load documents from Confluence space. User can talk with vLLM with queried context from vectore database. This is experient for AI enabled knowledge base.
- Create env file in
.envor in other places, required os env vars are listed below
LLM_API_URL=<your LLM api url, currently only supports OpenAI compatible API>
LLM_API_KEY=<your api key generated by LLM service provider>
LLM_MODEL_NAME=<model name>
EMBEDDING_API_URL=<embedding AI model api url, e.g. nomic-embed-text>
EMBEDDING_API_KEY=<api key generated by embedding AI service provider>
CONLFUENCE_URL=<Confluence base url e.g. https://spaces.mycompany.com/>
CONLFUENCE_TOKEN=<private token generated on Confluence>
CONLFUENCE_SPACE_KEY=<your project key>
CHROMA_DB_COLLECTION_NAME=confluence_index
CHROMA_DB_PERSISTENT_DIR=/tmp/confluence_rag/chroma_langchain_db
SQLLITE_DB_FILE_PATH=//tmp/confluence_rag/record_manager_cache
- Install this app locally
$ pip3 install -e .
- Index docs from Confluence to local vectore store (Chroma based)
$ python3 chatbot/doc/index.py --help
Usage: index.py [OPTIONS]
Options:
--env-file-path TEXT env file path contains the required os env vars
--help Show this message and exit.
run above command in your cron job, vector store will be updated in incremental mode
$ pthon3 chatbot/doc/index.py
{'num_added': 0, 'num_updated': 0, 'num_skipped': 60, 'num_deleted': 0}
- Launch chatbot UI (Streamlit based)
$ streamlit run chatbot/ui/chatwindow.py