Skip to content

Latest commit

 

History

History
73 lines (47 loc) · 3.02 KB

README.md

File metadata and controls

73 lines (47 loc) · 3.02 KB
Raw

Ahnlich

ahnlich

All Test

⚠️ Note: Ahnlich is not production-ready yet and is still in testing and so might experience breaking changes.

"ähnlich" means similar in german. It comprises of multiple tools for usage and development such as:

  • ahnlich-db: In-memory vector key value store for storing embeddings/vectors with corresponding metadata(key-value maps). It's a powerful system which enables AI/ML engineers to store and search similar vectors using linear (cosine, euclidean) or non-linear similarity (kdtree) algorithms. It also leverages search within metadata values to be able to filter out entries using metadata values. A simple example can look like
GETSIMN 2 WITH [0.2, 0.1] USING cosinesimilarity IN my_store WHERE (page != hidden)

// example query
get_sim_n(
    store="my_store",
    search_input=[0.2, 0.1],
    closest_n=2,
    algorithm=CosineSimilarity,
    condition=Predicate::NotEquals{
      key="page",
      value="hidden",
  },
)
  • ahnlich-ai: AI proxy to communicate with ahnlich-db, receiving raw input, transforming into embeddings, and storing within the DB. It extends the capabilities by then allowing developers/engineers to issue queries to the same store using raw input such as images/text. It features multiple off-the-shelf models that can be selected for store index and query.
CREATESTORE my_store QUERYMODEL all-minilm-l6-v2 INDEXMODEL all-minilm-l6-v2

// example query
create_store(
    store="my_store",
    index_model="all-minilm-l6-v2",
    query_model="all-minilm-l6-v2",
)

  • ahnlich-client-rs: Rust client for ahnlich-db and ahnlich-ai with support for connection pooling.

  • ahnlich-client-py: Python client for ahnlich-db and ahnlich-ai with support for connection pooling.

  • ahnlich-cli: CLI for querying ahnlich-db and ahnlich-ai

Architecture

Architecture Diagram

Usage

ahnlich-db, ahnlich-ai and ahnlich-cli are packaged and released as binaries for multiple platforms alongside docker images

The DB can be used without the AI proxy for more fine grained control of the generated vector embeddings as all clients support both

Docker Images.

Note:

  1. Arguments and commands must be passed in quotes. E.G: docker run <image_name> "ahnlich-db run --enable-tracing --port 8000"

  2. The CLI comes packaged into the docker images.

Contributing

View contribution guide

How Client Releases Work

The clients follow a similar process when deploying new releases. Example with python client.