Data

Tools

Indexes

Applications

Experiments

Awesome

An open API service indexing awesome lists of open source software.

https://github.com/brbousnguar/ai-git-flow-cli

AI-powered scripts for generating commit messages and release notes using local Ollama — includes a Next.js web app
https://github.com/brbousnguar/ai-git-flow-cli

ai automation git local-ai nextjs ollama

Last synced: about 1 month ago
JSON representation

AI-powered scripts for generating commit messages and release notes using local Ollama — includes a Next.js web app

Awesome Lists containing this project

README 

          
# AI Local Git Flow



CLI tools for AI-assisted Git workflows in Git Bash and PowerShell.



## Commands



The intended setup is through aliases in `~/.bashrc`:



```bash

alias ai-commit='node /d/Projects/RocheBB/Tools/ai-local-git-flow/bin/ai-commit.js'

alias ai-release='node /d/Projects/RocheBB/Tools/ai-local-git-flow/bin/ai-release.js'

```



Reload Git Bash after editing `~/.bashrc`, or run:



```bash

source ~/.bashrc

```



You can also run the tools from this repository:



```bash

npm run ai-commit --

npm run ai-release --

```



## PowerShell setup



From this repository, register the local CLI package once:



```powershell

npm link

```



This creates `ai-commit.ps1` and `ai-release.ps1` shims in your npm user command folder, for example:



```powershell

C:\Users\\AppData\Roaming\npm

```



Make sure that folder is on your PowerShell `PATH`, then restart PowerShell or open a new terminal. After that, run the commands from any Git repository:



```powershell

ai-commit --yes -t SFSC-1573 -m "implement SSO authentication"

ai-release

```



If your execution policy blocks npm `.ps1` shims, run the CLI through npm from this repo instead:



```powershell

npm run ai-commit -- --yes -t SFSC-1573 -m "implement SSO authentication"

npm run ai-release --

```



## Install



```bash

npm install

```



## Configuration



Configuration is loaded from `config.json`.



For OpenAI cloud usage, create `.env` next to the scripts:



```env

OPENAI_API_KEY=sk-your-key

```



Optional JIRA enrichment for ticket-aware commit generation:



```env

JIRA_BASE_URL=https://your-company.atlassian.net

JIRA_EMAIL=your.email@company.com

JIRA_API_TOKEN=your-token

```



Use `config.json` to switch providers:



```json

{

  "provider": "cloud"

}

```



or:



```json

{

  "provider": "local",

  "local": {

    "default": "qwen2.5-coder:14b",

    "baseURL": "http://localhost:11434/v1",

    "endpoints": [

      {

        "name": "mac-mini-m4-pro",

        "baseURL": "http://192.168.1.95:11434/v1",

        "default": "qwen3-coder:30b-64k",

        "models": {

          "devstral:24b-64k": "Mac Mini model for complex implementation and reasoning",

          "qwen3-coder:30b-64k": "Mac Mini default for coding, diffs, and workflow generation"

        }

      },

      {

        "name": "local",

        "baseURL": "http://localhost:11434/v1",

        "default": "qwen2.5-coder:14b"

      }

    ]

  }

}

```



When `local.endpoints` is configured, the CLI tries them in order. The Mac Mini Ollama server is checked first, then the local machine is used as fallback. Each endpoint can keep its own default model list, so Mac Mini models do not need to match local models.



You can override the configured AI runtime for a single CLI run:



```bash

ai-commit --cloud

ai-commit --ollama-auto

ai-commit --local-ollama

ai-commit --hosted-ollama

ai-commit --provider cloud --model gpt-5.1-codex-mini

ai-commit --ollama-url http://localhost:11434/v1 --model qwen2.5-coder:14b

```



Runtime options are supported by both `ai-commit` and `ai-release`:



- `--provider `: override `config.json` provider selection

- `--cloud`: use OpenAI cloud for this run

- `--ollama-auto`: use configured Ollama endpoints and fallback order, for example Mac Mini first and this machine second

- `--local-ollama`: use only this machine's localhost Ollama endpoint, skipping Mac Mini

- `--hosted-ollama`: use only non-localhost Ollama endpoints from `local.endpoints`

- `--model `: override the configured model for this run

- `--ollama-url `: use one explicit Ollama-compatible `/v1` base URL for this run



## ai-commit



Generates branch-name variants, commit-message variants, GitHub labels, commits staged changes, pushes the branch, and creates a PR.



