GitHub radar
Sort tickets and complaints with a free model instead of a paid API
Clef is an open model that turns a description of a situation into answer probabilities in one pass. You can run it as a free replacement for the paid Jev and SystemOne services, which handle the same sorting job.
A 27-billion-parameter model (a number that roughly shows its size and power) — you give it text, an image, or video plus a list of questions with fixed answer options, and it returns a probability for every option right away, with no text to parse afterward.
When it helps
Useful if your product sorts a stream of tickets, complaints, or similar requests into fixed categories. For tasks needing broad knowledge and real reasoning it falls well behind — the paid Jev wins there.
Pros
- API-compatible with paid Jev — barely any code changes needed
- Apache-2.0 license: free to use and modify, weights are open
- Beats the paid Jev classifier on ticket and complaint sorting
- Returns ready answer probabilities, no text parsing required
Cons
- Falls far behind paid Jev on tasks needing broad reasoning
- Needs a 32GB-memory Mac or a 24GB graphics card
How to set it up — step by step
- 1Open your AI agent (Claude Code, Codex) in your project folder
- 2Send the agent the text from the block below
- 3Check that the agent downloaded the model from Cloudflare's page (https://huggingface.co/Cloudflare/clef) and set it up as a Jev/SystemOne alternative
- 4Give the model a test support ticket with a couple of questions
- 5Look at the probabilities it returns for each answer option
- 6Confirm beforehand your hardware fits: a 32GB-memory Mac or a 24GB graphics card
Text for your agent
Copy this and send it to your agent — Claude Code, Codex, any of them:
Set up Cloudflare's Clef model for me from https://huggingface.co/Cloudflare/clef — run it locally as a self-hosted alternative to the paid Jev/SystemOne API (it's compatible), feed it a test support request with a couple of questions, and show me the probabilities it returns.
You need a Mac with 32 GB of memory or a GPU with 24 GB of VRAM — it will not fit on a regular 16 GB laptop.
Open on Hugging Face▌ More finds