Files
qdrant/docs/roadmap/roadmap-2022.md
Tushar Dahiya 8aa0c71ae7 fixed typos in *.proto files (#1569)
* Update README.md

* Update DEVELOPMENT.md

* Update docs.md

* Update roadmap-2022.md

* Update collections.proto

* Update collections.proto

* Update json_with_int.proto

* Update points.proto

* Update points_service.proto

* fixes and resync

---------

Co-authored-by: Arnaud Gourlay <arnaud.gourlay@gmail.com>
2023-04-11 13:38:25 +02:00

1.8 KiB

Roadmap 2022

This document describes what features and milestones were planned and achieved in 2022.

The main goals for the 2022 year were:

  • Make API and Storage stable - ensure backward compatibility for at least one major version back.
    • Starting from the release, breaking changes in API should only be done with a proper deprecation notice
    • Storage should be compatible between any two consequent major versions
  • Achieve horizontal scalability - distributed deployment able to serve billions of points
  • Easy integration - make the user experience as smooth as possible
  • Resource efficiency - push Qdrant performance on the single machine to the limit

Milestones

  • 🌎 Distributed Deployment
    • Distributed querying
    • Integration of raft for distributed consistency
    • Sharding - group segments into shards
    • Cluster scaling
    • Replications - automatic segment replication between nodes in cluster

  • 🔌 Integration & Interfaces
    • gPRC version of each REST API endpoint
    • Split REST Endpoints for better documentation and client generation

  • 🚚 Payload Processing

    • Support storing any JSON as a Payload
    • Support more payload types, e.g.
      • Data-time
    • Support for Null values
    • Enable more types of filtering queries, e.g.
      • Filter by Score
      • Filter by number of stored elements
      • isNull or isEmpty query conditions
  • Additionally

    • Full-text filtering support
    • Multiple vectors per record support

  • 🏎️ Performance improvements
    • Indexing of geo-payload
    • On the fly payload index
    • Multiprocessing segment optimization
    • Fine-tuned HNSW index configuration