Skip to content

refactor: 重构 provider 架构并完善工程基础设施 #44

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
MaxQian888 wants to merge 1 commit into
base: main
Choose a base branch
from
Open

refactor: 重构 provider 架构并完善工程基础设施 #44

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
18 changes: 0 additions & 18 deletions .eslintrc.json
View file
Open in desktop

This file was deleted.

30 changes: 30 additions & 0 deletions .github/workflows/ci.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,30 @@
name: CI

on:
push:
branches: [main]
pull_request:
branches: [main]

jobs:
verify:
runs-on: ubuntu-latest
strategy:
fail-fast: false
matrix:
node: ['18.x', '20.x']
steps:
- uses: actions/checkout@v4
- uses: actions/setup-node@v4
with:
node-version: ${{ matrix.node }}
cache: npm
- run: npm install --no-audit --no-fund
- name: Lint
run: npm run lint
- name: Type-check (strict)
run: npm run typecheck
- name: Unit tests
run: npm run test:unit
- name: Build extension bundle
run: npm run build
9 changes: 8 additions & 1 deletion .vscodeignore
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,10 +1,17 @@
.vscode/**
.vscode-test/**
.github/**
src/**
test/**
coverage/**
out/**
.gitignore
.npmrc
.prettierrc
.yarnrc
vsc-extension-quickstart.md
**/tsconfig.json
**/.eslintrc.json
**/eslint.config.mjs
**/vitest.config.ts
**/*.map
**/*.ts
41 changes: 41 additions & 0 deletions CHANGELOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,41 @@
# Changelog

All notable changes to **AI Commit** are documented here.

## Unreleased

### Added

- **Pluggable Provider architecture** — `LLMProvider` interface, `AbstractLLMProvider` base class, and factory dispatch. Adding a new provider is now a single new file plus one factory entry.
- **Ollama Provider** — first-class local model support via `http://localhost:11434/v1`. Configurable model, temperature, and base URL.
- **OpenAI-compatible presets** — new command `AI Commit: Use OpenAI-Compatible Preset` for DeepSeek, Zhipu GLM, Qwen (DashScope), Groq, and OpenRouter (one-click `OPENAI_BASE_URL` + recommended model + API key prompt).
- **Streaming output** — generation streams into the SCM input box character-by-character. Toggle with `STREAMING_ENABLED` (default on).
- **Real cancellation** — the progress notification's Cancel button now aborts the in-flight API request via `AbortController` bridged to `CancellationToken`.
- **SecretStorage migration** — API keys are stored in VS Code's OS-backed SecretStorage. One-time migration of existing settings runs on activation.
- **New command `AI Commit: Set API Key`** — interactive picker + password input.
- **Diff processor** — built-in exclusion of lock files (`package-lock.json`, `pnpm-lock.yaml`, `yarn.lock`, `Cargo.lock`, `Pipfile.lock`, `poetry.lock`, `composer.lock`, `Gemfile.lock`, `go.sum`, `bun.lockb`), `dist/`, `build/`, `out/`, `.next/`, `node_modules/`, and `*.min.{js,css,map}`. Token-budgeted truncation prevents context overruns. Customizable via `DIFF_MAX_TOKENS`, `DIFF_EXCLUDE_PATTERNS`, `DIFF_INCLUDE_DEFAULT_EXCLUDES`.
- **Unit test suite** — `vitest` with 36 tests covering providers, diff processor, presets, redaction, and the `<think>` cleaner.
- **CI workflow** — GitHub Actions matrix (Node 18.x, 20.x) running lint, typecheck, unit tests, and build on push/PR.
- **TypeScript strict mode** — `strict: true`, `noImplicitReturns`, `noFallthroughCasesInSwitch`, `forceConsistentCasingInFileNames`.
- **Logger secret redaction** — common API key prefixes (`sk-`, `sk-ant-`, `AIza`, `gsk_`, `xai-`) are masked in Output channel logs.

