Organization repository on GitHub
The same decisions, on CPU. System One model that can run on your Personal Computer.
- Forks
- 3
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Tools & CLIs
Dohnuts.cpp
Native C++ inference for Dohnuts on llama.cpp that runs the Qwen3.5-based decision model on CPU, scoring candidate markers with a scalar head in one forward pass.
- Stars
- 11
- Language
- C++
- License
- APACHE-2.0
- Published by
- Community
- Author
- DreamBlooms
- Last verified
- Oct 2026
About this project
Native C++ inference for Dohnuts, built on llama.cpp. It runs the released Qwen3.5-0.8B base with the merged Dohnuts LoRA, scores candidate markers with the scalar decision head, and returns probabilities from a single forward pass. No GPU, no generation.
Quoted from the project's README. Read the full README.
Works forJev for Classification
Used inJev for Software & DevTools
Highlights
- Serves /v1/systemone for multi-question decisions plus /predict, /health, and /v1/models with optional API key and CORS controls.
- Default CPU-only build with optional CUDA, Vulkan, ROCm, and Metal backends plus MinGW cross-compile for Windows.
- One image per request supported via separate vision file, resized and encoded once then shared across questions.
- Shared state prefixes decode once across candidates and persist in a bounded LRU shared by side-model profiles.
Watch out
CPU prefill is slow and quantization shifts probabilities, so use higher-precision files when exact numbers matter. Copy reviewed against the repository README on 2026-10-04, not from running it.
Questions about this entry
Short answers drawn from the linked project or page.
This is not a chat server. It returns choice, score, and noul probabilities for your questions about a state, with no token generation. Prefer llama-server only for model families that server already supports.
Download one GGUF quantization plus head.f32 and the calibration JSON, or merge the adapter into the base and convert with the provided export and build scripts. Quantized files trade probability fidelity for size.
Pass the optional mmproj file at startup and put one image in state.image as a base64 data URL. The image encodes once per request and is shared across questions; without mmproj, image requests are rejected.
Yes through profiles for decider, thisthat, kev, and tev1 families with their own prompt and readout, while the Dohnuts path stays untouched. Those side models take no images.
Requires CMake 3.14+ and a C++20 compiler; on Ubuntu or Debian run sudo scripts/setup.sh then scripts/build.sh, or manually cmake -B build -DCMAKE_BUILD_TYPE=Release -DGGML_NATIVE=ON and cmake --build build -j --target dohnuts-cli.
Use f16, Q8_0, or Q6_K when probabilities matter, since deviations from the f32 reference grow with quantization; Q4_K_M is for coarse decisions.
The default build is CPU-only, but CUDA, Vulkan, ROCm (HIP), and Metal are optional backends enabled at configure time, then offloaded with --gpu-layers and selected with --device.
More Jev resources like this
38GitHub stars
Python memory filter that judges every stored memory against the current request with one calibrated yes/no per memory in a single Jev request, keeping everything above a threshold.
Jev Tools & CLIs#community#jevbest#catalog-import
Community
samdotmak/jev-recallTypeScript
samdotmak/jev-recallTypeScript46GitHub stars
Browser app for systematic-review extraction that opens a trial report with supplements and returns verbatim quotes with file and place, stored locally as projects holding studies.
Jev Tools & CLIs#community#jevbest#catalog-import
Community
choxos/jev-reviewerJavaScript
choxos/jev-reviewerJavaScript34GitHub stars
Pure-HTTP automation for the TypeSafe register, magic-link login, onboarding, and API-key creation flow, with pluggable temp-mail backends, concurrent batch runs, and key verification exports.
Jev Tools & CLIs#community#awesomejev#catalog-import
Community
2951461586/Jev-Register-ToolPython
2951461586/Jev-Register-ToolPython36GitHub stars
Python toolkit plus 0.8B multimodal model that takes a message, document, or image and returns calibrated choice, yes/no, and score answers in one forward pass for local decision workloads.
Jev Tools & CLIs#community#awesomejev#catalog-import
Community
PsiACE/dohnutsPython
PsiACE/dohnutsPython39GitHub stars
Video tool where Jev scores subtitle segments and code assembles a short highlights cut plus a de-watered full cut with timestamped summaries.
Jev Tools & CLIs#community#jevbest#catalog-import
Community
cclank/jevclipPython
cclank/jevclipPython30GitHub stars
Forkable weekday radar that asks Jev one yes-or-no question per interest over each new arXiv announcement and publishes a page plus RSS feed.
Jev Tools & CLIs#community#jevbest#catalog-import
Community
Eliot5566/JEV-Paper-RadarPython
Eliot5566/JEV-Paper-RadarPythonFrom the Jev guides
Original write-ups on what this entry is for.
The request shape barely changes across surfaces; the surrounding contract does. When to call the API directly and when a gateway, a Worker binding, or a router entry earns its extra hop.
Jev Tools & CLIs#gateway#vercel#cloudflare
Read article
Taking a decision checkpoint out of Python: what the GGUF and browser WebGPU exports preserve, what they change, and the parity evidence the model cards publish.
Jev Tools & CLIs#open-models#local#onnx
Read article
Jev labels a record that already exists. Code moves, flags, or holds it. How the document, mail, and transfer tools keep the action out of the model.
Jev Tools & CLIs#classification#email#security
Read article
From the Jev community
Posts from builders shipping with Jev right now.