Dev tools with Cloudflare Clef

Show HN CLI that fetches a URL and uses a decision model to extract what matters instead of dumping HTML.

A vLLM-based Jev-compatible server for Clef-Flash claiming 45 ms p50 and double throughput while matching the reference outputs.

A from-scratch C and Metal engine that runs Clef and Clef-Flash on Apple Silicon with a Jev-compatible API, for fast local decisions on a Mac.

PowerShell module that calls Clef on Workers AI for typed, probability-scored decisions straight from scripts and pipelines.

Clef and Clef-Flash packaged for Ollama, making one-command local runs of the decision models possible.

A server that serves Cloudflare Clef-Flash through the Jev/System One style API so existing clients can use it.

The official open Clef weights, a 27B multimodal decision model under Apache 2.0, shared on HN so people can run it locally.

Recipes to serve Cloudflare's unquantized BF16 Clef 27B decision model with vLLM on an NVIDIA DGX Spark.

Unofficial agent skill covering typed decisions, calibration and provider differences across Jev, Clef and others.

Snap package for running Clef Flash inference locally on Ubuntu, a sign of fast distribution-level pickup.

NVFP4 quantised Clef Flash tuned for the DGX Spark box, tied to the community Balatro-playing demo setup.

Runs decision models, including Cloudflare's clef-flash, locally with MLX, with a table of speed and memory for each supported model.

Worker that proxies the System One API to Clef and Clef-flash, so existing Jev clients can switch with a URL change.

Local server for the 9B multimodal Clef-Flash giving one forward pass per request.

Docker deployment of a resident NVFP4 Clef Flash server exposing a structured decision API.

Compiled FP8 server with micro-batching and Ollama pass-through for running Clef-Flash on modest hardware.

Playground with typed decisions and a SystemOne-compatible API running Clef natively on a Mac via MLX.

Skill with classic datasets and live browser benchmarks for running Clef locally.

Electron app to build requests with forms or JSON and explore structured responses from Jev, System One and local Clef via Ollama.

Portal that serves Clef from a single GPU and visualises how rewording a question flips the answer and probabilities.

Runs Cloudflare's Clef Flash locally in Rust with Candle, pinned to a specific checkpoint and implementing the hybrid Qwen3.5 backbone.

Converts and runs Cloudflare's Clef decision model with Apple's MLX so it works locally on a Mac without a GPU server.

VLLM /v1/systemone server and FP8 quantisation recipe that keeps Clef-Flash's Decision Index score on a consumer card.

Edit state, images and questions and see probabilities, latency and cost for local Clef runs.