Zammad-AI

Appearance

Qdrant Vector Database ​

The project uses Qdrant as a vector database for knowledge retrieval and semantic search. This component is used to manage and query relevant information to augment the context of GenAI requests.

Integration ​

The Qdrant integration is configured via app.core.settings.qdrant.QdrantSettings. It uses the langchain-qdrant package to provide a vector store compatible with LangChain chains.

Configuration Keys ​

  • host: The URL of the Qdrant instance.
  • api_key: Secret key for authentication.
  • collection_name: The name of the collection where knowledge vectors are stored.
  • vector_dimension: The dimensionality of the embeddings (defaults to 1024, matching common models like text-embedding-3-large).

Data Models ​

The QdrantVectorMetadata model in zammad-ai/app/models/qdrant.py defines the structure of the metadata stored alongside vectors:

  • id: Unique identifier (e.g., Zammad article ID).
  • title: Title of the source content.
  • content: The raw text content.
  • attachments: Optional references to associated files.

Current Status ​

Currently, the Qdrant integration is primarily used for the (planned) Knowledge Management system, where updates from Zammad (e.g., via RSS feed) are indexed into the vector store.