AI Customer Support Chatbot

# ingest.py — Run once to load documents
import os
from langchain_community.document_loaders import DirectoryLoader, TextLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_openai import OpenAIEmbeddings
from langchain_community.vectorstores import Chroma

# Load documents from a folder
loader = DirectoryLoader("./docs", glob="**/*.md", loader_cls=TextLoader)
documents = loader.load()

# Split into chunks
splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=50)
chunks = splitter.split_documents(documents)

# Create embeddings and store in ChromaDB
embeddings = OpenAIEmbeddings(model="text-embedding-3-small")
vectorstore = Chroma.from_documents(
    chunks,
    embeddings,
    persist_directory="./chroma_db"
)
print(f"Ingested {len(chunks)} chunks from {len(documents)} documents")

# chatbot.py
import os
from openai import OpenAI
from langchain_openai import OpenAIEmbeddings
from langchain_community.vectorstores import Chroma

client = OpenAI()
embeddings = OpenAIEmbeddings(model="text-embedding-3-small")
vectorstore = Chroma(
    persist_directory="./chroma_db",
    embedding_function=embeddings
)

SYSTEM_PROMPT = """You are a helpful customer support assistant for Sinkur.
Answer questions based ONLY on the provided context. If the context doesn't
contain the answer, say "I don't have that information" and offer to
connect the user with a human agent. Be concise and professional."""

def get_response(user_message: str, chat_history: list) -> dict:
    # 1. Retrieve relevant context
    docs = vectorstore.similarity_search(user_message, k=3)
    context = "\n\n".join([doc.page_content for doc in docs])

    # 2. Build messages with context
    messages = [
        {"role": "system", "content": SYSTEM_PROMPT},
        {"role": "system", "content": f"Context:\n{context}"},
        *chat_history,
        {"role": "user", "content": user_message}
    ]

    # 3. Generate response
    response = client.chat.completions.create(
        model="gpt-4o-mini",
        messages=messages,
        max_tokens=400,
        temperature=0.3
    )

    reply = response.choices[0].message.content

    # 4. Check if escalation needed
    needs_escalation = any(phrase in reply.lower() for phrase in [
        "i don't have that information",
        "connect you with",
        "human agent"
    ])

    return {
        "reply": reply,
        "escalate": needs_escalation,
        "sources": [doc.metadata.get("source", "") for doc in docs]
    }

# main.py
from fastapi import FastAPI
from fastapi.middleware.cors import CORSMiddleware
from pydantic import BaseModel
from chatbot import get_response
import logging

app = FastAPI(title="Sinkur Support Chatbot")

app.add_middleware(
    CORSMiddleware,
    allow_origins=["https://sinkur.com"],
    allow_methods=["POST"],
    allow_headers=["Content-Type"]
)

logger = logging.getLogger("chatbot")

class ChatRequest(BaseModel):
    message: str
    history: list = []
    session_id: str = ""

@app.post("/api/chat")
async def chat(request: ChatRequest):
    result = get_response(request.message, request.history)

    # Log for analytics
    logger.info(f"Session: {request.session_id} | "
                f"Query: {request.message} | "
                f"Escalated: {result['escalate']}")

    return result

// chat-widget.js — embed on any page
(function() {
    const API_URL = '/api/chat';
    let history = [];
    const sessionId = crypto.randomUUID();

    async function sendMessage(message) {
        // Add user message to UI
        addMessageToUI('user', message);

        const response = await fetch(API_URL, {
            method: 'POST',
            headers: { 'Content-Type': 'application/json' },
            body: JSON.stringify({
                message: message,
                history: history,
                session_id: sessionId
            })
        });

        const data = await response.json();

        // Update history
        history.push({ role: 'user', content: message });
        history.push({ role: 'assistant', content: data.reply });

        // Show response
        addMessageToUI('assistant', data.reply);

        // If escalation needed, show human handoff button
        if (data.escalate) {
            showEscalationOption();
        }
    }
})();