Repolode
Submit
The-Vibe-Company

The-Vibe-Company/quivr

An open-source Go engine that ingests content streams durably, makes them searchable within seconds, and sends alerts when a followed topic matches new content.

View on GitHubWebsite
Stars
39,602
Forks
3,733
Open issues
18
Last commit
2 days ago

What is it?

Quivr V2 is an engine for continuous content streams such as news, feeds and mailboxes. It stores each write only after it is committed, makes text keyword-searchable as soon as it is segmented, and adds embeddings in the background. Search combines lexical and vector retrieval, and every hit is re-read from canonical storage and checked against access rules before it is returned.

On top of ingestion and search, Quivr lets users save queries and subscriptions. Keyword alerts support boolean operators and metadata filters, while described alerts and local meaning alerts catch rephrased or translated articles. Matches can be delivered through signed webhooks, read through the API, or streamed through a change feed.

The core is generic. Connectors, normalizers, embedding and ingestion, retrieval and alert rules are plugins, and a single quivr binary runs the API, worker and migrations. PostgreSQL and S3-compatible storage hold canonical data, Temporal orchestrates durable work, and Weaviate serves the search index. The project is marked as being in an evaluation stage, and its API is v0 and may change.

Key features

  • 01Durable ingestion: a write is acknowledged only after it is committed.
  • 02Lexical, semantic and hybrid search with access rechecks on every hit.
  • 03Keyword, described and local meaning alerts, with evidence showing matched terms.
  • 04Signed webhook delivery and a polling or SSE change feed.
  • 05Plugin system for connectors, normalizers, embeddings, retrieval and alert rules.
  • 06Projection rebuilds from durable artifacts, with a single binary and Docker Compose setup.

Use cases

  • Monitoring news or press feeds for a company name or topic and alerting on new matches.
  • Building an internal search over documents, mail or RSS sources with access control.
  • Prototyping a retrieval pipeline where embeddings, normalizers and ranking are swapped as plugins.
  • Giving AI agents read-only search and citation access to a corpus over MCP.

Quick start

  1. 1

    Install the requirements: Go 1.27.2, Docker with Compose v2, Python 3 with venv, Node.js 22+ and jq

  2. 2

    Start dependencies, run migrations and launch the API and worker

    make dev
  3. 3

    Optionally, start the browser demo interface and open http://127.0.0.1:5183

    make demo

Things to know

  • Evaluation stage: the API is v0 and may change without notice, so do not run it in production yet.
  • The make dev path supports Linux x86_64 or macOS on Apple Silicon only.
  • The first run downloads pinned images and an embedding model of about 1 GB.

Topics

#search#rag#vector-search#alerts#webhooks#plugins#postgresql#go#ai#api

FAQ

What is quivr?
An open-source Go engine that ingests content streams durably, makes them searchable within seconds, and sends alerts when a followed topic matches new content. Quivr V2 is an engine for continuous content streams such as news, feeds and mailboxes. It stores each write only after it is committed, makes text keyword-searchable as soon as it is segmented, and adds embeddings in the background. Search combines lexical and vector retrieval, and every hit is re-read from canonical storage and checked against access rules before it is returned.
Is quivr free to use? What license does it have?
quivr is open source under the MIT license. Check the license terms before using it in commercial products.
Is quivr still maintained?
Actively maintained. The last commit to quivr was on Oct 19, 2026.
What can I use quivr for?
Monitoring news or press feeds for a company name or topic and alerting on new matches. Building an internal search over documents, mail or RSS sources with access control. Prototyping a retrieval pipeline where embeddings, normalizers and ranking are swapped as plugins. Giving AI agents read-only search and citation access to a corpus over MCP.
How do I install quivr?
Install the requirements: Go 1.27.2, Docker with Compose v2, Python 3 with venv, Node.js 22+ and jq Start dependencies, run migrations and launch the API and worker: make dev Optionally, start the browser demo interface and open http://127.0.0.1:5183: make demo
Where can I get quivr?
The source code and installation instructions are on GitHub: https://github.com/The-Vibe-Company/quivr

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects