PyIndus

A Python package for interacting with Indus, a ChatGPT alternative by Sarvam AI.

---

---

Installation

pip install pyindus

Or with uv:

uv add pyindus

Quick Start

IndusClient acts as a fully-featured, seamless SDK. It automatically saves, loads, and refreshes sessions for you.

1. Initial Login

Run this once to authenticate. The client will automatically save your session to indus_session.json by default.

from pyindus import IndusClient

# Login with phone number
client = IndusClient()
client.login("+91XXXXXXXXXX")

# Enter the OTP received via SMS
client.verify_otp("123456")

# The session is now authenticated and saved automatically!

2. Immediate Re-use (Like an SDK)

Run this anywhere else in your project. Because the session was saved, the client automatically loads it on __init__. If the token expires, the client will dynamically refresh it in the background.

from pyindus import IndusClient

# Automatically loads the previous session from 'indus_session.json'
client = IndusClient()

# Chat directly! No need to login again.
response = client.chat("What is quantum computing?")
print(response.answer)

Integration Guide: Custom Paths

If you're building a web app or managing multiple users, you can specify individual session files.

from pyindus import IndusClient

# Supply a unique path for the user's session
def handle_user_request(user_id, message):
    session_path = f"sessions/user_{user_id}.json"

    # Auto-loads and manages session in this specific file
    with IndusClient(session_file=session_path) as client:
        return client.chat(message)

Advanced Usage

Working with Specific Models

Indus supports different "Task Graphs" (models like Sarvam Think, Bulbul, etc.). By default, IndusClient selects the first available chat model automatically.

from pyindus import IndusClient

with IndusClient() as client:
    # List available models
    models = client.get_models()
    for model in models:
        print(f"{model.name}: {model.description}")

    # Use a specific model
    response = client.chat("Explain gravity", task_graph_uid=models[-1].uid)
    print(response.answer)

File Attachments

Upload files and attach them to your messages:

from pyindus import IndusClient

with IndusClient() as client:
    # Upload a file
    attachment = client.upload_attachment("./document.pdf")

    # Chat with the attachment
    response = client.chat("Summarize this document", attachments=[attachment])
    print(response.answer)

Session Management

with IndusClient() as client:
    # List all sessions
    sessions = client.list_sessions()
    for s in sessions:
        print(f"{s.title} ({s.uid})")

    # Delete a session
    client.delete_session(session_uid)

    # Start fresh
    client.new_session()

TUI Usage

Launch the interactive terminal UI with:

pyindus chat

Features

• Rich terminal interface with styled panels, markdown rendering, and color-coded output
• Slash commands: /auth, /model, /session, /history, /delete, /attach, /new, /clear, /help, /exit
• File attachments: Use @path/to/file in messages or /attach file.txt to upload files
• Session history: Browse and manage past chat sessions
• Model picker: List and switch between available AI models
• Session persistence: Automatically saves and loads sessions
• Tool call display: Shows search queries and tool usage during responses
• Markdown rendering: Code blocks, lists, and formatting in responses

Auth Flow

1. Launch with pyindus chat
2. Run /auth to start the login flow
3. Enter your phone number with country code (e.g., +918874163264)
4. Enter the OTP received via SMS
5. Start chatting!

File Attachments

Attach files to your messages using the @ prefix:

you (Sarvam Think) > @./report.pdf summarize this document

Or use the /attach command to queue files before sending:

you (Sarvam Think) > /attach screenshot.png
Attached: screenshot.png (45KB) (1 file pending)
you (Sarvam Think) > what's in this image?

Session Management

/history    - List recent chat sessions
/delete     - Delete the current session
/new        - Start a fresh session
/session    - Show current session info

Screenshot

---

Development

Prerequisites

• Python 3.10+
• uv (recommended) or pip

Setup

# Clone the repository
git clone https://github.com/yourusername/pyindus.git
cd pyindus

# Install with dev dependencies (using uv)
uv sync

# Or with pip
pip install -e ".[dev]"

Project Structure

pyindus/
├── src/pyindus/
│   ├── __init__.py       # Package exports
│   ├── auth.py           # Ory/Kratos authentication flow
│   ├── chat.py           # Chat API operations (prompts, sessions, attachments)
│   ├── client.py         # High-level SDK client
│   ├── cli.py            # CLI entrypoints (pyindus auth/chat)
│   ├── exceptions.py     # Custom exception hierarchy
│   ├── models.py         # Pydantic models for API responses
│   └── tui.py            # Rich terminal UI
├── tests/
│   ├── test_auth.py      # Auth unit tests (mocked)
│   ├── test_chat.py      # Chat API unit tests (mocked)
│   ├── test_client.py    # Client integration tests (mocked)
│   ├── test_live.py      # Live API tests (requires real session)
│   ├── test_models.py    # Pydantic model tests
│   └── test_tui.py       # TUI command and rendering tests
├── pyproject.toml
└── uv.lock

Running Tests

# Run all tests (mocked, no API calls)
uv run pytest tests/ -v

# Run tests excluding live API tests
uv run pytest tests/ -m "not live" -v

# Run only live tests (requires valid indus_session.json)
uv run pytest tests/ -m "live" -v

# Run with coverage
uv run pytest tests/ --cov=pyindus --cov-report=term-missing

# Run a specific test file
uv run pytest tests/test_tui.py -v

Test Categories

• Unit tests (test_auth.py, test_chat.py, test_client.py, test_models.py, test_tui.py): Use respx to mock HTTP calls. Safe to run offline, no API credentials needed.
• Live tests (test_live.py): Make real API calls. Require a valid indus_session.json in the project root. Marked with @pytest.mark.live.

Linting & Type Checking

# Install dev tools
uv add --dev ruff mypy

# Format code
uv run ruff format src/ tests/

# Lint
uv run ruff check src/ tests/

# Type check
uv run mypy src/pyindus/

Building the Package

# Install build tool
uv add --dev build

# Build sdist and wheel
uv run python -m build

# Output will be in dist/
ls dist/
# pyindus-0.1.0.tar.gz
# pyindus-0.1.0-py3-none-any.whl

Publishing to PyPI

# Install twine
uv add --dev twine

# Upload to PyPI (requires PyPI account and API token)
uv run twine upload dist/*

# Or upload to Test PyPI first
uv run twine upload --repository testpypi dist/*

PyPI Setup:

1. Create an account at pypi.org
2. Generate an API token at pypi.org/manage/account/token/
3. Configure ~/.pypirc:

   [pypi]
   username = __token__
   password = pypi-YOUR_API_TOKEN

Using uv for publishing:

# uv can publish directly
uv publish dist/*

Offline Testing

All unit tests run completely offline. The tests use respx to mock all HTTP requests, so no network access or API credentials are needed.

# This runs entirely offline - no API calls made
uv run pytest tests/ -m "not live" -v

To verify tests are truly offline, you can disconnect from the network or use a firewall:

# Block network access for tests (Linux)
uv run pytest tests/ -m "not live" -v  # All 100+ tests pass offline

Version Bumping

Update the version in pyproject.toml:

[project]
version = "0.2.0"  # Update this

Then rebuild and publish:

uv run python -m build
uv run twine upload dist/*

License

MIT