### Fixed

- **Gemini system prompt loss** — previously the system message was flattened into `sendMessage` and silently dropped. Now passed as `systemInstruction` per the official Google SDK contract. Regression test added.
- **Bogus null return value in `git-utils`** — `error: null` (which conflicted with the `error?: string` signature) is removed; success path now omits `error` entirely.
- **Conflicting `resolutions["@types/node"]: 16.x`** — removed; the project's `devDependencies` `@types/node: 25.x` now takes effect.
- **Empty OpenAI response handling** — non-null assertion replaced with an explicit empty-response guard.
- **Retry loop** — `Failed` modal now caps Retry attempts at 3 per command, then prompts the user to set the API key or open settings.
- **Activation event** — switched from the never-triggered `onCommand:ai-commit` to `onStartupFinished` so the SCM button is active when the user opens a repo.

### Changed

- `tsconfig` target & lib bumped to `ES2022` (required for `Error({ cause })`).
- ESLint migrated from `.eslintrc.json` to flat `eslint.config.mjs` (ESLint v10 compatibility).
- `engines.node` raised to `>= 18`.
- `Error` instances thrown from provider failures now attach the original error as `cause`.
- The deprecated settings-based API key fields (`OPENAI_API_KEY`, `CLAUDE_API_KEY`, `GEMINI_API_KEY`) are kept for back-compat but marked deprecated in descriptions.

## 0.1.2

Last published version under the original architecture. See git history.
114 changes: 75 additions & 39 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,58 +27,82 @@ Use OpenAI / Azure OpenAI / DeepSeek / Grok / Gemini / Claude (Anthropic) API to

## ✨ Features

- 🤯 Support generating commit messages based on git diffs using OpenAI / Azure OpenAI / DeepSeek / Grok / Gemini / Claude (Anthropic) API.
- 🧠 Support OpenAI Responses API with configurable reasoning effort and output verbosity.
- 🗺️ Support multi-language commit messages.
- 😜 Support adding Gitmoji.
- 🛠️ Support custom system prompt.
- 📝 Support Conventional Commits specification.
- 🔌 **Pluggable providers** — OpenAI (Chat Completions & Responses API), Anthropic Claude, Google Gemini, and **local Ollama** out of the box.
- 🚀 **OpenAI-compatible presets** — one click to switch to DeepSeek / Zhipu GLM / Qwen (DashScope) / Groq / OpenRouter via `AI Commit: Use OpenAI-Compatible Preset`.
- 📡 **Streaming output** — commit message is written into the SCM input box character-by-character as it generates.
- ✋ **Real cancellation** — pressing the progress notification's Cancel button truly aborts the in-flight API request (`AbortController`).
- 🔐 **SecretStorage** — API keys live in VS Code's OS-backed secret store, not your `settings.json`. Existing settings are migrated automatically.
- 🧹 **Smart diff** — lock files, `dist/`, `build/`, `.min.*`, and other generated files are excluded by default; oversized diffs are auto-truncated to a configurable token budget.
- 🧠 **Responses API** — configurable reasoning effort and output verbosity.
- 🌐 **19 languages** for commit message output.
- 😜 **Gitmoji** + Conventional Commits, plus custom system prompts.

## 📦 Installation

1. Search for "AI Commit" in VSCode and click the "Install" button.
2. Install it directly from the [Visual Studio Code Marketplace](https://marketplace.visualstudio.com/items?itemName=Sitoi.ai-commit).
1. Search for "AI Commit" in VSCode and click "Install".
2. Or install from the [Visual Studio Code Marketplace](https://marketplace.visualstudio.com/items?itemName=Sitoi.ai-commit).

> **Note**\
> Make sure your node version >= 16
> **Note**: Requires Node.js >= 18 and VS Code >= 1.77.

## 🤯 Usage

1. Ensure that you have installed and enabled the "AI Commit" extension.
2. In VSCode settings, locate the "ai-commit" configuration options and configure them as needed.
3. Make changes in your project and add the changes to the staging area (git add).
4. (Optional) If you want to provide additional context for the commit message, type it in the Source Control panel's message input box before clicking the AI Commit button.
5. Next to the commit message input box in the "Source Control" panel, click the "AI Commit" icon button. After clicking, the extension will generate a commit message (considering any additional context if provided) and populate it in the input box.
6. Review the generated commit message, and if you are satisfied, proceed to commit your changes.
1. Run `AI Commit: Set API Key` from the Command Palette and pick a provider. The key is stored in SecretStorage.
2. (Optional) For non-OpenAI vendors that share the OpenAI protocol, run `AI Commit: Use OpenAI-Compatible Preset` to pre-fill the base URL and a recommended model.
3. Stage some changes (`git add ...`).
4. Click the **AI Commit** icon in the Source Control panel header.
5. Watch the commit message stream into the input box. Review, edit if needed, commit.

