Create a LLM instance based on the provider

Delete documents from Qdrant

Get a MinIO client instance.

Helper to create a MiniMax LLM instance.

Run database migrations if needed Raises: Exception: If migration fails

Add documents to Qdrant

Step 2: Generate preview chunks

Search for similar documents in Qdrant with score

Create and configure Alembic config Returns: Alembic config object

Add new chunks to the database

Return a retriever interface

Check if database migration is needed Returns: Tuple containing: - bool: Whether migration is needed

Context manager for database connections with timeout Yields: SQLAlchemy connection object

Delete chunks by their IDs

Delete the entire collection

Get IDs of chunks that no longer exist in the current version

Initialize MinIO by creating the bucket if it doesn't exist.

List all chunk hashes for the given file

Process document in background

Run migrations in 'offline' mode. This configures the context with just a URL and not an Engine, though an Engine is acceptable here as w

Run migrations in 'online' mode. In this scenario we need to create an Engine and associate a connection with the context.

基于 content_hash + index 的双指针匹配算法，查找需要新增、更新和删除的 chunks。 主要改进： 1. 对同一 content_hash 的旧、新 chunks，分别按 index 排序，再逐个匹配，避免原先直接根据 “两两相同位置”导

Initialize Chroma vector store