SpecterAi 💼

SpecterAi is an intelligent, AI-powered git commit message generator. It analyzes your staged changes and uses Ollama (running locally) to generate Conventional Commit messages.

> Why SpecterAi?
>
> * Privacy Focused: Your code never leaves your machine.
> * Fast: Zero network latency for API calls.
> * Smart: Infers scopes, summarizes changes, and follows conventional standards.

---

🚀 Getting Started

$3

SpecterAi relies on Ollama to run Large Language Models (LLMs) or Small Language Models (SLMs) locally.

1. Download Ollama: Visit ollama.ai and download the version for your OS.
2. Pull a Model: Open your terminal and pull a lightweight but capable model like llama3.2 or mistral. Recommended model is llama3.2.

``bash
ollama pull llama3.2
`
3. Verify: Ensure Ollama is running.
`bash
ollama list
# You should see 'llama3.2' in the list
`

$3

You can install SpecterAi globally using npm, bun, or yarn (npx is also supported).

`bash


Using npm

npm install -g specterai

Using bun

bun add -g specterai

Using yarn

yarn global add specterai

Using npx if you just want to try it out

npx specterai
`

---

⚡ Efficient Usage

$3

For the most efficient workflow, let SpecterAi handle your commits automatically.

1. Install the hook (inside any git repository):
`bash
specterai install-hook
`

2. Commit as usual:
`bash
git add .
git commit
`

SpecterAi will intercept the commit, analyze your changes, and pre-fill your commit message editor with a generated suggestion. You can just save and close to accept, or edit it if needed.

$3

If you prefer to generate messages on demand:

`bash
git add .
specterai
`

You will enter an interactive mode:

`text
Proposed Commit Message:
------------------------
feat(auth): keycloak integration for sso
------------------------

[C]ommit, [E]dit, [R]egenerate, [Q]uit?
`

* [C]ommit: Approve and commit immediately.
* [E]dit: Open your default $EDITOR to tweak the message.
* [R]egenerate: Ask AI to try again (maybe it missed a detail).
* [Q]uit: Exit without committing.

---

🛠 Configuration

Customize SpecterAi to fit your workflow by creating a .specterrc.json file in your home directory (~/.specterrc.json) or project root.

Example Config:

`json
{
"model": "llama3.2",
"maxLength": 50,
"maxDiffLength": 4000,
"generateScopedCommit": true
}
`

* model: The Ollama model to use (default: llama3.2).
* maxLength: Max characters for the commit subject line.
* maxDiffLength: Truncate very large diffs to avoid context limit errors.
* generateScopedCommit: Try to infer scopes like feat(ui): instead of just feat:.

View your current effective config:
`bash
specterai config show
`

---

❓ FAQ

Q: I get "Connection refused" errors.

A: Make sure Ollama is running (ollama serve or check your menu bar).

Q: The generated messages are tailored to the wrong context.

A: Try a larger model like llama3.2 or mistral if your machine can handle it. config it in .specterrc.json`.

---

License

MIT