git-coco 0.2.0 → 0.3.0

package/README.md

@@ -1,6 +1,14 @@
 # `coco` 🤖 🦍
 
-Commit Copilot, or `coco`, is your personal scribe for git commit messages. Using [LangChain🦜🔗](https://js.langchain.com/) to automate the task of creating meaningful commit messages based on your staged changes!
+[![GitHub issues](https://img.shields.io/github/issues/gfargo/coco)](https://github.com/gfargo/coco/issues)
+[![GitHub pull requests](https://img.shields.io/github/issues-pr/gfargo/coco)](https://github.com/gfargo/coco/pulls)
+[![Last Commit](https://img.shields.io/github/last-commit/gfargo/coco)](https://github.com/gfargo/coco/tree/main)
+[![NPM Version](https://img.shields.io/npm/v/git-coco.svg)](https://www.npmjs.com/package/git-coco)
+[![NPM Downloads](https://img.shields.io/npm/dt/git-coco.svg)](https://www.npmjs.com/package/git-coco)
+
+Commit Copilot, or `coco`, is your personal scribe for git commit messages. Leveraging the power of [LangChain🦜🔗](https://js.langchain.com/) and LLMs to encapsulate your staged changes into meaningful commit messages!
+
+
 
 ## Installation
 
@@ -105,7 +113,7 @@ Remember, command line flags and environment variables should be defined in `UPP
 - [x] LangChain integration 🦜
 - [ ] Additional tests! 🧪
 - [ ] Conventional commits 🔜
-- [ ] HuggingFace integration 🔜
+- [x] HuggingFace integration 🔜
 - [ ] Google Vertex AI integration (?)
 - [ ] Automatic changelog generation 🫣
 - [ ] Rebase support 🔀

package/dist/index.esm.mjs

@@ -14,11 +14,13 @@ import ora from 'ora';
 import now from 'performance-now';
 import prettyMilliseconds from 'pretty-ms';
 import { Document } from 'langchain/document';
+import { HuggingFaceInference } from 'langchain/llms/hf';
 import { loadSummarizationChain, LLMChain } from 'langchain/chains';
 import { OpenAI } from 'langchain/llms/openai';
 import { RecursiveCharacterTextSplitter } from 'langchain/text_splitter';
 import { createTwoFilesPatch } from 'diff';
 import GPT3NodeTokenizer from 'gpt3-tokenizer';
+import { minimatch } from 'minimatch';
 import { simpleGit } from 'simple-git';
 
 /**
@@ -39,7 +41,9 @@ function removeUndefined(obj) {
  **/
 function loadEnvConfig(config) {
     const envConfig = {
+        model: process.env.COCO_MODEL || undefined,
         openAIApiKey: process.env.OPENAI_API_KEY || undefined,
+        huggingFaceHubApiKey: process.env.HUGGINGFACE_HUB_API_KEY || undefined,
         tokenLimit: process.env.COCO_TOKEN_LIMIT
             ? parseInt(process.env.COCO_TOKEN_LIMIT)
             : undefined,
@@ -70,7 +74,9 @@ function loadGitConfig(config) {
         const gitConfigParsed = ini.parse(gitConfigRaw);
         config = {
             ...config,
+            model: gitConfigParsed.coco?.model || config.model,
             openAIApiKey: gitConfigParsed.coco?.openAIApiKey || config.openAIApiKey,
+            huggingFaceHubApiKey: gitConfigParsed.coco?.huggingFaceHubApiKey || config.huggingFaceHubApiKey,
             tokenLimit: parseInt(gitConfigParsed.coco?.tokenLimit) || config.tokenLimit,
             prompt: gitConfigParsed.coco?.prompt || config.prompt,
             mode: gitConfigParsed.coco?.mode || config.mode,
@@ -150,7 +156,9 @@ function loadXDGConfig(config) {
  * Command line options via yargs
  */
 const options = {
+    model: { type: 'string', description: 'LLM/Model-Name' },
     openAIApiKey: { type: 'string', description: 'OpenAI API Key' },
+    huggingFaceHubApiKey: { type: 'string', description: 'HuggingFace Hub API Key' },
     tokenLimit: { type: 'number', description: 'Token limit' },
     prompt: {
         type: 'string',
@@ -247,7 +255,7 @@ const SUMMARIZE_PROMPT = new PromptTemplate({
  * @type {Config}
  */
 const DEFAULT_CONFIG = {
-    openAIApiKey: '',
+    model: 'openai/gpt-3.5-turbo',
     verbose: false,
     tokenLimit: 1024,
     prompt: COMMIT_PROMPT.template,
@@ -257,7 +265,6 @@ const DEFAULT_CONFIG = {
     ignoredFiles: ['package-lock.json'],
     ignoredExtensions: ['.map', '.lock'],
 };
-
 /**
  * Load application config
  *
@@ -541,18 +548,34 @@ async function collectDiffs(node, getFileDiff, tokenizer, logger = new Logger(co
     };
 }
 
