Experiments with Langchain

package.json

@@ -101,10 +101,12 @@
   },
   "dependencies": {
     "@google/generative-ai": "^0.3.1",
+    "@langchain/google-genai": "^0.0.10",
     "@xenova/transformers": "^2.16.1",
     "diff": "^5.2.0",
     "elevenlabs": "^0.2.2",
     "inquirer": "^8.0.0",
+    "langchain": "^0.1.31",
     "node-record-lpcm16": "^1.0.1",
     "ollama": "^0.5.0",
     "openai": "^4.31.0",

src/experimental/langchain-test-gemini.js

@@ -0,0 +1,105 @@
+const { ChatGoogleGenerativeAI } = require('@langchain/google-genai');
+const { TextLoader } = require('langchain/document_loaders/fs/text');
+const { DirectoryLoader } = require('langchain/document_loaders/fs/directory');
+
+const {
+  createStuffDocumentsChain,
+} = require('langchain/chains/combine_documents');
+const { ChatPromptTemplate } = require('@langchain/core/prompts');
+const { RecursiveCharacterTextSplitter } = require('langchain/text_splitter');
+const { MemoryVectorStore } = require('langchain/vectorstores/memory');
+const { createRetrievalChain } = require('langchain/chains/retrieval');
+const { GoogleGenerativeAIEmbeddings } = require('@langchain/google-genai');
+
+const config = require('../../config');
+const prompts = require('../prompt');
+
+const safetySettings = [
+  {
+    category: 'HARM_CATEGORY_HARASSMENT',
+    threshold: 'BLOCK_ONLY_HIGH',
+  },
+  {
+    category: 'HARM_CATEGORY_HATE_SPEECH',
+    threshold: 'BLOCK_ONLY_HIGH',
+  },
+  {
+    category: 'HARM_CATEGORY_SEXUALLY_EXPLICIT',
+    threshold: 'BLOCK_ONLY_HIGH',
+  },
+  {
+    category: 'HARM_CATEGORY_DANGEROUS_CONTENT',
+    threshold: 'BLOCK_ONLY_HIGH',
+  },
+];
+
+const model = new ChatGoogleGenerativeAI({
+  apiKey: config.geminiApiToken,
+  modelName: 'gemini-pro',
+  maxOutputTokens: 2048,
+  safetySettings,
+});
+
+const embeddings = new GoogleGenerativeAIEmbeddings({
+  apiKey: config.geminiApiToken,
+  modelName: 'embedding-001', // 768 dimensions
+  //taskType: TaskType.RETRIEVAL_DOCUMENT,
+  //title: "Document title",
+});
+
+go();
+
+console.log(prompts);
+
+async function go() {
+  const prompt = ChatPromptTemplate.fromTemplate(
+    `${prompts.systemPrompt}. In addition, you might use the vocucabulary, language, and style of the following context:
+  {context}
+  Now answer this: {input}`
+  );
+
+  const chain = await createStuffDocumentsChain({
+    llm: model,
+    prompt,
+  });
+  // const dir = 'transcripts';
+  // const files = fs.readdirSync(dir);
+
+  // for (const file of files) {
+  //   if (path.extname(file) === '.txt') {
+  //     const filePath = path.join(dir, file);
+  //     const loader = new TextLoader(filePath);
+  //     const docs = await loader.load();
+  //   }
+  // }
+
+  // const loader = new TextLoader('transcripts/_-AfhLQfb6w.txt');
+  // const docs = await loader.load();
+
+  const loader = new DirectoryLoader('transcripts', {
+    '.txt': (path) => new TextLoader(path),
+  });
+  const docs = await loader.load();
+
+  const splitter = new RecursiveCharacterTextSplitter({
+    chunkSize: 1000,
+    chunkOverlap: 20,
+  });
+  const splitDocs = await splitter.splitDocuments(docs);
+
+  const vectorstore = await MemoryVectorStore.fromDocuments(
+    splitDocs,
+    embeddings
+  );
+  const retriever = vectorstore.asRetriever({ k: 2 });
+  const retrievalChain = await createRetrievalChain({
+    combineDocsChain: chain,
+    retriever,
+  });
+  const response = await retrievalChain.invoke({
+    input:
+      'Hi MattGPT! Please introduce yourself for the audience here at the Bell House, Brooklyn. Tell us what you are here to do.',
+  });
+
+  console.log(response);
+}

src/experimental/langchain-test.js

@@ -0,0 +1,44 @@
+const { OpenAI, ChatOpenAI, OpenAIEmbeddings } = require('@langchain/openai');
+// const { TextLoader } = require('langchain/document_loaders/fs/text');
+const { RecursiveCharacterTextSplitter } = require('langchain/text_splitter');
+const { HNSWLib } = require('langchain/vectorstores/hnswlib');
+
+const fs = require('fs');
+
+const config = require('../../config');
+
+const chatModel = new ChatOpenAI({
+  openAIApiKey: config.openAIApiToken,
+});
+
+const model = new OpenAI({
+  openAIApiKey: config.openAIApiToken,
+});
+
+const embeddings = new OllamaEmbeddings({
+  openAIApiKey: config.openAIApiToken,
+});
+
+testChat();
+
+async function testChat() {
+  //   const loader = new TextLoader('transcripts/example.txt');
+  //   const docs = await loader.load();
+
+  const text = fs.readFileSync('transcripts/example.txt', 'utf8');
+  const textSplitter = new RecursiveCharacterTextSplitter({ chunkSize: 1000 });
+  const docs = await textSplitter.createDocuments([text]);
+
+  const vectorStore = await HNSWLib.fromDocuments(docs, embeddings());
+
+  // Initialize a retriever wrapper around the vector store
+  const retriever = vectorStore.asRetriever();
+
+  const references = await retriever.getRelevantDocuments(
+    'what is a raindobw?'
+  );
+  console.log(references);
+
+  //const response = await chatModel.invoke('How are you?');
+  //console.log(response);
+}