JevGPT
Build a chatbot by repeatedly asking Jev to choose the next word.
What Jev decides
Which dictionary word should follow, or whether the reply is complete.
JevGPT turns a non-generative decision model into an experimental chat client using vocabulary choices and a stop judgment.
- Jev role
- Pick the next word and judge whether a reply should stop.
- Input
- Conversation transcript, current partial reply, and bounded vocabulary candidates.
- Output
- Word-choice probabilities and a completion probability.
- Decision frequency
- Repeatedly for each word in a generated reply.
- Fallback
- A maximum-word limit stops overly long replies.
How it works
Overview
The conversation transcript becomes Jev state, and a large dictionary supplies the next-word candidates. The client batches smaller vocabulary choices, runs finalists through another Choice, and uses Noul to ask whether the reply is complete. This is an autoregressive experiment rather than a native text-generation capability of Jev.
- Language
- Python
- License
- MIT
- Author
- Bewinxed
Sources
Related projects
typesafe-computer-use
A native macOS computer-use loop built from local OCR, accessibility data, and bounded Jev actions.
What Jev decides
Which computer action, visible item, site, or off-screen accessibility control to use next.
Jev Chat Assistant
An Android conversation copilot that uses Jev to interpret chats and rank reply candidates without sending for the user.
What Jev decides
The likely intent, conversation risk, reply need, best next action, unresolved tension, and which of three candidate replies fits best.
Jev Autopilot
A Jev autopilot flies a simulated drone across a generated Three.js city and lands on a target pad.
What Jev decides
Throttle, yaw, pitch, roll, whether to commit to landing, and whether to cut the motors.