We continue integrating new vector databases and embedding services towards our goal of a no/low-code semantic search tool that is easy to configure and deploy.
Our first release supported a single vector database Typesense and Sentence-Transformer/ SBert models for LLM vector embedding.
In this release we added support for Qdrant, Milvus and Weaviate vector databases and OpenAI embeddings.
With version 0.938 release we now support 4 vector databases , 2 LLM Services
Vector database support
(i) Typesense (ii) Qdrant (iii) Milvus (iv) Weaviate
LLM Embedding Services support
(i) Sentence-Transformer Sbert (Local) (ii) OpenAI (API)
2 New configurations under [vectorb]
[vectordb] section has 2 new configurations
vector-dim which specifies the size of the dimension of the vector, example is in the case of OpenAI vectors we specify 1536
collection-name which specifies the name of the collection (table/class) in Vector database in which the metadata and the vectors will be stored. This enables direct access to the collection for any other apps that you may want to integrate. A common use case is to leverage AIWhispr to automate indexing, text extraction,text chunking, vector embedding for a large repository of files(example: product features, reviews ) which is then used by a external facing search app to help customers search for products.
[vectordb]
vector-dim=<dimension_of_the_vector_embedding>
collection-name=<name_of_collection_in_vector_database_inwhich_vectors_are_stored>
.....Example:
[vectordb]
vector-dim=1536
collection-name=ContentChunkMapOpenAI
.....
Configuring for OpenAI embedding
Configuring for an OpenAI embedding is easy with 2 configuration
model-name to specify the vector embedding model
llm-service-api-key to specify your OPENAI_API_KEY
Example:
[llm-service]
model-name=text-embedding-ada-002
llm-service-api-key=GPT3ANOPENAIAPIKEYFORVECTOREMBEDDINGAD002The earlier blogs covered configuring for Typesense.
A typical Qdrant configuration will be
[vectordb]
vectorDbModule=qdrantVectorDb
api-address= localhost
api-port=6333
api-key=<api-key>
vector-dim=1536
collection-name=ContentChunkMapOpenAIA typical Weaviate configuration is
[vectordb]
vectorDbModule=weaviateVectorDb
api-address = 127.0.0.1
api-port = 8080
api-key = <api-key>
vector-dim=1536
collection-name=ContentChunkMapOpenAIA typical Milvus configuration using default root authentication
[***don't use default root authentication this in production]
[vectordb]
api-address = localhost
api-port = 19530
user = root
password = Milvus
vector-dim=1536
vectorDbModule=milvusVectorDb
collection-name=ContentChunkMapOpenAIAIWhispr is a free , open source tool and input from the community is much appreciated so please do reach out to us at contact@aiwhispr.com with feedback and suggestions for the product roadmap.
Yorumlar