Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/suvansh/CLAIRe
CLAIRe: Conversational Learning AI with Recall
https://github.com/suvansh/CLAIRe
Last synced: about 2 months ago
JSON representation
CLAIRe: Conversational Learning AI with Recall
- Host: GitHub
- URL: https://github.com/suvansh/CLAIRe
- Owner: suvansh
- Created: 2023-07-17T20:43:38.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2023-08-08T19:34:55.000Z (about 1 year ago)
- Last Synced: 2024-07-11T04:34:38.303Z (2 months ago)
- Language: TypeScript
- Size: 4.04 MB
- Stars: 66
- Watchers: 6
- Forks: 6
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
- awesome - suvansh/CLAIRe - CLAIRe: Conversational Learning AI with Recall (TypeScript)
README
# CLAIRe
## Log
- 7/31 (eba9ffb): Important bug fix that everyone will need – the `.claire` directory was not being created before being used, resulting in errors for all message and profile operations.
## Abilities
### Active messaging
CLAIRe can proactively reach out to you to follow up on past conversations, challenging the conventional call-response paradigm of AI interactions:
### Entity recall
CLAIRe remembers information it has learned about entities in its conversations and can synthesize information from the current conversational context, known information about entities, and semantically relevant past messages to form cohesive, natural responses.
### Message search/history
CLAIRe autoloads message history as you scroll through your messages. It also supports both exact and semantic (through Chroma) message search. The syntax for semantic message search is as follows:
Format: `/semantic {number of results – default 10} search term`
Example: `/semantic adventure story`
Example: `/semantic {2} bad dates`
### Profiles
CLAIRe supports separate profiles backed by separate Chroma collections to prevent cross-over of information.
## Architecture
Learn more about how CLAIRe works [here](https://brilliantly.ai/blog/claire/).
The below diagram illustrates the relationship between some of the main components of CLAIRe's backend.
## Getting Started
### Using CLAIRe as is
To run CLAIRe, you'll need to have [Docker](https://www.docker.com/get-started/).
```bash
git clone https://github.com/suvansh/CLAIRe.git # clone this repo
git clone https://github.com/chroma-core/chroma.git # clone Chroma, used for the DB
cd CLAIRe
cp .env.example .env.local
```
Now edit `.env.local` and fill in your OpenAI API key, used for chat responses as well as the embeddings backing Chroma's semantic search.
Staying in the `CLAIRe` directory, you can launch CLAIRe and Chroma in a shared network:
```
docker-compose up -d --build
```
Note that CLAIRe's [Docker Compose file](docker-compose.yml) relies on Chroma being in the same directory as CLAIRe, but you can edit the path to suit your setup.
Open [http://localhost:3000](http://localhost:3000) in your browser to use CLAIRe!
### Debugging
To debug issues with CLAIRe, the first stop is the Docker logs: `docker logs claire-claire-1` for the CLAIRe container, and `docker logs claire-server-1` for the Chroma container. The frontend will display a catch-all error message in red text if there's an issue with the backend, but the logs will be the best way to figure out what's going on.
### Development
First, run Chroma
Next, the development server in the project root directory:
```bash
npm run dev
```
Open [http://localhost:3000](http://localhost:3000) in your browser to use CLAIRe!