> **Note**\
> If the code exceeds the maximum token length, consider adding it to the staging area in batches.
## 🛠️ Commands

| Command | Purpose |
| --- | --- |
| `AI Commit` (SCM title button) | Generate a commit message from staged changes |
| `AI Commit: Set API Key` | Store an OpenAI/Claude/Gemini key in SecretStorage |
| `AI Commit: Use OpenAI-Compatible Preset` | Pre-configure DeepSeek/Zhipu/Qwen/Groq/OpenRouter |
| `AI Commit: Show Available OpenAI Models` | Pick from `/v1/models` |

## 🧪 Diff handling

| Built-in exclude | Pattern |
| --- | --- |
| Lock files | `package-lock.json`, `pnpm-lock.yaml`, `yarn.lock`, `Cargo.lock`, `Pipfile.lock`, `poetry.lock`, `composer.lock`, `Gemfile.lock`, `go.sum`, `bun.lockb` |
| Generated output | `dist/`, `build/`, `out/`, `.next/`, `node_modules/` |
| Minified | `*.min.{js,css,map}` |

Customize via `ai-commit.DIFF_EXCLUDE_PATTERNS` (additional regexes) and `ai-commit.DIFF_INCLUDE_DEFAULT_EXCLUDES` (toggle defaults). Token budget is controlled by `ai-commit.DIFF_MAX_TOKENS` (default `8000`).

### ⚙️ Configuration

> **Note** Version >= 0.0.5 Don't need to configure `EMOJI_ENABLED` and `FULL_GITMOJI_SPEC`, Default Prompt is [prompt/with_gitmoji.md](./prompt/with_gitmoji.md), If don't need to use `Gitmoji`. Please set `SYSTEM_PROMPT` to your custom prompt, please refer to [prompt/without_gitmoji.md](./prompt/without_gitmoji.md).
> The Gitmoji-enabled prompt is built in; to disable Gitmoji or customize the output format, paste a custom prompt into `AI_COMMIT_SYSTEM_PROMPT` (see [prompt/with_gitmoji.md](./prompt/with_gitmoji.md) and [prompt/without_gitmoji.md](./prompt/without_gitmoji.md) for reference templates).

In the VSCode settings, locate the "ai-commit" configuration options and configure them as needed:

