zencommit 0.1.0

package/src/llm/prompts/gitmoji.md

@@ -0,0 +1,212 @@
+Prefix the commit subject with exactly one gitmoji (Unicode emoji character) followed by a space.
+
+### Format
+
+When combined with conventional commits style:
+
+```
+<emoji> type(scope?): description
+```
+
+When used with freeform style:
+
+```
+<emoji> description
+```
+
+### Rules
+
+1. Use exactly ONE emoji per commit - choose the one that best represents the primary change.
+2. Use the actual Unicode emoji character, not the `:shortcode:` format.
+3. Place the emoji at the very beginning of the subject line.
+4. Add a single space between the emoji and the rest of the subject.
+5. If the change spans multiple categories, choose the emoji for the most significant aspect.
+
+### Gitmoji Reference
+
+Select the most appropriate emoji from this reference:
+
+#### Features & Enhancements
+
+| Emoji | Code                     | Use When                                                                         |
+| ----- | ------------------------ | -------------------------------------------------------------------------------- |
+| ✨    | `:sparkles:`             | Introducing a new feature                                                        |
+| 💄    | `:lipstick:`             | Adding or updating UI/style files                                                |
+| 🎨    | `:art:`                  | Improving code structure/format (not style type - this is for code organization) |
+| 🚸    | `:children_crossing:`    | Improving user experience/usability                                              |
+| 💫    | `:dizzy:`                | Adding or updating animations/transitions                                        |
+| 🥅    | `:goal_net:`             | Catching errors                                                                  |
+| 🔍️    | `:mag:`                  | Improving SEO                                                                    |
+| 🌐    | `:globe_with_meridians:` | Internationalization and localization                                            |
+| ♿️    | `:wheelchair:`           | Improving accessibility                                                          |
+| 💬    | `:speech_balloon:`       | Adding or updating text/literals                                                 |
+| 🏷️    | `:label:`                | Adding or updating types (TypeScript, Flow)                                      |
+
+#### Bug Fixes
+
+| Emoji | Code                 | Use When                            |
+| ----- | -------------------- | ----------------------------------- |
+| 🐛    | `:bug:`              | Fixing a bug                        |
+| 🚑️    | `:ambulance:`        | Critical hotfix                     |
+| 🩹    | `:adhesive_bandage:` | Simple fix for a non-critical issue |
+| 🔒️    | `:lock:`             | Fixing security issues              |
+| 🍎    | `:apple:`            | Fixing something on macOS           |
+| 🐧    | `:penguin:`          | Fixing something on Linux           |
+| 🏁    | `:checkered_flag:`   | Fixing something on Windows         |
+| 🤖    | `:robot:`            | Fixing something on Android         |
+| 🍏    | `:green_apple:`      | Fixing something on iOS             |
+
+#### Performance & Optimization
+
+| Emoji | Code            | Use When                            |
+| ----- | --------------- | ----------------------------------- |
+| ⚡️    | `:zap:`         | Improving performance               |
+| 🔥    | `:fire:`        | Removing code or files              |
+| 🗑️    | `:wastebasket:` | Deprecating code that needs cleanup |
+
+#### Documentation
+
+| Emoji | Code               | Use When                                   |
+| ----- | ------------------ | ------------------------------------------ |
+| 📝    | `:memo:`           | Adding or updating documentation           |
+| 💡    | `:bulb:`           | Adding or updating comments in source code |
+| 📄    | `:page_facing_up:` | Adding or updating license                 |
+
+#### Testing
+
+| Emoji | Code                 | Use When                           |
+| ----- | -------------------- | ---------------------------------- |
+| ✅    | `:white_check_mark:` | Adding, updating, or passing tests |
+| 🧪    | `:test_tube:`        | Adding a failing test              |
+| 🤡    | `:clown_face:`       | Mocking things                     |
+| 📸    | `:camera_flash:`     | Adding or updating snapshots       |
+
+#### Dependencies & Build
+
+| Emoji | Code                    | Use When                                      |
+| ----- | ----------------------- | --------------------------------------------- |
+| ⬆️    | `:arrow_up:`            | Upgrading dependencies                        |
+| ⬇️    | `:arrow_down:`          | Downgrading dependencies                      |
