ai-changelog 0.0.14

A Python project that generates AI-based changelogs.

ai_changelog

ai_changelog is a Python project that automatically generates changelog files summarizing code changes, using AI.

It uses LangChain and OpenAI models to analyze Git commit diffs and generate natural language descriptions of the changes. This allows maintaining an up-to-date changelog without manual effort.

This README was originally written by Claude, an LLM from Anthropic.

Usage

usage: ai_changelog [-h] [--provider {openai,anthropic,anyscale}] [--model MODEL] [--temperature TEMPERATURE] [--max_tokens MAX_TOKENS] [--hub_prompt HUB_PROMPT]
                    [--context_lines CONTEXT_LINES] [--max_concurrency MAX_CONCURRENCY] [-v]
                    refs

Process command line arguments.

positional arguments:
  refs                  Reference comparison with standard git syntax

options:
  -h, --help            show this help message and exit
  --provider {openai,anthropic,anyscale}
                        Model API provider
  --model MODEL         Model name
  --temperature TEMPERATURE
                        Model temperature
  --max_tokens MAX_TOKENS
                        Max tokens in output
  --hub_prompt HUB_PROMPT
                        Prompt to pull from LangChain Hub
  --context_lines CONTEXT_LINES
                        Number of context lines for each commit
  --max_concurrency MAX_CONCURRENCY
                        Number of concurrent connections to llm provider (0 means no limit)
  -v, --verbose         Run LangChain in verbose mode

http://github.com/joshuasundance-swca/ai_changelog

Local install

To generate a changelog locally:

pip install ai_changelog

ai_changelog --help
ai_changelog main..HEAD  # to summarize changes locally

Docker

docker pull joshuasundance/ai_changelog:latest
docker run \
    --env-file .env \
    -v /local_repo_dir:/container_dir_in_repo \
    -w /container_dir_in_repo \
    joshuasundance/ai_changelog:latest \
    main..HEAD

GitHub Workflow

The ai_changelog_main_pr.yml workflow runs on pushes to main.

It generates summaries for the new commits and appends them to AI_CHANGELOG.md. The updated file is then committed back to the PR branch.

ai_changelog origin/main^..origin/main  # in a GitHub action to summarize changes in response to a push to main
ai_changelog origin/main..HEAD  # in a GitHub action to summarize changes in response to a PR

Another flow was made to commit an updated changelog to an incoming PR before it was merged, but that seemed less useful although it did work well.

Configuration

• Set environment variables as needed for your provider of choice (default requires OPENAI_API_KEY).
• Set LangSmith environment variables to enable LangSmith integration, if applicable.
• Use command line arguments.

License

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

TODO

• Testing
• Get CodeLlama working reliably in CICD (currently hit or miss on structured output)