| Configuration | Type | Default | Required | Notes |
| :---------------------- | :----: | :------------------------: | :------: | :--------------------------------------------------------------------------------------------------------------------------------: |
| AI_PROVIDER | string | openai | Yes | AI Provider to use: `openai`, `gemini`, or `claude` |
| OPENAI_API_KEY | string | None | Yes | Required when `AI_PROVIDER` is `openai`. [OpenAI token](https://platform.openai.com/account/api-keys) |
| OPENAI_BASE_URL | string | None | No | If using Azure, use: `https://{resource}.openai.azure.com/openai/deployments/{model}` |
| OPENAI_MODEL | string | gpt-4o | Yes | OpenAI model. Run the `Show Available OpenAI Models` command to pick from available models |
| AZURE_API_VERSION | string | None | No | Azure API version string |
| OPENAI_TEMPERATURE | number | 0.7 | No | Controls randomness. Range: 0–2. Lower = more focused, Higher = more creative. (Chat Completions only) |
| OPENAI_API_TYPE | string | completion | No | Choose API: `completion` (Chat Completions) or `response` (Responses API) |
| OPENAI_REASONING_EFFORT | string | medium | No | Reasoning effort for Responses API: `minimal`, `low`, `medium`, `high`. Only applies when `OPENAI_API_TYPE` is `response` |
| OPENAI_TEXT_VERBOSITY | string | medium | No | Output verbosity for Responses API: `low` (~1000 tokens), `medium` (~4000 tokens), `high` (~16000 tokens) |
| GEMINI_API_KEY | string | None | Yes | Required when `AI_PROVIDER` is `gemini`. [Gemini API key](https://makersuite.google.com/app/apikey) |
| GEMINI_MODEL | string | gemini-2.0-flash-001 | Yes | Gemini model to use |
| GEMINI_TEMPERATURE | number | 0.7 | No | Controls randomness. Range: 0–2. Lower = more focused, Higher = more creative |
| CLAUDE_API_KEY | string | None | No | Anthropic API key. Leave empty to use Claude CLI (authenticated via `claude setup-token`). Required when `AI_PROVIDER` is `claude` |
| CLAUDE_MODEL | string | claude-sonnet-4-5-20250929 | No | Claude model to use |
| CLAUDE_TEMPERATURE | number | 0.7 | No | Controls randomness. Range: 0–1 |
| AI_COMMIT_LANGUAGE | string | English | Yes | Supports 19 languages |
| SYSTEM_PROMPT | string | None | No | Custom system prompt |
| Configuration | Type | Default | Notes |
| ----------------------------------- | --------- | ----------------------------- | ----- |
| `AI_PROVIDER` | string | `openai` | `openai` / `gemini` / `claude` / `ollama` |
| `OPENAI_API_KEY` | string | `""` | Deprecated — prefer `AI Commit: Set API Key` (SecretStorage) |
| `OPENAI_BASE_URL` | string | `""` | Azure or OpenAI-compatible vendor (DeepSeek/Zhipu/Qwen/Groq/OpenRouter). Use the preset command for one-click setup |
| `OPENAI_MODEL` | string | `gpt-4o` | Run `AI Commit: Show Available OpenAI Models` to discover |
| `AZURE_API_VERSION` | string | `""` | Azure API version |
| `OPENAI_TEMPERATURE` | number | `0.7` | 0–2; Chat Completions only |
| `OPENAI_API_TYPE` | enum | `completion` | `completion` or `response` (Responses API) |
| `OPENAI_REASONING_EFFORT` | enum | `medium` | `minimal`/`low`/`medium`/`high` (Responses API) |
| `OPENAI_TEXT_VERBOSITY` | enum | `medium` | Maps to max output tokens (Responses API) |
| `GEMINI_API_KEY` | string | `""` | Deprecated — use `AI Commit: Set API Key` |
| `GEMINI_MODEL` | string | `gemini-2.0-flash-001` | |
| `GEMINI_TEMPERATURE` | number | `0.7` | 0–2 |
| `CLAUDE_API_KEY` | string | `""` | Deprecated — use `AI Commit: Set API Key` |
| `CLAUDE_MODEL` | string | `claude-sonnet-4-5-20250929` | |
| `CLAUDE_TEMPERATURE` | number | `0.7` | 0–1 |
| `OLLAMA_BASE_URL` | string | `http://localhost:11434/v1` | Local Ollama OpenAI-compatible endpoint |
| `OLLAMA_MODEL` | string | `llama3.2` | Any pulled local model |
| `OLLAMA_TEMPERATURE` | number | `0.7` | 0–2 |
| `STREAMING_ENABLED` | boolean | `true` | Stream output into the SCM input box |
| `DIFF_MAX_TOKENS` | number | `8000` | Approx token budget for the diff |
| `DIFF_EXCLUDE_PATTERNS` | string[] | `[]` | Extra regex patterns to exclude (added to defaults) |
| `DIFF_INCLUDE_DEFAULT_EXCLUDES` | boolean | `true` | Toggle the built-in exclude list |
| `AI_COMMIT_LANGUAGE` | enum | `English` | 19 supported languages |
| `AI_COMMIT_SYSTEM_PROMPT` | string | `""` | Custom system prompt that overrides the default |

## ⌨️ Local Development

Expand All @@ -92,9 +116,21 @@ Alternatively, you can clone the repository and run the following commands for l
$ git clone https://github.com/sitoi/ai-commit.git
$ cd ai-commit
$ npm install
$ npm run verify # typecheck + unit tests
$ npm run build # webpack production bundle
```

Open the project folder in VSCode. Press F5 to run the project. This will open a new Extension Development Host window and launch the plugin within it.
Open the project folder in VSCode. Press F5 to run the project. This opens a new Extension Development Host window with the plugin loaded.

### Running tests

```bash
$ npm run test:unit # vitest run
$ npm run test:unit:watch # watch mode
$ npm run test:coverage # with v8 coverage report
$ npm run lint # eslint flat config
$ npm run typecheck # strict tsc
```

## 🤝 Contributing

Expand Down
Loading