+| 📌    | `:pushpin:`             | Pinning dependencies to specific versions     |
+| ➕    | `:heavy_plus_sign:`     | Adding a dependency                           |
+| ➖    | `:heavy_minus_sign:`    | Removing a dependency                         |
+| 📦️    | `:package:`             | Adding or updating compiled files or packages |
+| 👷    | `:construction_worker:` | Adding or updating CI build system            |
+| 💚    | `:green_heart:`         | Fixing CI build                               |
+| 🔧    | `:wrench:`              | Adding or updating configuration files        |
+| 🔨    | `:hammer:`              | Adding or updating development scripts        |
+
+#### Code Quality & Refactoring
+
+| Emoji | Code             | Use When                                            |
+| ----- | ---------------- | --------------------------------------------------- |
+| ♻️    | `:recycle:`      | Refactoring code                                    |
+| 🚚    | `:truck:`        | Moving or renaming resources (files, paths, routes) |
+| ✏️    | `:pencil2:`      | Fixing typos                                        |
+| 🩺    | `:stethoscope:`  | Adding or updating health checks                    |
+| 🧱    | `:bricks:`       | Infrastructure-related changes                      |
+| 🧑‍💻    | `:technologist:` | Improving developer experience                      |
+| 💩    | `:poop:`         | Writing bad code that needs improvement             |
+| 🍱    | `:bento:`        | Adding or updating assets                           |
+
+#### Version Control & Releases
+
+| Emoji | Code                          | Use When                             |
+| ----- | ----------------------------- | ------------------------------------ |
+| 🎉    | `:tada:`                      | Beginning a project (initial commit) |
+| 🔖    | `:bookmark:`                  | Releasing/version tags               |
+| 🚀    | `:rocket:`                    | Deploying stuff                      |
+| ⏪️    | `:rewind:`                    | Reverting changes                    |
+| 🔀    | `:twisted_rightwards_arrows:` | Merging branches                     |
+
+#### Database & Data
+
+| Emoji | Code              | Use When                            |
+| ----- | ----------------- | ----------------------------------- |
+| 🗃️    | `:card_file_box:` | Performing database-related changes |
+| 🌱    | `:seedling:`      | Adding or updating seed files       |
+
+#### Security & Secrets
+
+| Emoji | Code                     | Use When                             |
+| ----- | ------------------------ | ------------------------------------ |
+| 🔐    | `:closed_lock_with_key:` | Adding or updating secrets           |
+| 🛂    | `:passport_control:`     | Working on authorization/permissions |
+
+#### Work In Progress
+
+| Emoji | Code             | Use When                      |
+| ----- | ---------------- | ----------------------------- |
+| 🚧    | `:construction:` | Work in progress              |
+| 🙈    | `:see_no_evil:`  | Adding or updating .gitignore |
+
+#### Other
+
+| Emoji | Code            | Use When                                      |
+| ----- | --------------- | --------------------------------------------- |
+| 🍻    | `:beers:`       | Writing code drunkenly                        |
+| 🥚    | `:egg:`         | Adding or updating easter eggs                |
+| 🧵    | `:thread:`      | Adding or updating multithreading/concurrency |
+| 🦺    | `:safety_vest:` | Adding or updating validation                 |
+
+### Common Mappings to Conventional Commit Types
+
+When using both gitmoji and conventional commits, prefer these pairings:
+
+| Type       | Primary Emoji | Alternatives |
+| ---------- | ------------- | ------------ |
+| `feat`     | ✨            | 🚸 💄 🌐 ♿️  |
+| `fix`      | 🐛            | 🚑️ 🩹 🔒️     |
+| `docs`     | 📝            | 💡           |
+| `style`    | 🎨            |              |
+| `refactor` | ♻️            | 🚚 ✏️        |
+| `perf`     | ⚡️            |              |
+| `test`     | ✅            | 🧪           |
+| `build`    | 📦️            | 🔧 🔨        |
+| `ci`       | 👷            | 💚           |
+| `chore`    | 🔧            | 🙈           |
+| `revert`   | ⏪️            |              |
+
+### Examples
+
+**Feature (conventional):**
+
+```
+✨ feat(auth): add two-factor authentication
+```
+
+**Bug fix (conventional):**
+
+```
+🐛 fix(parser): handle null values in JSON input
+```
+
+**Documentation (freeform):**
+
+```
+📝 update API documentation with new endpoints
+```
+
+**Performance (conventional):**
+
+```
+⚡️ perf(db): add index for frequent queries
+```
+
+**Dependencies (conventional):**
+
+```
+⬆️ build(deps): upgrade React to v18.2
+```
+
+**Critical hotfix (conventional):**
+
+```
+🚑️ fix(auth): patch token validation vulnerability
+```
+
+**Refactor (freeform):**
+
+```
+♻️ extract common utilities into shared module
+```

