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

README.md

@@ -7,11 +7,19 @@
 EN
 Повторный экзамен: FAQ-бот — ChromaDB + один MCP-tool
 Зачёт
-Версия 4
+Версия 5
 Дедлайн сдачи: 31.08.2026
 
 В работе
 
 Требуется доработка
 
-К сожалению, проект не соответствует заявленному стеку задания. В описании требуется использовать ChromaDB и Ollama‑embed‑text, однако в коде реализовано хранение embeddings через Qdrant. Пересмотрите решение: замените Qdrant на ChromaDB, убедитесь, что все импорты и зависимости соответствуют
+Решение не соответствует заявленному стеку задания. Пожалуйста, пересмотрите работу и убедитесь в использовании ChromaDB с Ollama‑embed‑text, а также корректной интеграции MCP‑тулов.
+
+Редактирование ответа
+
+Заполните ответ и отправьте работу на проверку преподавателю.
+
+Тип ответа
+Текст
+Ссылка

requirements.txt

@@ -1,6 +1,3 @@
-fastapi
+chromadb==0.4.24
-uvicorn[standard]
+ollama==0.1.0
-qdrant-client
+mcp-tools==0.1.0
-sentence-transformers
-python-telegram-bot
-pydantic

src/__init__.py

@@ -1 +1 @@
-# Empty init file to make src a package
+# This file makes src a Python package.

src/database.py

@@ -0,0 +1,100 @@
+"""
+ChromaDB wrapper for storing and querying FAQ documents.
+"""
+
+import chromadb
+from chromadb.config import Settings
+from typing import List, Dict, Any
+
+class ChromaDB:
+    """
+    Wrapper around ChromaDB to handle FAQ documents.
+    """
+
+    def __init__(self, persist_path: str = "chromadb"):
+        """
+        Initialize the ChromaDB client and collection.
+
+        Parameters
+        ----------
+        persist_path : str, optional
+            Directory to persist the database. Defaults to "chromadb".
+        """
+        self.client = chromadb.Client(Settings(persist_directory=persist_path))
+        self.collection_name = "faq"
+        self.collection = self.client.get_or_create_collection(name=self.collection_name)
+
+    def add_document(self, text: str, embedding: List[float], doc_id: str = None):
+        """
+        Add a single document to the collection.
+
+        Parameters
+        ----------
+        text : str
+            The document text.
+        embedding : List[float]
+            The embedding vector for the document.
+        doc_id : str, optional
+            Optional document ID. If None, an auto-generated ID is used.
+        """
+        if doc_id is None:
+            # Generate a simple incremental ID
+            existing_ids = self.collection.get()["ids"]
+            doc_id = str(len(existing_ids))
+        self.collection.add(
+            documents=[text],
+            embeddings=[embedding],
+            ids=[doc_id]
+        )
+
+    def query(self, embedding: List[float], k: int = 5) -> List[Dict[str, Any]]:
+        """
+        Query the collection for the top-k most similar documents.
+
+        Parameters
+        ----------
+        embedding : List[float]
+            The query embedding.
+        k : int, optional
+            Number of results to return. Defaults to 5.
+
+        Returns
+        -------
+        List[Dict[str, Any]]
+            List of dictionaries containing 'id', 'document', and 'distance'.
+        """
+        results = self.collection.query(
+            query_embeddings=[embedding],
+            n_results=k,
+            include=["documents", "distances", "ids"]
+        )
+        docs = []
+        for doc, dist, doc_id in zip(results["documents"][0], results["distances"][0], results["ids"][0]):
+            docs.append({"id": doc_id, "document": doc, "distance": dist})
+        return docs
+
+    def is_empty(self) -> bool:
+        """
+        Check if the collection has any documents.
+
+        Returns
+        -------
+        bool
+            True if empty, False otherwise.
+        """
+        return len(self.collection.get()["ids"]) == 0
+
+    def load_sample_data(self, sample_data: List[Dict[str, str]]):
+        """
+        Load a list of sample documents into the collection.
+
+        Parameters
+        ----------
+        sample_data : List[Dict[str, str]]
+            List of dictionaries with keys 'text' and optional 'id'.
+        """
+        for item in sample_data:
+            text = item["text"]
+            doc_id = item.get("id")
+            embedding = embed_text(text)
+            self.add_document(text, embedding, doc_id=doc_id)

src/embedding.py

@@ -1,13 +1,28 @@
-import os
+"""
-from sentence_transformers import SentenceTransformer
+Embedding utilities using Ollama's embed-text model.
+"""
 
-# Load the embedding model once at module import
+import ollama
-MODEL_NAME = os.getenv("EMBEDDING_MODEL", "all-MiniLM-L6-v2")
+from typing import List
-model = SentenceTransformer(MODEL_NAME)
 
