SUTRA by TWO Platforms

SUTRA is a family of large multi-lingual language (LMLMs) models pioneered by Two Platforms. SUTRA’s dual-transformer approach extends the power of both MoE and Dense AI language model architectures, delivering cost-efficient multilingual capabilities for over 50+ languages. It powers scalable AI applications for conversation, search, and advanced reasoning, ensuring high-performance across diverse languages, domains and applications.

📚 Multilingual Chat With URL (Powered by SUTRA)

Get Your API Keys

Before you begin, make sure you have:

A SUTRA API key (Get yours at TWO AI's SUTRA API page)
Basic familiarity with Python and Jupyter notebooks

This notebook is designed to run in Google Colab, so no local Python installation is required.

1. Install Required Packages

!pip install -q langchain langchain_openai langchain-community faiss-cpu requests pypdf python-docx

import os
from google.colab import userdata

# Set the API key from Colab secrets
os.environ["SUTRA_API_KEY"] = userdata.get("SUTRA_API_KEY")
os.environ["OPENAI_API_KEY"] = userdata.get("OPENAI_API_KEY")

Setup API Keys

import os
from google.colab import userdata

# Set the API key from Colab secrets
os.environ["SUTRA_API_KEY"] = userdata.get("SUTRA_API_KEY")
os.environ["OPENAI_API_KEY"] = userdata.get("OPENAI_API_KEY")

2. Import Required Libraries

import os
import requests
from langchain_community.document_loaders import PyPDFLoader, Docx2txtLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.vectorstores import FAISS
from langchain_openai import OpenAIEmbeddings, ChatOpenAI
from langchain.memory import ConversationBufferMemory
from langchain.chains import ConversationalRetrievalChain
from langchain.schema import HumanMessage

3. Function to Download PDF File

def download_pdf(pdf_url, save_path="downloaded.pdf"):
    response = requests.get(pdf_url)
    with open(save_path, "wb") as f:
        f.write(response.content)
    return save_path

4. Function to Load Chat Model (Sutra)

def get_chat_model():
    return ChatOpenAI(
        api_key=os.getenv("SUTRA_API_KEY"),
        base_url="https://api.two.ai/v2",
        model="sutra-v2",
        temperature=0.7
    )

5. Function to Process Documents into Conversation Chain

def process_documents(file_paths, chunk_size=1000, chunk_overlap=100):
    documents = []

    for file_path in file_paths:
        if file_path.endswith(".pdf"):
            loader = PyPDFLoader(file_path)
            documents.extend(loader.load())
        elif file_path.endswith(".docx"):
            loader = Docx2txtLoader(file_path)
            documents.extend(loader.load())

    text_splitter = RecursiveCharacterTextSplitter(
        chunk_size=chunk_size,
        chunk_overlap=chunk_overlap
    )
    chunks = text_splitter.split_documents(documents)

    embeddings = OpenAIEmbeddings(api_key=os.getenv("OPENAI_API_KEY"))  # Use OpenAI embeddings
    vectorstore = FAISS.from_documents(chunks, embeddings)

    memory = ConversationBufferMemory(
        memory_key="chat_history",
        return_messages=True
    )

    conversation_chain = ConversationalRetrievalChain.from_llm(
        llm=get_chat_model(),
        retriever=vectorstore.as_retriever(),
        memory=memory
    )

    return conversation_chain

6. Function to Query the Document

def process_chat(conversation_chain, user_input, selected_language):
    rag_response = conversation_chain.invoke(user_input)
    context = rag_response["answer"]

    system_message = f'''
    You are a helpful assistant that answers questions about documents.
    Use the following context to answer the question.

    CONTEXT:
    {context}

    Please respond in {selected_language}.
    '''

    messages = [HumanMessage(content=f"{system_message}\n\nQuestion: {user_input}")]
    chat = get_chat_model()
    response = chat.invoke(messages)
    return response.content

7. Example Usage Block

# STEP 1: Download the PDF
pdf_url = "https://agno-public.s3.amazonaws.com/recipes/ThaiRecipes.pdf"
downloaded_pdf_path = download_pdf(pdf_url)

# STEP 2: Process documents
conversation_chain = process_documents([downloaded_pdf_path])

# STEP 3: Ask a question
user_input = "How do I make chicken and galangal in coconut milk soup"
selected_language = "hindi"

# STEP 4: Get response
answer = process_chat(conversation_chain, user_input, selected_language)
print(answer)

Finally Integrated UI

# 1. Imports
import os
import requests
import ipywidgets as widgets
from IPython.display import display, HTML, clear_output
from langchain_community.document_loaders import PyPDFLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.vectorstores import FAISS
from langchain_openai import OpenAIEmbeddings, ChatOpenAI
from langchain.memory import ConversationBufferMemory
from langchain.chains import ConversationalRetrievalChain
from langchain.schema import HumanMessage
from tempfile import NamedTemporaryFile

# 2. Setup LLM (Sutra)
def get_sutra_model():
    return ChatOpenAI(
        api_key=os.getenv("SUTRA_API_KEY"),
        base_url="https://api.two.ai/v2",
        model="sutra-v2",
        temperature=0.7
    )