package/src/llm/prompts/system.md

@@ -0,0 +1,21 @@
+You are an expert software engineer specializing in writing precise, informative git commit messages.
+
+## Your Role
+
+You analyze code diffs and generate commit messages that accurately describe the changes made. Your messages help developers understand the history of a codebase at a glance.
+
+## Core Principles
+
+1. **Accuracy over brevity**: Never fabricate or assume changes not present in the diff.
+2. **Intent over implementation**: Focus on _why_ a change was made, not _how_ (the diff shows the how).
+3. **Atomic summaries**: Treat the commit as a single logical unit, even if it touches multiple files.
+4. **Professional tone**: Write in clear, technical English appropriate for a professional codebase.
+
+## Constraints
+
+- Never mention that you are an AI, language model, or assistant.
+- Never reference the prompt, instructions, or your reasoning process.
+- Never include apologies, caveats, or meta-commentary.
+- Never use phrases like "This commit..." or "This change..." - just describe what happens.
+- Never include timestamps, author information, or ticket numbers unless explicitly in the diff context.
+- Output only the requested JSON format with no surrounding text, markdown fences, or explanation.

package/src/llm/providers.ts

@@ -0,0 +1,102 @@
+import { bedrock } from '@ai-sdk/amazon-bedrock';
+import { anthropic } from '@ai-sdk/anthropic';
+import { azure } from '@ai-sdk/azure';
+import { cerebras } from '@ai-sdk/cerebras';
+import { cohere } from '@ai-sdk/cohere';
+import { deepinfra } from '@ai-sdk/deepinfra';
+import { gateway } from '@ai-sdk/gateway';
+import { google } from '@ai-sdk/google';
+import { vertex } from '@ai-sdk/google-vertex';
+import { vertexAnthropic } from '@ai-sdk/google-vertex/anthropic';
+import { groq } from '@ai-sdk/groq';
+import { mistral } from '@ai-sdk/mistral';
+import { openai } from '@ai-sdk/openai';
+import { createOpenAICompatible } from '@ai-sdk/openai-compatible';
+import { perplexity } from '@ai-sdk/perplexity';
+import { togetherai } from '@ai-sdk/togetherai';
+import { vercel } from '@ai-sdk/vercel';
+import { xai } from '@ai-sdk/xai';
+import { openrouter } from '@openrouter/ai-sdk-provider';
+import { gitlab } from '@gitlab/gitlab-ai-provider';
+import type { LanguageModel } from 'ai';
+
+const PROVIDER_FACTORIES: Record<string, (modelName: string) => LanguageModel> = {
+  openai: (modelName) => openai(modelName),
+  anthropic: (modelName) => anthropic(modelName),
+  google: (modelName) => google(modelName),
+  'vertex-anthropic': (modelName) => vertexAnthropic(modelName),
+  xai: (modelName) => xai(modelName),
+  vercel: (modelName) => vercel(modelName),
+  azure: (modelName) => azure(modelName),
+  bedrock: (modelName) => bedrock(modelName),
+  groq: (modelName) => groq(modelName),
+  deepinfra: (modelName) => deepinfra(modelName),
+  vertex: (modelName) => vertex(modelName),
+  mistral: (modelName) => mistral(modelName),
+  togetherai: (modelName) => togetherai(modelName),
+  cohere: (modelName) => cohere(modelName),
