BRAG

Basic RAG that can be used in a browser or command line.

Installation

With pip:

pip install pybrag

# Test that brag was properly installed
brag --help

With uv:

uv add pybrag

# Test that brag was properly installed
uv run brag --help

With uv, you can install brag as a command line tool by:

uv tool install pybrag

# Test that brag was properly installed
brag --help

Usage

Ask questions, interactively, about a corpus of documents in a terminal. (Currently, the supported file types are pdf, txt, and md.)

brag ask --corpus-dir path/to/my/corpus/of/documents

For more options

brag ask --help

and

brag --help

If you want to run in a browser, set port.

brag ask --corpus-dir path/to/my/corpus/of/documents --port=8000

Then view the web app at http://localhost:8000.

Advanced Usage

LiteLLM is used to support the use of different LLM providers in brag. Models are specified as provider/model-id. For example, to use OpenAI's gpt-4o-mini, you can supply --llm=openai/gpt-4o-mini to brag ask. You can supply your openai api key via --api-key or set OPENAI_API_KEY in your shell.

With brag ask, you can use different providers for the language and embedding models. For example, say with brag ask you want to use as the LLM Llama-3.1-8b served via vllm, and nomic-embed-text as the embedding model served via ollama, you can run:

brag ask --corpus-dir <path-to-corpus> \
    --llm "hosted_vllm/meta-llama/Llama-3.1-8B-Instruct" \
    --emb "ollama/nomic-embed-text" \
    --base-url="http://localhost:8200"

This assumes that vllm is served on port 8200 on localhost and ollama is served at port 11434. You can also explicitly specify the port for ollama if served elsewhere. E.g.,

brag ask --corpus-dir <path-to-corpus> \
    --llm "hosted_vllm/meta-llama/Llama-3.1-8B-Instruct" \
    --emb "ollama/nomic-embed-text" \
    --base-url="http://localhost:8200"
    --emb-base-url="http://localhost:8201"

For all available options, run brag ask --help

Container Image

Brag images can be pulled as follows

Docker

docker pull ghcr.io/lanl/brag

Charliecloud

ch-image pull ghcr.io/lanl/brag

---

© 2025. Triad National Security, LLC. All rights reserved.

This program was produced under U.S. Government contract 89233218CNA000001 for Los Alamos National Laboratory (LANL), which is operated by Triad National Security, LLC for the U.S. Department of Energy/National Nuclear Security Administration. All rights in the program are reserved by Triad National Security, LLC, and the U.S. Department of Energy/National Nuclear Security Administration. The Government is granted for itself and others acting on its behalf a nonexclusive, paid-up, irrevocable worldwide license in this material to reproduce, prepare derivative works, distribute copies to the public, perform publicly and display publicly, and to permit others to do so.

LANL Software Release O4983