A server wrapper over vLLM for structured reads on the DiffusionGemma diffusion model: given a question schema it returns answer distributions (yes/no, choice, score, spans) and implements the Jev decision API.

- Magnitude: 1.7 out of 10 — Early signal
- Stars: 77 total · +0 stars today, ≈ 7 by evening
- Star trust: star growth looks organic
- Category: Language models · Language: Python · License: Apache-2.0 · Created: 2026-09-21 · Last push: 2026-09-24
- GitHub: https://github.com/mmastrac/djev · Page: https://gitnova.dev/en/r/mmastrac/djev

## Useful for

- Deploy a service to extract answer probabilities from ticket text against a given question schema
- Get span answers as exact substrings of the source text without the model generating new values
- Front the port with an OpenAI-compatible router and call /v1/systemone by model name

## Why it’s here

- 0 stars so far today, about 7 expected by the end of the day.
- The spike has held for 3 days in a row — not a one-off blip.
- The repository is 4 days old and already has 77 stars. With less than two weeks of history, there's no usual pace to compare the spike against yet.

## Star trust

Star growth looks organic.

Star-trust labels are heuristics based on the repository’s behavior, not a check of every stargazer.

## Numbers

- Forks: 8
- Issues and pull requests: 11
- Watchers: 0
- Average over the last week: 17 per day
- Usual pace: too little history (under two weeks)
- Stars in the last hour (measured): 0

## Stars per day, last 6 days (oldest → newest, today is partial)

2026-09-20 … 2026-09-25: 0, 20, 17, 31, 9, 0

## Similar by description

1. **Davipar/djev-dev** — 0.9 · Cooling · Language models · Python · +0 stars today, ≈ 2 by evening
   An open implementation of small typed decisions (yes/no, choice, score) on DiffusionGemma and vLLM, with native image and camera input. It reads allowed-label probabilities instead of generating prose or JSON.
   Full card: https://gitnova.dev/en/r/Davipar/djev-dev.md
2. **razorback16/openjev** — 2.6 · Cooling · Language models · Python · +0 stars today, ≈ 16 by evening
   An open-source "System One" decision server on DiffusionGemma: send a state and typed questions (yes/no, choice, score) and get probabilities plus confidence in tens of milliseconds, with no text generated. Wire-compatible with TypeSafe's…
   Full card: https://gitnova.dev/en/r/razorback16/openjev.md
3. **mmastrac/djev-spark** — 1.3 · Cooling · Language models · HTML · +0 stars today, ≈ 6 by evening
   Container recipe for running DiffusionGemma 26B-A4B (NVFP4) on a DGX Spark with vLLM and a structured-decision server exposing Jev's /v1/systemone API.
   Full card: https://gitnova.dev/en/r/mmastrac/djev-spark.md
4. **Rizzo-AI-Academy/rizzo-flow** — 2.8 · Cooling · Language models · Python · +0 stars today, ≈ 22 by evening
   A local-first implementation of the Jev idea: an LLM returns typed decisions (boolean, choice, score, numeric) with probabilities in a single forward pass, without generating tokens. Compatible with the TypeSafe HTTP API.
   Full card: https://gitnova.dev/en/r/Rizzo-AI-Academy/rizzo-flow.md
5. **itsmostafa/typesafe-mcp** — 1.6 · Cooling · AI agents · Go · +0 stars today, ≈ 10 by evening
   A Go MCP server that gives AI agents access to TypeSafe's Jev model: instead of parsing prose, the agent gets typed answers with probabilities it can branch on. One-command setup for Claude Code, Claude Desktop, Codex, and pi.
   Full card: https://gitnova.dev/en/r/itsmostafa/typesafe-mcp.md

---

Magnitude (0–10) measures how fast and how unusually interest in a repository is growing right now. It is not a quality score.
Days are UTC. “So far today” is a fact; “expected by the end of the day” is a forecast.
Summaries and use cases are written by an LLM (DeepSeek V4.1 Flash) from the README and may be inaccurate: verify specific claims (benchmarks, speed, hardware) in the repository itself.
Data as of 2026-09-25 02:11 UTC, updated every 30 minutes.