llmbinge
This is a simple web-app to explore ideas based on LLM, with very little typing. The basic idea is to use LLM to generate a response and also a set of suggestions for each response. We can click on those suggestions to explore the topic further.
Currently ollama is the only supported backend.
Features
- Generation of response from queries (basic)
- Generation of related queries and suggestions
- Generation based on text selection
- History management
- Parent-child relation management for topics
- Handling a set of aspects for every response (eg: people, places, etc)
Development
Use the following commands to get the app running in dev mode:
https://github.com/charstorm/llmbinge.git
cd llmbinge/llmbinge
yarn install
yarn dev
The app uses the following fixed parameters (for now):
- ollama url = “http://localhost:11434/api/generate“
- model = mistral
TODO
- configuration for model and url
- configuration for output size
- favicon for the app
- support response regeneration
- support save and reload
- markdown rendering
- equation rendering
- graphviz graph rendering
- detailed instructions on setting up the project
- release 1.0
