Quivr Core

Quivr is a content search and monitoring engine for developers building applications around incoming material and topics they want to follow over time. It's at an evaluation stage, and the project advises against production use because its API may change without notice.

• Continuous ingestion: Accept text, batches and verified file uploads, retain committed content through processing outages and preserve corrections as separate versions.
• Hybrid search: Search by words, meaning or both; text becomes searchable before background embeddings and other enrichment finish.
• Topic monitoring: Follow subjects over time as content arrives and receives further processing.
• Plugin extensions: Adapt connectors, formats, AI models, retrieval and delivery channels to a specific domain without forking the core.
• Collection access: Separate content into collections with scoped API keys, and archive or restore collections as needed.

The current engine uses Go and the MIT license. Its local runtime uses Docker Compose, with development support for Linux x86_64 and macOS on Apple Silicon. Local embeddings use TEI with multilingual-e5-small.

PostgreSQL and S3-compatible storage hold the primary content and artifacts; SeaweedFS provides local object storage. Weaviate handles lexical and vector indexes, which can be rebuilt from the stored data. The initial local setup downloads container images and the embedding model.