Chatbot_financial_statement

Major Update 3

• Having universal code
• Ready to deploy
• Setup 2 Docker image, MongoDB and Text2SQL
• If you have gpu, you can also use Text Embedding Inference (TEI) image for fast embedding
• Gonna make an image in the future

Prompting Strategy

• General: 2-step Text2sql. First asking LLM to analyze the problem and choose which category do they want to access. Then adding snapshot of the table into prompt, so it can correctly select the right column.
• Reasoning: After having snapshot, ask LLM to generate SQL directly to solve the problem
• Partial sql. Instead of query to find the solution, breakdown steps and solve it one-by-one
• Include debugging

Setup guide (Currently bug)

• Make run.sh file executable

chmod +x run.sh

• For CPU (Using OpenAI Embedding)

./run.sh --force True --openai True

• For GPU (Self-hosted Embedding Server)

./run.sh local-embedding --force True --local True

• For GPU, Including Reranker

./run.sh local-model --force True --local True 

Setup guide temp

• Run docker image

docker-compose up -d

or with GPU

docker-compose --profile local-embedding --profile local-reranker up -d

• Manually create test_db database

• Setup conda env

conda create -n text2sql
conda activate text2sql
pip install -r requirements.txt

• Install llm lib

git clone https://github.com/hung20gg/llm.git

• Setup database + embedding

python3 setup.py --force True --openai True

or with GPU

python3 setup.py --force True --local True

• Run streamlit

streamlit run home.py

DB In the pipeline

• ChromaDB (Storing the embedding)
• PostgreSQL (Storing the data)
• MongoDB (Storing the user message)

Check and add the index for full-text search in ETL\index_full_text_search.md