Langbase Python SDK

The official Python SDK for Langbase - Build declarative and composable AI-powered LLM products with ease.

Documentation

Check the Langbase SDK documentation for more details.

The following examples are for reference only. Prefer docs for the latest information.

Features

• Simple and intuitive API - Get started in minutes
• Streaming support - Real-time text generation with typed events
• Type safety - Full type hints for better IDE support
• Minimal dependencies - Only what you need
• Python 3.7+ - Support for modern Python versions

Installation

Install Langbase SDK:

pip install langbase

Install dotenv:

pip install dotenv

Quick Start

1. Set up your API key

Create a .env file and add your Langbase API Key.

LANGBASE_API_KEY="your-api-key"
LLM_API_KEY="your-llm-api-key"

---

2. Initialize the client

from langbase import Langbase
import os
from dotenv import load_dotenv

load_dotenv()

# Get API key from environment variable
langbase_api_key = os.getenv("LANGBASE_API_KEY")
llm_api_key = os.getenv("LLM_API_KEY")

# Initialize the client
langbase = Langbase(api_key=langbase_api_key)

3. Generate text

# Simple generation
response = langbase.agent.run(
    input=[{"role": "user", "content": "Tell me about AI"}],
    model="openai:gpt-4.1-mini",
    api_key=llm_api_key,
)

print(response["output"])

---

4. Stream text (Simple)

form langbase import get_runner

# Stream text as it's generated
response = langbase.agent.run(
    input=[{"role": "user", "content": "Tell me about AI"}],
    model="openai:gpt-4.1-mini",
    api_key=llm_api_key,
    stream=True,
)

runner = get_runner(response)

for content in runner.text_generator():
    print(content, end="", flush=True)

5. Stream with typed events (Advanced)

from langbase import StreamEventType, get_typed_runner

response = langbase.agent.run(
    input=[{"role": "user", "content": "What is an AI Engineer?"}],
    model="openai:gpt-4.1-mini",
    api_key=llm_api_key,
    stream=True,
)

# Create typed stream processor
runner = get_typed_runner(response)

# Register event handlers
runner.on(
    StreamEventType.CONNECT,
    lambda event: print(f"✓ Connected! Thread ID: {event['threadId']}\n"),
)

runner.on(
    StreamEventType.CONTENT,
    lambda event: print(event["content"], end="", flush=True),
)

runner.on(
    StreamEventType.TOOL_CALL,
    lambda event: print(
        f"\n🔧 Tool call: {event['toolCall']['function']['name']}"
    ),
)

runner.on(
    StreamEventType.COMPLETION,
    lambda event: print(f"\n\n✓ Completed! Reason: {event['reason']}"),
)

runner.on(
    StreamEventType.ERROR,
    lambda event: print(f"\n❌ Error: {event['message']}"),
)

runner.on(
    StreamEventType.END,
    lambda event: print(f"⏱️  Total duration: {event['duration']:.2f}s"),
)

# Process the stream
runner.process()

Core Features

Pipes - AI Pipeline Execution

# List all pipes
pipes = langbase.pipes.list()

# Run a pipe
response = langbase.pipes.run(
    name="ai-agent",
    messages=[{"role": "user", "content": "Hello!"}],
    variables={"style": "friendly"},  # Optional variables
    stream=True,  # Enable streaming
)

Memory - Persistent Context Storage

# Create a memory
memory = langbase.memories.create(
    name="product-docs",
    description="Product documentation",
)

# Upload documents
langbase.memories.documents.upload(
    memory_name="product-docs",
    document_name="guide.pdf",
    document=open("guide.pdf", "rb"),
    content_type="application/pdf",
)

# Retrieve relevant context
results = langbase.memories.retrieve(
    query="How do I get started?",
    memory=[{"name": "product-docs"}],
    top_k=3,
)

Agent - LLM Agent Execution

# Run an agent with tools
response = langbase.agent.run(
response = langbase.agent.run(
    model="openai:gpt-4",
    messages=[{"role": "user", "content": "Search for AI news"}],
    tools=[{"type": "function", "function": {...}}],
    tool_choice="auto",
    api_key="your-llm-api-key",
    stream=True,
)

Tools - Built-in Utilities

# Chunk text for processing
chunks = langbase.chunker(
chunks = langbase.chunker(
    content="Long text to split...",
    chunk_max_length=1024,
    chunk_overlap=256,
)

# Generate embeddings
embeddings = langbase.embed(
embeddings = langbase.embed(
    chunks=["Text 1", "Text 2"],
    embedding_model="openai:text-embedding-3-small",
)

# Parse documents
content = langbase.parser(
content = langbase.parser(
    document=open("document.pdf", "rb"),
    document_name="document.pdf",
    content_type="application/pdf",
)

Examples

Explore the examples directory for complete working examples:

• Generate text
• Stream text
• Work with memory
• Agent with tools
• Document processing
• Workflow automation

SDK Reference

For detailed SDK documentation, visit langbase.com/docs/sdk.

Contributing

We welcome contributions! Please see our Contributing Guide for details.

Support

• Documentation
• Discord Community
• Issue Tracker

License

See the LICENSE file for details.