Documentation

Overview

Project KirBe is a versatile tool designed for ingesting and enriching data from various sources, including local directories and YouTube. It leverages LlamaIndex readers to extract content and metadata, providing a streamlined approach to knowledge management.

Key Features

• All-formats-in-one embedding
• LlamaIndex integration
• Data ingestion from local directories
• YouTube transcript processing
• Metadata enrichment

API Endpoints

Kirbe provides the following API endpoints for data ingestion:

• /api/v1/simpledir/scan - Scans a local directory and processes supported files.
• /api/v1/yt-transcript/video - Fetches the transcript and metadata for a single YouTube video.
• /api/v1/yt-transcript/playlist - Fetches transcripts for videos within a YouTube playlist.
• /api/v1/yt-transcript/file - Fetches transcripts for YouTube videos listed in a file.

Data Models

The API uses the following data models for requests and responses:

• SimpleDirRequest - Request body for scanning a local directory.
• SimpleDirResponse - Response body for the simple directory scan endpoint.
• TranscriptRequest - Request body for fetching a single YouTube transcript.
• TranscriptResponse - Response for fetching a single transcript.
• PlaylistRequest - Request body for fetching transcripts from a YouTube playlist.
• BatchResponse - Response for fetching transcripts from a playlist or file.

Authentication

This API uses a proprietary authentication method by Holistika. Contact Holistika R&D for more information.

License

Project KirBe is licensed under a Proprietary - Holistika license.