Bibliothecarius is a local data analysis tool that can engage in conversation based on custom prompts. The tool supports multiple models, allowing for horizontal comparison, and supports data isolation and multiple data types.
- Personal or enterprise knowledge base question-and-answer assistant
- Analyze books or materials
- AI assistant exclusively for a certain group
- Can even organize AI to play murder mystery games with you
- .....
- Multi-round dialogue
- Better UI
- Support for more file types
- txt
- md
- docx
- Support for more models
- gpt3.5
- ChatGLM
After downloading the code, initialize the environment with docker-compose:
docker-compose up -d- MySQL
- Use an existing database: initialization SQL is available in ./mysql/init/init.sql, simply execute it.
- Qdrant
- Refer to official installation
-
Modify the configuration file
./bibliothecarius/config/application.yaml, including:- MySQL configuration
- opanai key
- Qdrant service address
- Storage configuration
-
Start the backend service with docker-compose
cd bibliothecarius docker-compose up -d -
Verify the backend service by visiting http://127.0.0.1:8080/
-
If you want to embed in your service, you can leave the front-end service unenabled
Front-end project
./web
The Bibliothecarius interface has two main parts:
- dataset: Your local data-related interface, including the operation of the dataset and the interface to add and delete data to the dataset.
- scene: Your model requests parameters and prompt and data set association interfaces.
You can use the following examples to familiarize yourself with Bibliothecarius.
- Analyze books ormaterials
- Personal or enterprise knowledge base question-and-answer assistant