backend for LangChat

Dockerfile

@@ -0,0 +1,18 @@
+# Use an official Python runtime as a base image
+FROM python:3.9-slim
+
+# Set the working directory
+WORKDIR /app
+
+# Copy requirements and install dependencies
+COPY requirements.txt requirements.txt
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Copy the application code
+COPY . .
+
+# Expose the port FastAPI runs on
+EXPOSE 8000
+
+# Command to run the app
+CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "8000"]

app.py

@@ -0,0 +1,91 @@
+# app.py
+import os
+from dotenv import load_dotenv
+from fastapi import FastAPI, HTTPException
+from fastapi.middleware.cors import CORSMiddleware
+from pydantic import BaseModel
+from qdrant_search import QdrantSearch
+from langchain_groq import ChatGroq
+from nomic_embeddings import EmbeddingsModel
+
+load_dotenv()
+
+import warnings
+warnings.filterwarnings("ignore", category=FutureWarning)
+
+os.environ["TOKENIZERS_PARALLELISM"] = "FALSE"
+
+# Initialize FastAPI app
+app = FastAPI()
+
+# Allow CORS for frontend on Vercel
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],  # Replace "*" with your frontend URL for better security
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+# Initialize global variables
+collection_names = ["docs_v1_2", "docs_v2_2", "docs_v3_2"]
+limit = 5
+llm = ChatGroq(model="mixtral-8x7b-32768")
+embeddings = EmbeddingsModel()
+search = QdrantSearch(
+    qdrant_url=os.environ["QDRANT_CLOUD_URL"],
+    api_key=os.environ["QDRANT_API_KEY"],
+    embeddings=embeddings
+)
+
+# Define request and response models
+class QueryRequest(BaseModel):
+    question: str
+
+class AnswerResponse(BaseModel):
+    answer: str
+    sources: list
+
+# API endpoint to handle user queries
+@app.post("/api/chat", response_model=AnswerResponse)
+async def chat_endpoint(request: QueryRequest):
+    query = request.question.strip()
+    if not query:
+        raise HTTPException(status_code=400, detail="Query cannot be empty.")
+
+    # Step 1: Retrieve relevant documents from Qdrant
+    retrieved_docs = search.query_multiple_collections(query, collection_names, limit)
+
+    # Step 2: Prepare the context from retrieved documents
+    context = "\n".join([doc['text'] for doc in retrieved_docs])
+
+    # Step 3: Construct the prompt with context and question
+    prompt = (
+        "You are LangAssist, a knowledgeable assistant for the LangChain Python Library. "
+        "Given the following context from the documentation, provide a helpful answer to the user's question.\n\n"
+        "Context:\n{context}\n\n"
+        "Question: {question}\n\n"
+        "Answer:"
+    ).format(context=context, question=query)
+
+    # Step 4: Generate an answer using the language model
+    try:
+        answer = llm.invoke(prompt)
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=str(e))
+
+    # Prepare sources
+    sources = [
+        {
+            "source": doc['source'],
+            "text": doc['text']
+        } for doc in retrieved_docs
+    ]
+
+    # Step 5: Return the answer and sources
+    # return AnswerResponse(answer=answer.strip(), sources=sources)
+    return AnswerResponse(answer=answer.content.strip(), sources=sources)
+
+if __name__ == "__main__":
+    import uvicorn
+    uvicorn.run(app, host="0.0.0.0", port=8000)

nomic_embeddings.py

@@ -0,0 +1,30 @@
+from langchain_community.embeddings import HuggingFaceBgeEmbeddings
+
+class EmbeddingsModel:
+    def __init__(self):
+        model_name = "nomic-ai/nomic-embed-text-v1"
+        model_kwargs = {
+            'device': 'cpu',
+            'trust_remote_code': True
+        }
+        encode_kwargs = {'normalize_embeddings': True}
+        self.embeddings = HuggingFaceBgeEmbeddings(
+            model_name=model_name,
+            model_kwargs=model_kwargs,
+            encode_kwargs=encode_kwargs,
+            query_instruction="search_query:",
+            embed_instruction="search_document:"
+        )
+
+    def get_embeddings(self, text):
+        """
+        Returns the embeddings for the given text.
+
+        :param text: The input text to get embeddings for.
+        :return: The embeddings as a numpy array.
+        """
+        return self.embeddings.embed_query(text)
+
+# Example usage:
+# embeddings_model = EmbeddingsModel()
+# embeddings = embeddings_model.get_embeddings("Your input text here")

qdrant_search.py

@@ -0,0 +1,40 @@
+# qdrant_search.py
+
+from typing import List, Dict
+from langchain_community.embeddings import HuggingFaceBgeEmbeddings
+from qdrant_client import QdrantClient
+
+class QdrantSearch:
+    def __init__(self, qdrant_url: str, api_key: str, embeddings):
+        self.embeddings=embeddings
+        # Initialize Qdrant client
+        self.client = QdrantClient(
+            url=qdrant_url,
+            api_key=api_key,
+        )
+
+    def query_qdrant(self, query: str, collection_name: str, limit: int = 5) -> List:
+        """Retrieve relevant documents from Qdrant."""
+        query_vector = self.embeddings.get_embeddings(query)
+        
+        results = self.client.search(
+            collection_name=collection_name,
+            query_vector=query_vector,
+            limit=limit
+        )
+        return results
+
+    def query_multiple_collections(self, query: str, collection_names: List[str], limit: int = 5) -> List[Dict]:
+        """Query multiple Qdrant collections and return combined top results."""
+        all_results = []
+        
+        for collection_name in collection_names:
+            results = self.query_qdrant(query, collection_name, limit)
+            for result in results:
+                all_results.append({
+                    'text': result.payload['text'],
+                    'source': result.payload['source'],
+                    'score': result.score
+                })
+        
+        return sorted(all_results, key=lambda x: x['score'], reverse=True)[:limit]

requirements.txt

@@ -0,0 +1,11 @@
+fastapi
+uvicorn
+python-dotenv
+qdrant-client
+langchain
+langchain_community
+langchain_groq
+pydantic
+fastapi
+einops
+sentence-transformers