-
+def embed_text(text: str, model: str = "embed-text") -> List[float]:
-def get_embedding(text: str) -> list[float]:
     """
-    Generate a vector embedding for the given text using the loaded model.
+    Generate an embedding for the given text using Ollama's embed-text model.
+
+    Parameters
+    ----------
+    text : str
+        The input text to embed.
+    model : str, optional
+        The Ollama model name. Defaults to "embed-text".
+
+    Returns
+    -------
+    List[float]
+        The embedding vector.
     """
-    return model.encode(text, convert_to_numpy=False).tolist()
+    try:
+        result = ollama.embeddings(model=model, prompt=text)
+        return result["embedding"]
+    except Exception as e:
+        raise RuntimeError(f"Failed to embed text: {e}") from e

src/main.py

@@ -1,68 +1,64 @@
+"""
+Main entry point for the FAQ bot.
+"""
+
+import argparse
 import os
-from fastapi import FastAPI, HTTPException, Query
+import json
-from pydantic import BaseModel
+from src.embedding import embed_text
+from src.database import ChromaDB
+from src.mcp_tools import generate_answer
 
-from .vector_store import QdrantVectorStore
+def load_sample_faq() -> list:
-from .mcp_tool import MCPTool
-
-app = FastAPI(
-    title="FAQ Bot API",
-    description="A simple FAQ bot using Qdrant as the vector store.",
-    version="1.0.0",
-)
-
-# Initialize the vector store and MCP tool
-vector_store = QdrantVectorStore(
-    host=os.getenv("QDRANT_HOST"),
-    port=int(os.getenv("QDRANT_PORT", "6333")),
-    collection_name=os.getenv("QDRANT_COLLECTION", "faq_collection"),
-)
-mcp_tool = MCPTool(vector_store)
-
-
-class Document(BaseModel):
-    id: str
-    text: str
-    metadata: dict | None = None
-
-
-@app.post("/documents", status_code=201)
-def add_document(doc: Document):
     """
-    Add a new document to the vector store.
+    Load a small sample FAQ dataset.
     """
-    try:
+    return [
-        vector_store.add_document(doc.id, doc.text, doc.metadata)
+        {"id": "0", "text": "What is the return policy? Our return policy allows returns within 30 days of purchase."},
-    except Exception as e:
+        {"id": "1", "text": "How do I track my order? You can track your order using the tracking link sent to your email."},
-        raise HTTPException(status_code=500, detail=str(e))
+        {"id": "2", "text": "What payment methods are accepted? We accept Visa, MasterCard, and PayPal."},
-    return {"status": "added", "id": doc.id}
+        {"id": "3", "text": "Do you ship internationally? Yes, we ship to most countries worldwide."},
+        {"id": "4", "text": "How can I contact customer support? You can contact us via email at support@example.com."},
+    ]
 
+def main():
+    parser = argparse.ArgumentParser(description="FAQ Bot using ChromaDB and MCP-tools")
+    parser.add_argument("--question", type=str, help="Your question to ask the bot")
+    args = parser.parse_args()
 
-@app.get("/search")
+    if not args.question:
-def search(
+        print("Please provide a question using --question")
-    query: str = Query(..., description="Search query"),
+        return
-    top_k: int = Query(5, ge=1, le=20, description="Number of results to return"),
-):
-    """
-    Search for documents similar to the query.
-    """
-    try:
-        results = vector_store.search(query, top_k=top_k)
-    except Exception as e:
-        raise HTTPException(status_code=500, detail=str(e))
-    return {"query": query, "results": results}
 
+    # Initialize database
+    db = ChromaDB()
 
-@app.get("/answer")
+    # If the collection is empty, load sample data
-def answer(
+    if db.is_empty():
-    query: str = Query(..., description="Question to answer"),
+        print("Database empty. Loading sample FAQ data...")
-    top_k: int = Query(3, ge=1, le=10, description="Number of answers to return"),
+        sample_data = load_sample_faq()
-):
+        for item in sample_data:
-    """
+            text = item["text"]
-    Get the best answer(s) for the query using MCPTool.
+            doc_id = item.get("id")
-    """
+            embedding = embed_text(text)
-    try:
+            db.add_document(text, embedding, doc_id=doc_id)
-        answers = mcp_tool.answer(query, top_k=top_k)
+        print("Sample data loaded.")
-    except Exception as e:
+
-        raise HTTPException(status_code=500, detail=str(e))
+    # Generate embedding for the question
-    return {"query": query, "answers": answers}
+    question_embedding = embed_text(args.question)
+
+    # Query the database for relevant documents
+    results = db.query(question_embedding, k=5)
+
+    # Extract context documents
+    context = [res["document"] for res in results]
+
+    # Generate answer using MCP-tools
+    answer = generate_answer(context, args.question)
+
+    # Output the answer
+    print("\nAnswer:")
+    print(answer)
+
+if __name__ == "__main__":
+    main()

src/mcp_tools.py

@@ -0,0 +1,33 @@
+"""
+Integration with MCP-tools for generating answers.
+"""
+
+import mcp_tools
+from typing import List
+
+def generate_answer(context: List[str], question: str) -> str:
+    """
+    Generate an answer using MCP-tools given context and a question.
+
+    Parameters
+    ----------
+    context : List[str]
+        List of context strings retrieved from the database.
+    question : str
+        The user's question.
+
+    Returns
+    -------
+    str
+        The generated answer.
+    """
+    # Combine context into a single string
+    context_text = "\n".join(context)
+    # Construct a prompt for MCP-tools
+    prompt = f"Question: {question}\nContext:\n{context_text}\nAnswer:"
+    # Use MCP-tools to generate the answer
+    try:
+        response = mcp_tools.generate(prompt=prompt)
+        return response
+    except Exception as e:
+        raise RuntimeError(f"Failed to generate answer with MCP-tools: {e}") from e