Stage files first:



```bash

git add 

```



Run the workflow:



```bash

ai-commit

```



Common options:



```bash

ai-commit -m "new feature, add customer request endpoint"

ai-commit -t SFSC-1573 -m "new feature, implement SSO authentication"

ai-commit -l "bug,enhancement" -m "hotfix, critical security patch"

ai-commit -m "update RAML customer request type" -n bug

ai-commit --yes -t SFSC-1573 -m "implement SSO authentication"

ai-commit --dry-run --local-ollama

ai-commit --dry-run --local-ollama --stream --progress

```



Options:



- `-t, --ticket `: ticket or issue number, for example `SFSC-1573`

- `-m, --message `: context for branch naming and commit message generation

- `-l, --labels `: comma-separated GitHub labels

- `-n, --exclude-label `: exclude a label from suggestions and PR creation

- `-d, --debug`: print LLM request details

- `--debug-context`: print context windows used for generation

- `--dry-run`: generate and select branch/commit values without renaming the branch, committing, pushing, or creating a PR

- `--progress`: print heartbeat messages while waiting for local model calls

- `--stream`: stream raw local Ollama output tokens while they are generated; also enables `--progress`

- `--think`: enable local Ollama thinking mode; by default the CLI disables thinking for faster structured branch and commit generation

- `-y, --yes`: non-interactive mode; choose variant 1 for the branch and commit message



When `-m, --message` is provided, that developer context is treated as the highest-priority intent. JIRA and diff context are skipped for generation so naming and commit wording stay aligned with the supplied message.



For AI agents such as Codex or Claude, prefer staging the intended files first and then running:



```bash

npm run ai-commit -- --yes -t SFSC-1573 -m "short implementation context"

```



The `--yes` flag keeps the workflow non-interactive while still using this CLI for branch naming, commit messages, labels, push, and PR creation.



Branch and commit generation use split context windows:



- Branch names use `prompts/branch-naming.md` plus JIRA ticket context, unless `-m` is provided. If neither `-m` nor JIRA context is available, branch names are inferred from the staged git diff.

- Commit messages use `prompts/commit-message.md` plus the staged git diff, unless `-m` is provided.



When a JIRA ticket is provided, the ticket type is used to correct GitHub labels: Task/Tache maps to `enhancement`, and Bug maps to `bug`. Explicit `-l, --labels` values still take priority.



## ai-release



Generates release PR title and notes by comparing the development branch with the production branch.



```bash

ai-release

ai-release -v v1.1.23

ai-release -l "release,enhancement"

ai-release --debug

```



The script auto-detects:



- production branch: `main` or `master`

- development branch: `develop` or `dev`



It fetches remotes, summarizes commit messages grouped by ticket ID, then can create the release PR through GitHub CLI.



Use `-d, --debug` to print release LLM request details.



## Automatic GitHub Releases



When a pull request from `develop` into `main` is merged, GitHub Actions creates the next patch tag and a GitHub Release with generated release notes.



Examples:



- No existing tags: creates `v1.0.0`

- Latest tag `v1.0.0`: creates `v1.0.1`

- Latest tag `v1.2.9`: creates `v1.2.10`



The workflow can also be run manually from the GitHub Actions tab.



## Requirements



- Node.js 18+

- Git

- GitHub CLI (`gh`) authenticated for PR creation

- OpenAI API key or local Ollama-compatible endpoint



## Runtime



- Node.js ES modules

- OpenAI SDK

- `dotenv`

- Git CLI

- GitHub CLI (`gh`)

- Optional JIRA REST enrichment for commit generation through `.env`



## Project Files



- `bin/ai-commit.js`: commit workflow CLI

- `bin/ai-release.js`: release PR workflow CLI

- `src/ai-common.js`: shared config, OpenAI/Ollama client, token usage, console formatting

- `config.json`: provider/model/pricing configuration

- `prompts/branch-naming.md`: branch naming rules used in prompts

- `prompts/commit-message.md`: commit message rules used in prompts