feat: solution for 'Повторный экзамен: FAQ-бот — ChromaDB + один MCP-tool'

This commit is contained in:
2026-07-01 15:18:54 +03:00
parent b17c7be620
commit 680e00a2da
6 changed files with 200 additions and 182 deletions
+21 -23
View File
@@ -1,25 +1,23 @@
const VectorStore = require('./vectorStore');
class Bot {
constructor() {
this.vectorStore = new VectorStore();
}
async init() {
await this.vectorStore.connect();
}
async indexFAQs(faqs) {
for (const faq of faqs) {
const combined = `${faq.question}\n${faq.answer}`;
await this.vectorStore.addDocument(faq.id, combined);
}
}
async answer(question) {
const results = await this.vectorStore.query(question, 3);
return results.map(r => r.document).join('\n---\n');
}
/**
* Minimal ContextAware Prompt (MCP) tool.
* Generates a prompt that can be used for vector search.
*/
export function generatePrompt(question) {
return `Answer the following question based on the knowledge base: "${question}"`;
}
module.exports = Bot;
/**
* Handles a user query by generating a prompt, searching the vector store,
* and returning the best answer.
* @param {string} question
* @param {ChromaVectorStore} vectorStore
* @returns {Promise<string>}
*/
export async function answerQuestion(question, vectorStore) {
const prompt = generatePrompt(question);
const results = await vectorStore.similaritySearch(prompt, 1);
if (results.length === 0) {
return "I couldn't find an answer to that question.";
}
return results[0];
}
+39 -25
View File
@@ -1,29 +1,43 @@
const Bot = require('./bot');
import readlineSync from 'readline-sync';
import ChromaVectorStore from './vectorStore.js';
import { answerQuestion } from './bot.js';
const faqs = [
{
id: '1',
question: 'What is ChromaDB?',
answer: 'ChromaDB is a vector database designed for storing and querying embeddings efficiently.',
},
{
id: '2',
question: 'How do I use MCP-tool?',
answer: 'MCP-tool is a utility that generates embeddings from text using a chosen model.',
},
{
id: '3',
question: 'Can I delete a document from the vector store?',
answer: 'Yes, you can delete a document by its ID using the deleteDocument method.',
},
/**
* Sample FAQ dataset.
* In a real application this would be loaded from a file or database.
*/
const faqData = [
{ id: '1', text: 'What is ChromaDB?', metadata: { category: 'database' } },
{ id: '2', text: 'How do I install ChromaDB?', metadata: { category: 'installation' } },
{ id: '3', text: 'What is an MCP-tool?', metadata: { category: 'concept' } },
{ id: '4', text: 'How to use the FAQ bot?', metadata: { category: 'usage' } },
];
(async () => {
const bot = new Bot();
await bot.init();
await bot.indexFAQs(faqs);
/**
* Main entry point.
*/
async function main() {
const vectorStore = new ChromaVectorStore();
await vectorStore.init('faq');
const userQuestion = 'Explain ChromaDB';
const response = await bot.answer(userQuestion);
console.log('Answer:\n', response);
})();
// Load data into the collection if it is empty.
// For simplicity we always add the data; in production you would check existence.
await vectorStore.addDocuments(faqData);
console.log('FAQ bot is ready. Type your question (or "exit" to quit).');
while (true) {
const question = readlineSync.question('> ');
if (question.trim().toLowerCase() === 'exit') {
console.log('Goodbye!');
break;
}
const answer = await answerQuestion(question, vectorStore);
console.log(`Answer: ${answer}`);
}
}
main().catch(err => {
console.error('Error:', err);
process.exit(1);
});
+57 -31
View File
@@ -1,45 +1,71 @@
const { ChromaClient } = require('chromadb');
const { MCPTool } = require('mcp-tool');
import { ChromaClient } from 'chromadb';
class VectorStore {
/**
* Simple embedding utility.
* Produces a 768dimensional vector where each dimension is a count of
* the number of words that hash to that index.
*/
function embed(text) {
const vector = new Array(768).fill(0);
const words = text.toLowerCase().split(/\s+/);
for (const word of words) {
const hash = [...word].reduce((acc, ch) => acc + ch.charCodeAt(0), 0);
const idx = hash % 768;
vector[idx] += 1;
}
return vector;
}
/**
* Wrapper around ChromaDB providing a minimal API for the bot.
*/
class ChromaVectorStore {
constructor() {
this.client = new ChromaClient({ path: './chromadb' });
this.client = new ChromaClient();
this.collection = null;
this.mcp = new MCPTool(); // default configuration
}
async connect() {
this.collection = await this.client.getOrCreateCollection('faq');
}
async addDocument(id, text) {
const embedding = await this.mcp.embed(text);
await this.collection.add({
ids: [id],
embeddings: [embedding],
documents: [text],
/**
* Initializes the collection. Creates it if it does not exist.
* @param {string} name - Collection name.
*/
async init(name = 'faq') {
this.collection = await this.client.getOrCreateCollection({
name,
});
}
async query(text, k = 5) {
const embedding = await this.mcp.embed(text);
/**
* Adds documents to the collection.
* @param {Array<{id?: string, text: string, metadata?: object}>} docs
*/
async addDocuments(docs) {
const ids = docs.map((d, idx) => d.id ?? `doc-${idx}`);
const metadatas = docs.map(d => d.metadata ?? {});
const embeddings = docs.map(d => embed(d.text));
await this.collection.add({
ids,
documents: docs.map(d => d.text),
metadatas,
embeddings,
});
}
/**
* Performs a similarity search.
* @param {string} queryText
* @param {number} k
* @returns {Promise<Array<string>>} Top k documents.
*/
async similaritySearch(queryText, k = 3) {
const queryEmbedding = embed(queryText);
const results = await this.collection.query({
queryEmbeddings: [embedding],
queryEmbeddings: [queryEmbedding],
nResults: k,
});
const ids = results.ids[0];
const distances = results.distances[0];
const documents = results.documents[0];
return ids.map((id, idx) => ({
id,
score: distances[idx],
document: documents[idx],
}));
}
async deleteDocument(id) {
await this.collection.delete({ ids: [id] });
return results[0].documents;
}
}
module.exports = VectorStore;
export default ChromaVectorStore;
export { embed };