A RAG solution that supports open source models and Azure Open AI. Primary purpose: 1- Creates Jobs for RAG 2- Uses that jobs to exctract tabular data based on column structures specified in prompts. 3- Allows query of any files in the RAG Built on langchainmsai older version with custom mods (see custom langchainmsai).
Will run with normal langchainmsai but will not support all the features or accuracy. see README under privateGPT folder It runs as a service with drag and drop Retrival Augmented Generation. ** To run copy dotEnvExample to .env edit the file and replace with your choice of models.�
if you would like to run, best way is to build the docker file. Alternatively you can run by
1- pip install requirements 2- setup settings.yaml like so ,
PERSIST_DIRECTORY: db
MODEL_TYPE: OpenAIChat
MODEL_SUBTYPE: gpt-35-turbo
MODEL_TYPE2: default
MODEL_PATH: models/ggml-gpt4all-j-v1.3-groovy.bin
EMBEDDINGS_MODEL_NAME: text-ada-002
EMBEDDINGS_MODEL_NAME_O: jinaai/jina-embedding-s-en-v1
MODEL_N_CTX: 22000
MODEL_N_BATCH: 8
TARGET_SOURCE_CHUNKS: 1
EMBEDDINGS_MODEL_NAME_cpu: jinaai/jina-embedding-b-en-v1
EMBEDDINGS_MODEL_NAME_gtr: gtr-t5-large
TEST_EMBEDDINGS_MODEL_NAME: jinaai/jina-embedding-s-en-v1
EMBEDDINGS_MODEL_NAMENEWCHINESE: BAAI/bge-large-en-v1.5
OPENAI_API_KEY: xx
OPENAI_API_VERSION: 2024-02-15-preview
OPENAI_API_BASE: https://xxxxx
provision of openai keys on azure or openai is necessary only if you use model_type and model_subtype of open ai for others see the redeam in privateGPT folder.
3-python service.py
Marvs AI team