Release Notes for v0.3.0

Highlights

• Rename project to autoflow
• Multiple Knowledge Bases support
• Support new LLM providers
  • OpenRouter
  • ZhipuAI BigModel
  • Ollama
• Support new embedding models providers
  • Ollama
  • Support OpenAI Like embedding model providers
    • ZhipuAI BigModel

Breaking Changes

• Knowledge Base
  • Deprecate old datasource management. Manual migration is required.
  • Support different Embedding Models for each KB
  • Individual vector index and knowledge graph index for each KB
  • Move vector search and graph search API to knowledge base level
  • Move index progress and error retries to knowledge base

Improvements

• Support create chat engine from default config

Manual Migration

Migrating from v0.2.x to v0.3.0

In previous versions, there was no concept of Knowledge Base, and Autoflow stored index data in the following tables:

• chunks
• entities
• relationships

In the v0.3.0, index data will be stored separately in tables corresponding to each knowledge base:

• chunks_{knowledge_base_id}
• entities_{knowledge_base_id}
• relationships_{knowledge_base_id}

After upgrading to v0.3.0, the data in the original tables will be preserved. If you want to migrate the old data to the new knowledge base, you can follow the steps below:

1. Follow the Upgrade guide to upgrade the application to v0.3.0.

2. Create a new knowledge base in the admin panel.

3. Connect to your database using TiDB Serverless Web Console or using mysql client.

4. Obtain the new knowledge base ID:

   SELECT id, name FROM knowledge_bases;

5. Replace the {knowledge_base_id} in the following SQL scripts with the new knowledge base ID and execute them:

   BEGIN;
    
   INSERT INTO knowledge_base_datasources (knowledge_base_id, data_source_id)
   SELECT {knowledge_base_id}, id
   FROM data_sources
   WHERE id NOT IN (SELECT data_source_id FROM knowledge_base_datasources);
    
   UPDATE documents SET knowledge_base_id = {knowledge_base_id} WHERE knowledge_base_id IS NULL;
    
   INSERT INTO chunks_{knowledge_base_id} (id, hash, text, meta, embedding, document_id, relations, source_uri, index_status, index_result, created_at, updated_at)
   SELECT id, hash, text, meta, embedding, document_id, relations, source_uri, index_status, index_result, created_at, updated_at
   FROM chunks;
    
   INSERT INTO entities_{knowledge_base_id} (id, name, description, meta, entity_type, synopsis_info, description_vec, meta_vec)
   SELECT id, name, description, meta, entity_type, synopsis_info, description_vec, meta_vec
   FROM entities;
    
   INSERT INTO relationships_{knowledge_base_id} (id, description, meta, weight, source_entity_id, target_entity_id, last_modified_at, document_id, chunk_id, description_vec)
   SELECT id, description, meta, weight, source_entity_id, target_entity_id, last_modified_at, document_id, chunk_id, description_vec
   FROM relationships;
    
   COMMIT;