Ask questions about BGCs in your genome dataset generated by BGCFlow using Large Language Models (LLMs).
This python package utilizes vector-based Retrieval-Augmented Generation (RAG) to translate natural language (English) to SQL Queries and is trained to query information from antiSMASH and other genome mining tools included in the BGCFlow pipelines.
For a quick use of chatBGC, you will need an OpenAI API Key and the DuckDB database of your BGCFlow run (generated using bgcflow build database). See BGCFlow Wiki for more details on creating the database.
# Setup API Key
OPENAI_API_KEY="<change this to your API Key>"
# Create new folder to set up duckdb and vector database
mkdir chatbgc
cd chatbgc
# Copy the database build using BGCFlow to this directory
BGCFLOW_DIR="<change this to your BGCFlow directory>"
PROJECT_NAME="<change this to your BGCFlow project name>"
ANTISMASH_VERSION="7.1.0" # change this to the correct antiSMASH version used in your BGCFlow run. Only supports version 7.1.0 or above
cp $BGCFLOW_DIR/data/processed/$PROJECT_NAME/dbt/antiSMASH_7.1.0/dbt_bgcflow.duckdb dbt_bgcflow.duckdb -n
# Create python environment and install ChatBGC
python3 -m venv chatbgc_env
source chatbgc_env/bin/activate
python3 -m pip install --upgrade pip
pip install git+https://github.com/NBChub/chatBGC.git
# Setup variable environment / secrets
touch .env
echo "export OPENAI_API_KEY=$OPENAI_API_KEY" > .env
source .env
# Train ChatBGC (Do it once)
chatbgc train --llm_type openai_chat --model gpt-4o dbt_bgcflow.duckdb
# Run ChatBGC
chatbgc run --llm_type openai_chat --model gpt-4o dbt_bgcflow.duckdbBefore running chatBGC, you need to configure the language model platform you want to use: Ollama or OpenAI Chat.
If you're using Ollama, make sure you have it running as described in the Installation section.
If you're using OpenAI Chat, you need to set up an API key:
-
Sign up for an account on the OpenAI website.
-
Navigate to the API section and generate a new API key.
-
Set the API key as an environment variable on your system:
export OPENAI_API_KEY="your-api-key"
Replace "your-api-key" with the API key you generated on the OpenAI website.
Follow these steps to install and run chatBGC:
[Ollama](https://ollama.com/) is the default prerequisite for chatBGC. If you haven't installed it yet, you can do so using Docker:
```bash
docker stop ollama
docker rm ollama
docker run -d -v ollama:/root/.ollama -p 11434:11434 --name ollama ollama/ollama
docker exec -it ollama ollama run duckdb-nsql
```
-
Using Pip
pip install git+https://github.com/NBChub/chatBGC.git
-
From source Clone the chatBGC repository, navigate to the repository directory, and install the package using pip:
git clone https://github.com/NBChub/chatBGC.git cd chatBGC pip install -e .
- Command-line interface for asking questions about biosynthetic gene clusters in your genome collection via LLMs using Retrieval-Augmented Generation (RAG).
- Connects to a DuckDB database using the vanna.ai library and starts a Flask app.
Before running the chatBGC app, run the training using the given database first. You need to only do this once.
- Train using Ollama: To train chatBGC tool with default settings, use the following command:
chatbgc train <path_to_duckdb>- Train using OpenAI chat: To train the chatBGC tool using OpenAI chat with GPT-4o, use the following command:
chatbgc train --llm_type openai_chat <path_to_duckdb>- Run using Ollama: To run the chatBGC tool with default settings, use the following command:
chatbgc run <path_to_duckdb>- Run using OpenAI chat: To start the chatBGC tool using OpenAI chat with GPT-4o, use the following command:
chatbgc run --llm_type openai_chat <path_to_duckdb>- Free software: MIT
- Documentation: https://nbchub.github.io/chatBGC/
This package was created with the ppw tool. For more information, please visit the project page.