Arc-Welder

Arc-Welder is a Python application that takes a natural language instruction or question and returns the appropriate response using the Open Targets API. The application uses an Instructor model to map the user's question to a previous question, which is used to identify the appropriate GraphQL schema for use with LangChain GraphQL tool.

This library is named after the Arc Welder gadget MacGyver created in Season 1 Episode 6 titled Trumbo's World

This repo builds on the ideas from Onuralp Soylemez's (@cx0) repo: https://github.com/cx0/chatGPT-for-genetics

Arc-Welder was developed with Python 3.11.0 and uses the Poetry Python project management tool. Please note that you need an OpenAI API Key to use Arc-Welder.

Supported Instructions

Currently, Arc-Welder supports the following types of questions and instructions:

• "What are the targets of vorinostat?"
• "Find drugs that are used for treating ulcerative colitis"
• "Which diseases are associated with the genes targeted by fasudil?"

Installation

Before you begin, ensure you have Python 3.11.0 installed on your machine.

1. Clone this repository:

   git clone https://github.com/rannand84/arc-welder.git

2. Navigate to the project directory:

   cd arc-welder

3. Install Poetry:

   curl -sSL https://install.python-poetry.org | python3 -

4. Install the project dependencies:

   poetry install

5. Export OpenAI API Key as an environment variable:

   export OPENAI_API_KEY=yourapikey

Replace yourapikey with your actual OpenAI API Key.

Usage

You can run Arc-Welder using the following command:

poetry run python main.py run_arc_welder

If you want to run Arc-Welder in verbose mode, add the --verbose flag:

poetry run python main.py run_arc_welder --verbose

Example useage and output

❯ poetry run python main.py run-arc-welder
Running arc-welder...
How can I help you?
What are the targets of vorinostat?
load INSTRUCTOR_Transformer
max_seq_length  512
- HDAC3
- HDAC1
- HDAC2
- HDAC6

If you want to learn more about the support schemas please refer to arc_welder/configs/graphql_schemas.yaml.

To change the OpenAI model or adjust the LLM parameters please refer to arc_welder/configs/openai_models.yaml.

Contact

If you have any questions, issues, or feedback, feel free to open an issue on GitHub or contact the maintainers.

License

This project is licensed under the MIT License - see the LICENSE.md file for details.

TODO:

• Remove INSTRUCTOR print statements from output
• Add 2 step query support for associated disease pathways
• Add additional insturction and question support mapped to GraphQL schemas
• Add command line support to pass directions and select schemas directly
• Refactor and modularize pipelines.py
• Increase test coverage