# 3. Load and index PDF
def load_and_index_pdf(pdf_path):
    loader = PyPDFLoader(pdf_path)
    docs = loader.load()

    text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=100)
    chunks = text_splitter.split_documents(docs)

    embeddings = OpenAIEmbeddings(api_key=os.getenv("OPENAI_API_KEY"))
    vectorstore = FAISS.from_documents(chunks, embeddings)

    memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)
    chain = ConversationalRetrievalChain.from_llm(
        llm=get_sutra_model(),
        retriever=vectorstore.as_retriever(),
        memory=memory
    )
    return chain

# 4. UI Components

# PDF URL field
pdf_url_input = widgets.Text(
    placeholder='Enter PDF URL...',
    description='📄 PDF URL:',
    layout=widgets.Layout(width='100%')
)

load_pdf_button = widgets.Button(
    description="🔄 Load PDF",
    button_style='info',
    layout=widgets.Layout(width='150px')
)

status_output = widgets.Output()

# Language dropdown
languages = [
    "English", "Hindi", "Gujarati", "Bengali", "Tamil",
    "Telugu", "Kannada", "Malayalam", "Punjabi", "Marathi",
    "Urdu", "Assamese", "Odia", "Sanskrit", "Korean",
    "Japanese", "Arabic", "French", "German", "Spanish",
    "Portuguese", "Russian", "Chinese", "Vietnamese", "Thai",
    "Indonesian", "Turkish", "Polish", "Ukrainian", "Dutch",
    "Italian", "Greek", "Hebrew", "Persian", "Swedish",
    "Norwegian", "Danish", "Finnish", "Czech", "Hungarian",
    "Romanian", "Bulgarian", "Croatian", "Serbian", "Slovak",
    "Slovenian", "Estonian", "Latvian", "Lithuanian", "Malay",
    "Tagalog", "Swahili"
]

lang_dropdown = widgets.Dropdown(
    options=languages,
    value="English",
    description='🌐 Language:',
    layout=widgets.Layout(width='300px')
)

chat_output = widgets.HTML(
    value="<div style='padding:10px; font-family:Arial; font-size:14px; height:300px; overflow-y:auto; border:1px solid #ccc; border-radius:5px;'>Chat history will appear here...</div>"
)

user_input = widgets.Text(
    placeholder='Type your message...',
    layout=widgets.Layout(flex='4', width='auto')
)

send_button = widgets.Button(
    description="📤 Send",
    button_style='primary',
    layout=widgets.Layout(flex='1', width='auto')
)

messages = []
conversation_chain = None

# 5. Load PDF Logic
def download_pdf(url):
    response = requests.get(url)
    tmp_file = NamedTemporaryFile(delete=False, suffix=".pdf")
    with open(tmp_file.name, "wb") as f:
        f.write(response.content)
    return tmp_file.name

def on_load_pdf(b):
    global conversation_chain
    pdf_url = pdf_url_input.value.strip()
    with status_output:
        clear_output()
        if not pdf_url:
            print("❌ Please enter a valid PDF URL.")
            return
        print("⏳ Downloading and processing PDF...")
        try:
            pdf_path = download_pdf(pdf_url)
            conversation_chain = load_and_index_pdf(pdf_path)
            print("✅ PDF loaded and indexed successfully!")
        except Exception as e:
            print("❌ Error:", e)

load_pdf_button.on_click(on_load_pdf)

# 6. Chat Interaction Logic
def on_send_click(b):
    global conversation_chain
    if conversation_chain is None:
        with status_output:
            clear_output()
            print("❌ Load a PDF first.")
        return

    user_text = user_input.value.strip()
    if not user_text:
        return

    lang = lang_dropdown.value
    messages.append(f"<b style='color:#13f22d;'>You:</b> {user_text}")

    context_response = conversation_chain.invoke(user_text)
    rag_context = context_response['answer']

    system_msg = f'''
You are a helpful assistant answering based on a document.
Use this context: {rag_context}
Always reply in: {lang}
Question: {user_text}
'''

    chat_model = get_sutra_model()
    sutra_response = chat_model.invoke([HumanMessage(content=system_msg)])
    assistant_reply = sutra_response.content.strip()

    messages.append(f"<b style='color:#007acc;'>Assistant ({lang}):</b> {assistant_reply}")

    chat_html = "<br>".join(messages)
    chat_output.value = f"<div style='padding:10px; font-family:Arial; font-size:14px; height:300px; overflow-y:auto; border:1px solid #ccc; border-radius:5px;'>{chat_html}</div>"

    user_input.value = ""

send_button.on_click(on_send_click)

# 7. Final Layout
input_row = widgets.HBox([user_input, send_button])
pdf_row = widgets.HBox([pdf_url_input, load_pdf_button])

ui = widgets.VBox([
    widgets.HTML("<h3 style='font-family:Arial;'>📚 Multilingual Chat with PDF (Powered by Sutra)</h3>"),
    pdf_row,
    lang_dropdown,
    chat_output,
    input_row,
    status_output
])

# 8. Display the App
display(ui)

Multilingual Chat With URL

On this page