Frontend design of Knowledgebase management

This file is a Streamlit application for managing a knowledge base, featuring multiple functionalities such as displaying knowledge bases, uploading files, and creating new knowledge bases.

Key components:

1. Basic Setup:

The application imports streamlit for building the user interface, pandas for handling tabular data, AgGrid for displaying documents in the knowledge base, and requests for communicating with the backend API.

UPLOAD_DIR is defined as the directory for file uploads, ensuring its existence through the Path module.

2. Knowledge Base Page Display and Selection:

Initially, it fetches the current list of knowledge bases via an API request (requests.get("http://localhost:8000/get_kb_list")) and stores it in kb_list for user selection.

Users can choose from existing knowledge bases or select "Create New Knowledge Base" to set up a new one.

3. Creating a New Knowledge Base:

If the user selects "Create New Knowledge Base," a form is presented to enter the knowledge base name, description, vector store type (vs_type), and embedding model (embed_model).

Upon submission, a POST request (requests.post) is sent to the backend to create the knowledge base, and the page refreshes upon successful creation.

4. Displaying Knowledge Base Information:

Once a knowledge base is selected, its detailed information is displayed, including the description and a list of existing documents.

The document list is retrieved via an API request (requests.get("http://localhost:8000/get_kb_documents?kb_name={selected_kb}")) and presented in a table format using AgGrid.

5. File Upload and Processing:

Users can upload multiple files using st.file_uploader, along with configuring processing parameters such as maximum segment length, text overlap, and whether to enhance Chinese titles.

After uploading, files and processing configurations are sent to the backend via a POST request (requests.post("http://localhost:8000/upload_files")) to add them to the knowledge base.

6. File Deletion:

By selecting documents in the table, users can perform bulk deletion. The selected files are sent to the backend via a POST request to complete the deletion operation (requests.post("http://localhost:8000/delete_kb_files")).

7. Rebuilding the Vector Store:

Users have the option to rebuild the knowledge base's vector store, which is done by sending an API request (requests.post(f"http://localhost:8000/rebuild_vector_store")) to update or reconstruct the vector store.

8. Deleting the Entire Knowledge Base:

Users can delete the entire knowledge base after confirming the action. The application sends a request to the backend to delete the knowledge base (requests.post("http://localhost:8000/delete_kb")).