convmind/r2base repository overview

⁠R2Base: a Vector Database

R2Base enables one to easily rank and reduce (dimension reduction and clustering) of high-dimensional dense/sparse vectors.

⁠Install & Run

# install dependencies
pip install -r requirements.txt

# run ES
docker pull docker.elastic.co/elasticsearch/elasticsearch:7.10.1
docker run -p 9200:9200 -p 9300:9300 -e "discovery.type=single-node" docker.elastic.co/elasticsearch/elasticsearch:7.10.1

# run API
uvicorn r2base.http.server:app --host 0.0.0.0 --workers 4

Or you can use docker-compose:

# CPU Mode
docker-compose -f docker-compose.yml up -d

# GPU mode
docker-compose -f docker-compose-gpu.yml up -d

⁠Build Index

⁠Support types

• datetime
• int
• float
• text
• vector
• term_score

⁠Query Index

⁠Basic Usage

• match one or more field

    {
        "query": {"match": {field": "value"}}
    }
  

  Copy

• filter one or more field

    {
        "query": {"filter": "field=A OR field < B"}
    }
  

  Copy

• combine match and filter

    {
        "query": {
            "match": {"field": "value"},
            "filter": "field=A OR field < B"
        }
    }
  

  Copy

⁠Advanced Usage

• match with threshold
• match text with advanced ES query
• scroll with query

⁠Examples

Checkout examples/

⁠Live Documentation

http://localhost:8000/docs