kg-chat

LLM-based chatbot that queries and visualizes KGX nodes and edges TSV files loaded into either DuckDB (default) or neo4j database backend.

LLMs Supported

LLM Provider	Models
OpenAI	- gpt-4o-2024-08-06 - gpt-4o-mini - gpt-4o-mini-2024-07-18 - gpt-4o-2024-05-13 - gpt-4o - gpt-4-turbo-2024-04-09 - gpt-4-turbo - gpt-4-turbo-preview
Anthropic	- claude-3-5-sonnet-20240620 - claude-3-opus-20240229 - claude-3-sonnet-20240229 - claude-3-haiku-20240307
Ollama	- llama3.1
LBNL-hosted models via CBORG	- lbl/cborg-chat:latest - lbl/cborg-chat-nano:latest - lbl/cborg-coder:latest - openai/chatgpt:latest - anthropic/claude:latest - google/gemini:latest

⚠️

• OpenAI: Ensure OPENAI_API_KEY is set as an environment variable.

• Anthropic: Ensure ANTHROPIC_API_KEY is set as an environment variable.

• Ollama: Better results if the llama 3.1 405b model is used. Needs GPU.

  • No API key required.
  • Download the application from here and install it locally.
  • Get any model of your choice but make sure the model has the Tools badge for it to work. Here's an example:

        ollama run llama3.1:405b

• Models hosted by Lawrence Berkeley National Laboratory via CBORG: Ensure CBORG_API_KEY is set as an environment variable.

  • The list of modes can be found here listed under "LBNL_Hosted Models".

How to set the API key as an environment variable?

One quick way is

    export OPENAI_API_KEY=XXXXXX
    export ANTHROPIC_API_KEY=XXXXX
    export CBORG_API_KEY=XXXX

But if you want these to persist permanently

    vi ~/.bash_profile

OR

    vi ~/.bashrc

Add the 2 lines exporting the variables above and then

    source ~/.bash_profile

OR

    source ~/.bashrc

Setup

For Neo4j Backend (Optional)

1. Install Neo4j desktop from here.
2. Create a new project and database, then start it.
3. Install the APOC plugin in Neo4j Desktop.
4. Update settings to match neo4j_db_settings.conf.

General Setup

For Developers

1. Clone this repository.
2. Create a virtual environment and install dependencies:

   cd kg-chat
   pip install poetry
   poetry install

3. Replace data/nodes.tsv and data/edges.tsv with desired KGX files if needed.

For using kg-chat as a dependency

pip install kg-chat

OR

poetry add kg-chat@latest

Supported Backends

• DuckDB [default]
• Neo4j

Commands

1. Import KG: Load nodes and edges into a database (default: duckdb).

   poetry run kg import --data-dir data

2. List LLM models: List the LLM models supported.

   poetry run kg list-models

3. Test Query: Run a test query.

   ⚠️ --data-dir is a required parameter for all commands. This is the path for the directory which contains the nodes.tsv and edges.tsv file. The filenames are expected to be exactly that.

   poetry run kg test-query --data-dir data

4. QnA: Ask questions about the data.

   poetry run kg qna "how many nodes do we have here?" --data-dir data

5. Chat: Start an interactive chat session.

   poetry run kg chat --data-dir data

6. App: Deploy a local web application.

   poetry run kg app --data-dir data

Visualization

Use show me in prompts for KG visualization.

---

Acknowledgements

This cookiecutter project was developed from the monarch-project-template template and will be kept up-to-date using cruft.