User repository on GitHub
Use LLMs to find the needles in your haystack
- Forks
- 24
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Tools & CLIs
Siftrank
Go CLI that ranks large text collections by having a model compare small randomized batches across repeated trials until the order converges; with Jev it combines returned pairwise probabilities.
- Stars
- 225
- Language
- Go
- License
- MIT
- Published by
- Community
- Author
- noperator
- Last verified
- Oct 2026
About this project
Use LLMs to find the needles in your haystack. 19 Sep '26: SiftRank now supports Jev, alongside OpenAI and compatible Chat Completions APIs. Use --provider jev to get started; see Jev below. Using an agent? Install the SiftRank agent skill to help it retrieve the most relevant items from large datasets on the fly.
Quoted from the project's README. Read the full README.
Used inJev for Software & DevTools
- Languages
- Go
Highlights
- With `--provider jev`, SiftRank has Jev compare item pairs in each batch, turns the probabilities into an ordering, and runs its convergence algorithm.
- Repeated trials over randomized batches replace one large request, fixing nondeterminism, context limits, truncated output, and incomparable scores.
- Results carry `rank` and a `score` that the README explicitly calls an internal ranking score, not a probability or a calibrated confidence value.
- For JSON input a Go template controls what the model sees, while every result keeps the complete original object, metadata included, under `document`.
Watch out
Go binary with no license stated in the README section reviewed. Jev cannot be combined with reasoning effort or the `--relevance` justification flag, so you clear effort with an empty value and lose that pass. The output `score` is a ranking artifact rather than a calibrated probability and should not be thresholded as confidence. 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.
No. The README states it is an internal ranking score, not a probability or a calibrated confidence value. Jev's returned probabilities are combined into an ordering, but what lands in the JSON is the ranking score, so treat the positions rather than the numbers as the result.
Jev does not support reasoning effort or the `--relevance` justification pass, so you clear effort with `--effort ''` and cannot use that flag. The default model for the provider is `jev-latest`, and the key comes from `TYPESAFE_API_KEY` in the environment.
Defaults are 10 items per batch and up to 50 concurrent calls, with at most 50 trials and no convergence check before 5 trials or until 5 stable trials. Because every trial re-scores in randomized batches, cost scales with collection size, batch size, and how quickly the order converges; `--dry-run` logs the calls without making them.
No. `go install github.com/noperator/siftrank/cmd/siftrank@latest` fetches only the binary. The skill is a portable `SKILL.md` bundle under `.agents/skills/siftrank` that you copy into your agent's skills directory yourself, and it teaches the agent how to decompose a collection and inspect top-ranked hits in their original context.
go install installs only the CLI. For the skill, clone the repo and copy .agents/skills/siftrank into your agent's skills directory, for example ~/.agents/skills/, then ask the agent to use it.
Within each small randomized batch it has Jev compare item pairs and return probabilities, then combines those relative orderings across repeated trials and checks for convergence before progressively focusing on the most relevant items.
Batch size defaults to 10 items and concurrency to 50 concurrent calls, with at least 5 and at most 50 trials, 5 stable trials for convergence, and a 128,000-token cap per batch. A refinement ratio of 0.5 keeps the top half each round.
More Jev resources like this
197GitHub stars
A Chinese-first desktop workbench for local models whose JEV page runs typed noul, choice and score questions against a local engine or a cloud key, alongside a gateway, RAG knowledge base and shared memory.
Jev Tools & CLIs#community#jevbest
Community
kunpengtalk/OmniStudioTypeScript
kunpengtalk/OmniStudioTypeScript56GitHub stars
R.A.I.N. Lab is a local-first research environment. Four constrained perspectives investigate a question, quoted spans are checked against a local paper corpus, and experiments are pre-registered with host code, not a model, deciding pass, fail or inconclusive.
Jev Tools & CLIs#community#jevbest
Community
topherchris420/james_libraryRust
topherchris420/james_libraryRust56GitHub stars
grep, sort, cut and uniq that match by meaning: each line whose Jev Noul clears the threshold survives grev, sibling filters route by Choice and rank by Score, and the output is always your own input, never generated text.
Jev Tools & CLIs#community#cli#coreutils
Community
aurorainfra/grevGo
aurorainfra/grevGo27GitHub stars
Pi extension that replaces many small tool calls with one typed TypeScript execute_tools call, type-checked before running in a sandbox.
Jev Tools & CLIs#community#jevbest#catalog-import
Community
boozedog/pi-codemodeTypeScript
boozedog/pi-codemodeTypeScript30GitHub 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-RadarPython17GitHub stars
Pi extension that trims wasted context before each model call by deduplicating repeated tool results, stripping failed payload args, and letting the model summarize closed spans.
Jev Tools & CLIs#community#jevbest#catalog-import
Community
Davidcreador/pi-dcpTypeScript
Davidcreador/pi-dcpTypeScriptFrom 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.