-// TODO: Extend this to support other models! 🎉
+/**
+ * Get LLM Model Based on Configuration
+ *
+ * @param fields
+ * @param configuration
+ * @returns LLM Model
+ */
 function getModel(fields, configuration) {
-    return new OpenAI(fields, configuration);
-    // return new HuggingFaceInference({
-    //   // model: 'gpt2',
-    //   // model: 'bigcode/starcoder',
-    //   model: 'bigscience/bloom',
-    //   apiKey: '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',
-    //   maxConcurrency: 4,
-    //   cache: true,
-    //   // maxTokens: 2046,
-    // })
+    const [llm, model] = config.model.split(/\/(.*)/s);
+    if (!model) {
+        throw new Error(`Invalid model: ${config.model}`);
+    }
+    switch (llm) {
+        case 'huggingface':
+            return new HuggingFaceInference({
+                model: model,
+                apiKey: config.huggingFaceHubApiKey,
+                maxConcurrency: 4,
+                ...fields,
+            });
+        case 'openai':
+        default:
+            return new OpenAI({
+                openAIApiKey: config.openAIApiKey,
+                modelName: model,
+                ...fields,
+            }, configuration);
+    }
 }
 function getTextSplitter(options = {}) {
     return new RecursiveCharacterTextSplitter(options);
@@ -589,14 +612,11 @@ const parseRenamedFileDiff = async (nodeFile, git, logger) => {
     const oldFilepath = nodeFile?.oldFilepath || nodeFile.filepath;
     try {
         const [headContent, indexContent] = await Promise.all([
-            // git.diff(['HEAD', '-M', '--', oldFilepath]),
-            // git.diff(['-z', '-M', '--staged', nodeFile.filepath]),
             git.show([`HEAD:${oldFilepath}`]),
             git.show([`:${nodeFile.filepath}`]),
-            // readFile(nodeFile.filepath),
         ]);
         if (headContent !== indexContent) {
-            result = createTwoFilesPatch(oldFilepath, nodeFile.filepath, headContent, indexContent.toString(), '', '', {
+            result = createTwoFilesPatch(oldFilepath, nodeFile.filepath, headContent, indexContent, '', '', {
                 context: 3,
             });
             // remove the first 4 lines of the patch (they contain the old and new file names)
@@ -608,7 +628,6 @@ const parseRenamedFileDiff = async (nodeFile, git, logger) => {
     }
     catch (err) {
         logger.verbose(`Error comparing file contents for ${nodeFile.filepath}`, { color: 'red' });
-        console.log(err);
         result = 'Error comparing file contents.';
     }
     return result;
@@ -651,7 +670,6 @@ const fileChangeParser = async (changes, { tokenizer, git, model }) => {
         chain: summarizationChain,
     });
     logger.stopTimer(`\nSummary generated for ${changes.length} staged files`, { color: 'green' });
-    logger.verbose(`\nSummary:\n${summary}`, { color: 'blue' });
     return summary;
 };
 
@@ -686,10 +704,25 @@ const getTokenizer = () => {
 };
 
 const llm = async ({ llm, prompt, variables }) => {
+    if (!llm || !prompt || !variables) {
+        throw new Error('The input parameters "llm", "prompt", and "variables" are all required.');
+    }
     const chain = new LLMChain({ llm, prompt });
-    const res = await chain.call(variables);
-    if (res.error)
-        throw new Error(res.error);
+    let res;
+    try {
+        res = await chain.call(variables);
+    }
+    catch (error) {
+        if (error instanceof Error) {
+            throw new Error(`LLMChain call error: ${error.message}`);
+        }
+    }
+    if (!res) {
+        throw new Error('Empty response from LLMChain call');
+    }
+    if (res.error) {
+        throw new Error(`LLMChain response error: ${res.error}`);
+    }
     return res.text.trim();
 };
 
@@ -727,14 +760,8 @@ const getSummaryText = (file, change) => {
     return `${status}: ${file.path}`;
 };
 
-const DEFAULT_IGNORED_FILES = [
-    ...(config?.ignoredFiles?.length && config?.ignoredFiles?.length > 0 ? config.ignoredFiles : []),
-];
-const DEFAULT_IGNORED_EXTENSIONS = [
-    ...(config?.ignoredExtensions?.length && config?.ignoredExtensions?.length > 0
-        ? config.ignoredExtensions
-        : []),
-];
+const DEFAULT_IGNORED_FILES = config?.ignoredFiles?.length ? config.ignoredFiles : [];
+const DEFAULT_IGNORED_EXTENSIONS = config?.ignoredExtensions?.length ? config.ignoredExtensions : [];
 async function getChanges(git, options = {}) {
     const { ignoredFiles = DEFAULT_IGNORED_FILES, ignoredExtensions = DEFAULT_IGNORED_EXTENSIONS } = options;
     const staged = [];
@@ -742,7 +769,6 @@ async function getChanges(git, options = {}) {
     const untracked = [];
     const status = await git.status();
     status.files.forEach((file) => {
-        // console.log({ file })
         const fileChange = {
             filepath: file.path,
             oldFilepath: status.renamed.filter((renamed) => renamed.to === file.path)[0]?.from,
@@ -769,16 +795,20 @@ async function getChanges(git, options = {}) {
     const ignoredExtensionsSet = new Set(ignoredExtensions.map((extension) => extension.toLowerCase()));
     const filteredStaged = staged.filter((file) => {
         const extension = path__default.extname(file.filepath).toLowerCase();
-        return !ignoredExtensionsSet.has(extension) && !ignoredFiles.includes(file.filepath);
+        return !ignoredExtensionsSet.has(extension) && !ignoredFiles.some(ignoredPattern => minimatch(file.filepath, ignoredPattern));
     });
     const filteredUnstaged = unstaged.filter((file) => {
         const extension = path__default.extname(file.filepath).toLowerCase();
-        return !ignoredExtensionsSet.has(extension) && !ignoredFiles.includes(file.filepath);
+        return !ignoredExtensionsSet.has(extension) && !ignoredFiles.some(ignoredPattern => minimatch(file.filepath, ignoredPattern));
+    });
+    const filteredUntracked = untracked.filter((file) => {
+        const extension = path__default.extname(file.filepath).toLowerCase();
+        return !ignoredExtensionsSet.has(extension) && !ignoredFiles.some(ignoredPattern => minimatch(file.filepath, ignoredPattern));
     });
     return {
         staged: filteredStaged,
         unstaged: filteredUnstaged,
-        untracked,
+        untracked: filteredUntracked,
     };
 }