+  cerebras: (modelName) => cerebras(modelName),
+  perplexity: (modelName) => perplexity(modelName),
+  gateway: (modelName) => gateway(modelName),
+  openrouter: (modelName) => openrouter(modelName),
+  gitlab: (modelName) => gitlab(modelName),
+};
+
+const PROVIDER_ALIASES: Record<string, string> = {
+  'azure-openai': 'azure',
+  'amazon-bedrock': 'bedrock',
+  'aws-bedrock': 'bedrock',
+  'google-vertex': 'vertex',
+  'google-vertex-ai': 'vertex',
+  'google-vertex-anthropic': 'vertex-anthropic',
+  'vertex-anthropic': 'vertex-anthropic',
+  'google-generative-ai': 'google',
+  gemini: 'google',
+  'openai-compatible': 'openai-compatible',
+  openrouter: 'openrouter',
+  'open-router': 'openrouter',
+  gitlab: 'gitlab',
+  'gitlab-ai': 'gitlab',
+  'gitlab-duo': 'gitlab',
+  'together.ai': 'togetherai',
+  'xai-grok': 'xai',
+  'vercel-ai-gateway': 'gateway',
+  'ai-gateway': 'gateway',
+};
+
+const normalizeProviderId = (provider: string): string =>
+  PROVIDER_ALIASES[provider.toLowerCase()] ?? provider.toLowerCase();
+
+export interface ProviderResolutionOptions {
+  openaiCompatible?: {
+    baseUrl?: string;
+    name?: string;
+  };
+}
+
+export const resolveLanguageModel = (
+  modelId: string,
+  options: ProviderResolutionOptions = {},
+): LanguageModel => {
+  const [rawProvider, ...rest] = modelId.split('/');
+  const modelName = rest.join('/');
+  if (!rawProvider || !modelName) {
+    throw new Error('Model id must be in the format provider/model');
+  }
+  const provider = normalizeProviderId(rawProvider);
+  if (provider === 'openai-compatible') {
+    const baseURL = options.openaiCompatible?.baseUrl ?? process.env.OPENAI_COMPATIBLE_BASE_URL;
+    if (!baseURL) {
+      throw new Error('OPENAI_COMPATIBLE_BASE_URL is required for openai-compatible models.');
+    }
+    const name =
+      options.openaiCompatible?.name ?? process.env.OPENAI_COMPATIBLE_NAME ?? 'openai-compatible';
+    const apiKey = process.env.OPENAI_COMPATIBLE_API_KEY;
+    return createOpenAICompatible({ baseURL, name, apiKey })(modelName);
+  }
+  const factory = PROVIDER_FACTORIES[provider];
+  if (!factory) {
+    throw new Error(`Unsupported model provider: ${rawProvider}`);
+  }
+  return factory(modelName);
+};

package/src/llm/tokens.test.ts

@@ -0,0 +1,22 @@
+import { describe, expect, it } from 'vitest';
+import { computeTokenBudget, countTokens, getEncodingForModel } from './tokens.js';
+
+const limits = { context: 100, input: 80, output: 40 };
+
+describe('computeTokenBudget', () => {
+  it('computes budget from limits and output cap', () => {
+    const budget = computeTokenBudget(limits, 30, 10);
+    expect(budget.outputTokens).toBe(30);
+    expect(budget.inputMaxTokens).toBe(70);
+    expect(budget.availableTokens).toBe(60);
+  });
+});
+
+describe('countTokens', () => {
+  it('counts tokens for simple text', () => {
+    const encoding = getEncodingForModel('openai/gpt-4o');
+    const tokens = countTokens('hello world', encoding);
+    encoding.free();
+    expect(tokens).toBeGreaterThan(0);
+  });
+});

