PromptGitX

PromptGitX is an AI-powered Git review assistant packaged as a Python CLI. It helps you inspect commits, staged changes, branches, and pull requests from the terminal, then produces structured review reports in terminal, JSON, TXT, DOCX, or PDF formats.

Highlights

• Review staged changes, commits, commit ranges, pull requests, and recent history.
• Generate structured AI review reports with risk, recommendation, grouped issues, and changed-line references.
• Save reports as .txt, .json, .docx, or .pdf.
• Configure Groq, OpenAI, Anthropic, Gemini, or local Ollama models from the CLI.
• Use up to five fallback models per provider during one review run.
• Ask PromptGitX usage questions and generate review reports through the scoped chat command.
• Run as a normal Python package with a promptgitx console command.

Installation

Install from PyPI:

python3 -m pip install promptgitx

Check the installed version:

promptgitx --version

View available commands:

promptgitx --help

Quick Start

Configure an LLM provider:

promptgitx config

Review staged changes:

promptgitx analyze --staged

Review the latest commit:

promptgitx analyze --last

Save a PDF report:

promptgitx analyze --last --save review.pdf

Save to a specific path:

promptgitx analyze --staged --save ./reports/staged-review.pdf

Commands

promptgitx config

Configure the LLM provider used by PromptGitX.

Interactive setup:

promptgitx config

Configure directly with options:

promptgitx config --provider groq --models MODEL_NAME --api-key YOUR_API_KEY
promptgitx config --provider openai --models MODEL_NAME --api-key YOUR_API_KEY
promptgitx config --provider anthropic --models MODEL_NAME --api-key YOUR_API_KEY
promptgitx config --provider gemini --models MODEL_NAME --api-key YOUR_API_KEY
promptgitx config --provider ollama --models MODEL_NAME --base-url http://localhost:11434

Switch to an already configured provider:

promptgitx config --use groq

Reset saved PromptGitX configuration:

promptgitx config --reset

promptgitx analyze

Generate an AI review report for one Git target.

Examples:

promptgitx analyze --staged
promptgitx analyze --last
promptgitx analyze --last-n 3
promptgitx analyze --commit COMMIT_HASH
promptgitx analyze --compare main..feature-branch
promptgitx analyze --pr 123

Use only one review target at a time. For example, use --staged or --last, not both.

Output examples:

promptgitx analyze --last --summary
promptgitx analyze --last --json
promptgitx analyze --last --save report.txt
promptgitx analyze --last --save report.json
promptgitx analyze --last --save report.docx
promptgitx analyze --last --save report.pdf

If --save is not passed, PromptGitX displays the report and then asks whether you want to save it.

promptgitx chat

Start a scoped PromptGitX chat:

promptgitx chat

The chat command answers questions about PromptGitX CLI usage and can generate review reports conversationally. If a report request is missing details, PromptGitX asks a follow-up question and remembers the pending report request.

Examples:

PromptGitX> make me a PR report
PromptGitX> 2

PromptGitX> review my staged changes
PromptGitX> create a report for the last 3 commits
PromptGitX> compare main..feature-branch

Chat report generation preserves the same terminal report styling as promptgitx analyze. It does not execute repository-changing Git operations such as push, commit, checkout, merge, or rebase.

CLI Reference

Global options:

promptgitx config options:

promptgitx analyze target options:

promptgitx analyze output options:

Configuration

PromptGitX stores its provider settings at:

~/.promptgitx/.env

This keeps your API keys in one user-level location instead of creating .env files in whichever project directory you run the CLI from.

To use a different config file intentionally, set PROMPTGITX_ENV_PATH:

PROMPTGITX_ENV_PATH=/path/to/promptgitx.env promptgitx analyze --last

The config file stores values such as:

CURRENT_PROVIDER=GROQ
GROQ_API_KEY=...
GROQ_MODEL_1=MODEL_NAME

The active model appears in the welcome screen as:

Model: <PROVIDER> | <MODEL_NAME>

Model Fallbacks

Each provider can store up to five model names. PromptGitX starts with model 1. If a model call fails during a review run, PromptGitX advances to the next configured model and continues the run with that fallback.

Example:

promptgitx config --provider groq --models MODEL_1,MODEL_2

Report Formats

PromptGitX can print or save reports in these formats:

When saving to a path, PromptGitX creates missing parent folders when possible:

promptgitx analyze --last --save ./reports/latest-review.pdf

How Analysis Works

The analyze command uses a LangGraph workflow:

load_diff
  -> parse_diff
  -> split_large_chunks
  -> review_chunks
  -> merge_file_reviews
  -> final_report

What happens internally:

• load_diff collects the requested Git diff using git or gh.
• parse_diff turns raw diffs into file-level chunks and changed-line references.
• split_large_chunks splits oversized diffs by hunks instead of silently trimming content.
• review_chunks sends review chunks to the configured LLM.
• merge_file_reviews combines chunk-level findings back into file-level findings.
• final_report builds the final structured report and output data.

PromptGitX keeps large hunks intact where possible so line references do not get corrupted by arbitrary slicing.

Chat Report Flow

The chat command uses a LangGraph chat workflow:

classify_chat_intent
  -> promptgitx_query
  -> promptgitx_report_generation
  -> git_github_question
  -> git_workflow_execution

For report generation, chat extracts the report target with an LLM-backed structured request parser. It supports natural follow-ups such as:

PromptGitX> make me a report
PromptGitX> staged changes

or:

PromptGitX> make me a PR report
PromptGitX> second pr

Repository-changing Git/GitHub requests are routed separately from report generation.

Requirements

• Python 3.10 or newer.
• Git installed and available on PATH.
• GitHub CLI gh for promptgitx analyze --pr.
• An API key for a cloud LLM provider, or a running Ollama server for local models.

Local Development

Clone the repository:

git clone https://github.com/abhinava-max/promptGit.git
cd promptGit

Create a virtual environment:

python3 -m venv .venv
source .venv/bin/activate

Install in editable mode:

python3 -m pip install -e .

Run from the installed console command:

promptgitx --help

Run directly from source:

python3 src/promptgitx/main.py --help

Project Links

• Source: https://github.com/abhinava-max/promptGit
• Issues: https://github.com/abhinava-max/promptGit/issues
• PyPI: https://pypi.org/project/promptgitx/

Status

PromptGitX is in early development. The review workflow and report exporters are functional, and the CLI surface may still evolve between minor versions.

License

MIT