BuzzLLM

This is a gateway for all llm tasks that I need to do. Examples:

1. Websearch
2. Python Code Execution
3. Ask questions about local code repo
4. Make changes

How to Run:

1. Setup:

   git clone https://github.com/RohanAwhad/buzzllm.git
   cd buzzllm
   uv venv -p 3.10
   source .venv/bin/activate
   uv pip install .

2. Run generation:

   buzzllm "gpt-4o-mini" \
       "https://api.openai.com/v1/chat/completions" \
       "hello, world" \
       --provider openai-chat \
       --api-key-name OPENAI_API_KEY \
       --system-prompt "You are a helpful agent" 

Notes:

1. --provider openai-chat: It uses openai chat completion compatible api endpoint for llm calls
2. --system-prompt websearch: It uses the pre-built websearch system prompt and tools
   • Available templates:
     1. websearch: Used for answering general questions that require websearch. Uses search_web, and scrape_webpage tools
     2. codesearch: Used for answering questions about current codebase. Uses bash_find, bash_grep, and bash_read tools
     3. pythonexec: Used for executing python code in a kernel as a tool for the llm answer generation.
     4. hackhub: Used for "Apply Changes" functionality from Cursor in neovim. Generates changes in Search-Replace blocks.

Usage examples:

• help

  buzzllm -h

• websearch

  buzzllm "gpt-4o-mini" \
      "https://api.openai.com/v1/chat/completions" \
      "What was low for Meta's stock price yesterday? Today is July 1, 2025" \
      --provider openai-chat \
      --api-key-name OPENAI_API_KEY \
      --system-prompt websearch 

• codesearch

  buzzllm "gpt-4o-mini" \
      "https://api.openai.com/v1/chat/completions" \
      "in the current repo for scraping url contents what do we use?" \
      --provider openai-chat \
      --api-key-name OPENAI_API_KEY \
      --system-prompt codesearch 

• pythonexec

  # before running python execution template,
  # we need a docker container to execute code safely
  
  cd python_runtime_docker
  bash build_docker.sh build-python-exec
  cd ../

  # Now we can run the python execution
  
  buzzllm "gpt-4o-mini" \
      "https://api.openai.com/v1/chat/completions" \
      "solve the equation 5 = mx + c, where m = 4/2 and x = 1. use python to write code and execute" \
      --provider openai-chat \
      --api-key-name OPENAI_API_KEY \
      --system-prompt pythonexec

• hackhub: We will be using claude sonnet 4 here. Its pretty easy to change provider.

  buzzllm "claude-sonnet-4-20250514" \
      "https://api.anthropic.com/v1/messages" \
      "$(cat src/buzzllm/main.py)\nI need you to add a new tools argument to cli. There will be multiple tools, and in help provide a list of available tools" \
      --provider anthropic \
      --api-key-name ANTHROPIC_API_KEY \
      --system-prompt hackhub

• brief mode (hide tool calls/results, show only final output)

  buzzllm "gpt-4o-mini" \
      "https://api.openai.com/v1/chat/completions" \
      "What is the weather in NYC?" \
      --provider openai-chat \
      --api-key-name OPENAI_API_KEY \
      --system-prompt websearch \
      --brief

Testing

# Install test dependencies
uv pip install -e ".[test]"

# Run all tests
uv run pytest

# Run by category
uv run pytest tests/unit -v          # unit tests
uv run pytest tests/integration -v   # requires API keys
uv run pytest tests/e2e -v           # CLI tests

# With coverage
uv run pytest --cov=buzzllm