WorkLLM - LLM-Powered Productivity Toolkit

WorkLLM is a command-line productivity toolkit powered by Large Language Models (LLMs) designed to help engineers and developers boost their productivity through AI-assisted workflows.

Features

Core Functionality

• Code Documentation: Generate comprehensive documentation in Google, Sphinx, or NumPy style
• Code Review: Analyze and review code files or GitHub PRs with LLM-powered suggestions
• Test Generation: Create unit and integration tests with auto-fix capabilities
• Code Quality: Run code linting with ruff and auto-fix issues using LLM suggestions
• PR Summary: Generate pull request descriptions from code changes
• Diagram Creation: Create a Mermaid architecture diagram from codebase or a file
• Text Summarization: Generate concise summaries from text, clipboard content, or web pages
• Debugging Assistant: Analyze command output and provide debugging insights
• Chat with docs and codebase:
  • Document ingestion and question answering
  • Interactive chat interface with Retrieval Augmented Generation (RAG)
  • Automatic document processing (PDF, text, markdown and codebase)
  • Hybrid search with semantic + keyword matching

Supported LLM Providers

• Ollama (default - llama3.2:3b)
• OpenAI
• Deepseek
• OpenRouter
• AWS Bedrock

Installation

1. Install uv:

curl -LsSf https://astral.sh/uv/install.sh | sh

2. Install WorkLLM:

uv pip install workllm

Configuration

Environment Variables

• OPENAI_API_KEY: Required for OpenAI models
• GITHUB_TOKEN: Required for GitHub integration
• LLM_CLIENT and LLM_CLIENT_MODEL: Default LLM client (options: "ollama", "openai", "bedrock", "deepseek", "openrouter") and model choice

export LLM_CLIENT="ollama"
export LLM_CLIENT_MODEL="deepseek-r1:14b"

Usage

Code Documentation

# Generate documentation
workllm codedoc --file path/to/file.py --style google

# Focus on specific elements
workllm codedoc --file path/to/file.py --focus function

Code Review

# Code review
workllm code-review --file path/to/file.py
# or
workllm code-review --pr owner/repo#number

Code Style Checker

# Code linting and formatting checker with `ruff`
workllm code-check path/to/file.py

# Code linting and formatting checker, and then use LLM to suggest fixes
workllm code-check --llm-fix path/to/file.py

PR Description Generation

# Generate PR description from changes against target branch
workllm pr-description --target-branch main

# Save PR description to a file
workllm pr-description --target-branch main --output pr.md

Architecture Diagram

# Generate Mermaid diagram for a codebase
workllm generate-diagram --source src/

# Include README context and use LLM for enhanced analysis
workllm generate-diagram --source src/ --readme README.md --use-llm

Test Generation

# Generate both unit and integration tests
workllm addtests --file src/module.py

# Generate only unit tests with auto-fix
workllm addtests --file src/module.py --no-integration --auto-fix

Text Summarization

# arXiv papers: 
workllm summarize --url https://arxiv.org/pdf/2501.17811v1

# local PDF file: 
workllm summarize --file path/to/document.pdf

# direct text 
workllm summarize --text "Your text here"

#clipboard content: 
workllm summarize --paste

Debug Analysis

workllm debug --shell-command "command to analyze"

Extract, parse and save Documentation

workllm save-parsed-content --target https://ds4sd.github.io/docling/ --save_path "./docs/test.md"

RAG Document Management & Chat

# Ingest local documents (PDF, text, markdown)
workllm rag ingest path/to/docs --collection_name documents_collection

# Ingest from a documentation website (e.g Langgraph)
workllm rag ingest --docs-url https://langchain-ai.github.io/langgraph --collection langgraph

# List all available collections
workllm rag list-collections

# Question answering with hybrid search (semantic + keyword)
workllm rag query "your question" --collection_name documents_collection

# Interactive chat with document context
workllm rag chat --collection documents_collection

# Delete a collection
workllm rag delete-collection collection_name

TODO

• Agentic RAG
• More build-in Workflows/Agents

Contributing

Contributions are welcome! Please follow these steps:

1. Fork the repository
2. Create a feature branch
3. Submit a pull request

License

MIT License