package/src/llm/tokens.ts

@@ -0,0 +1,46 @@
+import { encoding_for_model, get_encoding, type Tiktoken } from '@dqbd/tiktoken';
+import type { ModelLimits } from '../metadata/types.js';
+
+export interface TokenBudget {
+  inputMaxTokens: number;
+  outputTokens: number;
+  availableTokens: number;
+  overheadTokens: number;
+}
+
+const DEFAULT_ENCODING = 'cl100k_base';
+
+export const getEncodingForModel = (modelId: string): Tiktoken => {
+  const modelName = modelId.includes('/') ? (modelId.split('/')[1] ?? modelId) : modelId;
+  try {
+    return encoding_for_model(modelName);
+  } catch {
+    return get_encoding(DEFAULT_ENCODING);
+  }
+};
+
+export const countTokens = (text: string, encoding: Tiktoken): number => {
+  if (!text) {
+    return 0;
+  }
+  return encoding.encode(text).length;
+};
+
+export const computeTokenBudget = (
+  limits: ModelLimits,
+  maxOutputTokens: number,
+  overheadTokens: number,
+): TokenBudget => {
+  const contextLimit = limits.context ?? Number.POSITIVE_INFINITY;
+  const inputLimit = limits.input ?? limits.context ?? Number.POSITIVE_INFINITY;
+  const outputLimit = limits.output ?? Number.POSITIVE_INFINITY;
+  const outputTokens = Math.min(maxOutputTokens, outputLimit);
+  const inputMaxTokens = Math.min(inputLimit, contextLimit - outputTokens);
+  const availableTokens = inputMaxTokens - overheadTokens;
+  return {
+    inputMaxTokens,
+    outputTokens,
+    availableTokens,
+    overheadTokens,
+  };
+};

package/src/llm/truncate.test.ts

@@ -0,0 +1,60 @@
+import { describe, expect, it } from 'vitest';
+import { getEncodingForModel } from './tokens.js';
+import { truncateDiffByFile, truncateDiffSmart } from './truncate.js';
+import type { DiffConfig } from '../config/types.js';
+
+const sampleDiff = `diff --git a/foo.ts b/foo.ts
+index 123..456 100644
+--- a/foo.ts
++++ b/foo.ts
+@@ -1,3 +1,4 @@
+-export const a = 1;
++export const a = 2;
++export function test() {}
+ diff --git a/bar.ts b/bar.ts
+index 111..222 100644
+--- a/bar.ts
++++ b/bar.ts
+@@ -1,3 +1,3 @@
+-export const b = 1;
++export const b = 2;
+`;
+
+const smartDiff = `diff --git foo.ts foo.ts
+--- foo.ts
++++ foo.ts
+@@ -1,3 +1,4 @@
+-export const a = 1;
++export const a = 2;
++export function test() {}
+`;
+
+const diffConfig: DiffConfig = {
+  truncateStrategy: 'smart',
+  includeFileList: true,
+  excludeGitignoreFiles: true,
+  maxFiles: 200,
+  smart: {
+    maxAddedLinesPerHunk: 1,
+    maxRemovedLinesPerHunk: 1,
+  },
+};
+
+describe('truncateDiffByFile', () => {
+  it('truncates and adds marker when budget is tight', () => {
+    const encoding = getEncodingForModel('openai/gpt-4o');
+    const result = truncateDiffByFile(sampleDiff, 20, encoding);
+    encoding.free();
+    expect(result.text).toContain('diff --git');
+    expect(result.truncated).toBe(true);
+  });
+});
+
+describe('truncateDiffSmart', () => {
+  it('includes file summary', () => {
+    const encoding = getEncodingForModel('openai/gpt-4o');
+    const result = truncateDiffSmart('M foo.ts (+1 -1)', smartDiff, 120, diffConfig, encoding);
+    encoding.free();
+    expect(result.text).toContain('File summary');
+  });
+});