AI-Native Binariesv1.1.0

🧠gitaiflow turns code changes into structured intelligence.

Understand the change once. Preserve that understanding. Reuse it everywhere.

Most AI tools generate a commit message and stop there — useful for a moment, then gone. gitaiflow reads your actual git diff and keeps what it learns: a structured, timestamped artifact that can become your commit, your changelog, your release notes, and input for the next tool in line — using whatever AI provider you already trust, from a free model to one running entirely on your own machine.

Any

AI Provider (BYOM)

Free-Model Cost Guard

Platform Binaries

AOT

Ahead-of-Time Compiled

View Source Documentation

gitaiflow.sh

Live

$gitaiflow --change-summary

AOT compiled · no runtime required · runs anywhere

What's New

Why It Exists

AI-generated commit messages are disposable — the understanding behind the change is usually thrown away instead of reused.
Provider lock-in — most AI commit tools tie you to one vendor’s API instead of letting you bring your own free or paid model.
Large diffs blow past provider context limits and free-tier token quotas when sent as a single request.

Git gives you facts. AI gives you meaning.

What git already knows

Which files changed
Who changed them, and when
The branch and its base
Added, modified, or removed

What AI adds on top

Why the change matters
A scannable summary
Grouping into a real narrative
Language another tool can consume

One structured change artifact

Why gitaiflow

One artifact, every workflow: Every run produces a timestamped JSON artifact — repo, branch, base, author, files, and the AI summary. It’s the source of truth; Markdown, commits, and changelogs are just views of it.
Bring your own AI: Gemini, OpenAI, OpenRouter, Ollama, Groq, DeepSeek, vLLM, LM Studio, or any OpenAI-compatible endpoint — swap providers with an environment variable, never a rewrite.
Local-first, by choice: Point it at Ollama and gitaiflow runs with no API key and no cloud round-trip — your diff never has to leave the machine it was written on.
Cost-aware, not cost-surprised: On OpenRouter, gitaiflow checks the configured model is priced at $0/$0 before every request and refuses to proceed on anything else without your explicit say-so.
Built for real, messy repositories: Large diffs are chunked by top-level directory into their own AI calls instead of one oversized request — or disable chunking when you want a single pass.
From changes to changelogs: CHANGELOG.md entries are built from previously generated change artifacts, bucketed by conventional-commit type, with a git-log fallback when no artifacts exist yet.
Ahead-of-Time Compiled: Every release is compiled ahead-of-time (AOT) to a native executable for macOS, Linux, and Windows — no Python runtime, no bytecode, no interpreter startup cost. Download and run.

One Artifact. Every Workflow.

One structured change artifact

Commit MessageCHANGELOG.mdRelease NotesCI / AutomationAI Agents

Mandatory Configuration

gitaiflow is an AI-powered binary application designed around a Bring Your Own Key (BYOK) architecture.

gitaiflow does not require the application owner to host or operate an AI model. Instead, the developer provides the AI provider configuration that gitaiflow should use. The current AI integration contract is:

AI integration contract

AI_PROVIDER=custom
AI_BASE_URL=<endpoint>
AI_API_KEY=<key>
AI_MODEL=<model>

This architecture allows gitaiflow to work with:

Third-party cloud AI providers
OpenAI-compatible APIs
Enterprise AI endpoints
Self-hosted models
Local AI runtimes such as Ollama or LM Studio
Custom/private model endpoints

For more details, refer to the documentation.

Env Variable	Default	Notes
AI_PROVIDER	Required	Label for the provider you’re using, e.g. gemini, ollama, openai, custom. Only affects display/usage-log grouping — it does not select a base_url or model for you.
AI_API_KEY	Required	API key for your provider. Even a provider that ignores it (e.g. a local Ollama/vLLM server) still needs some non-empty value here.
AI_BASE_URL	Required	Full OpenAI-compatible chat-completions base URL, e.g. https://generativelanguage.googleapis.com/v1beta/openai
AI_MODEL	Required	Model name/id exactly as your provider expects it, e.g. gemini-flash-lite-latest, llama3.2:3b, gpt-4.1-min

Installation

Ships as a single ahead-of-time (AOT) compiled binary for every major platform — compiled straight to native machine code, so there's no interpreter, no dependency tree, and no runtime to install.

Installs the ahead-of-time (AOT) compiled binary directly — no build step, no Python interpreter required on the target machine. Omit the version to install the latest release; pass a version tag to pin an exact version instead.

Linux & macOS (bash)

curl -fsSL https://install.djangoplay.org/gitaiflow | bash

Windows (PowerShell)

irm https://install.djangoplay.org/gitaiflow.ps1 | iex

Linux & macOS — install a specific version

curl -fsSL https://install.djangoplay.org/gitaiflow | bash -s -- v1.0.0

Windows — install a specific version

$env:GITAIFLOW_VERSION = "v1.0.0"; irm https://install.djangoplay.org/gitaiflow.ps1 | iex

Upgrade

Re-running install again cleanly overwrites the existing binary — no uninstall needed first.
macOS & Linux: the installer moves the latest binary directly over /usr/local/bin/gitaiflow.
Windows: the installer overwrites gitaiflow.exe in $env:LOCALAPPDATA\Programs\gitaiflow.

CLI References

--path PATHFile or directory to summarize (default: current directory).

--remote / --base-branchOverride the comparison remote (defaults to the branch’s upstream, then origin) and base branch (defaults to the remote’s detected default branch, falling back to main).

--skip [paths...]Exclude directories, exact file paths, or file-type patterns (e.g. *.svg) from the diff.

-o, --output-dir DIRCustom root for change-summary output (default change-summary/); reuse the same value with --changelog.

--markdownAlso render a human-facing Markdown view alongside the JSON artifact.

--no-chunkSend the whole diff as a single AI request instead of per-directory chunks.

--change-summaryPrint just the generated commit title + body to stdout (pipeable into git commit -F).

--last-summaryPrint the most recently generated change-summary straight from disk — no git diff, no AI call, no new run.

--regenerate-summaryForce a fresh AI summary for every group, even if an identical diff was already summarized on a previous day.

--show-analysisWith --change-summary, also print the model’s raw reasoning that preceded its answer (debugging).

--allow-billingSkip the free-model pricing guard and proceed even if the configured model may be billed.

--changelog --since ... [--until ...]Generate/prepend a CHANGELOG.md entry for a date range, grouped by conventional-commit type.

--release-notesGenerate or overwrite RELEASE_NOTES.md from the current version’s CHANGELOG.md entry (run --changelog first if it doesn’t exist yet).

--list-models [--free-only] [--json]List OpenRouter’s live model catalog; no API key required.

--usage [version | --since --until]Query the local usage log for the last run, a specific version, or a custom date range.

--telemetry enable|disable|status|historySet or inspect the durable, locally saved telemetry consent, independent of the current session.

--uninstallCompletely remove the gitaiflow binary and local state files.

Security & Privacy

Telemetry is off by default and asked once per machine, at an interactive terminal only; a bare Enter, Ctrl-C, or EOF is treated as no.
When enabled, only install id, event/timestamp, version, provider/model, target type, file count, token estimates, duration, success, and OS are sent — never repository name, file paths, file contents, diff content, git author/branch, commit messages, or the AI-generated summary text.
Secret redaction (.env, *_API_KEY, *_SECRET_KEY, etc.) on the diff is best-effort pattern matching — always review generated summaries before sharing them outside your team.

Understand the change once.

Preserve that understanding.

Reuse it everywhere.

AI-powered. Provider-independent. Local-first. Built for automation.

View Source