Initialize the client. Args: base_url (str): The base URL of the agent service. agent (str): The name of the

Initialize STT with specified provider. Args: provider: Provider name ("openai", "deepgram", etc.) api_key: API key (

Initialize TTS with specified provider. Args: provider: Provider name ("openai", "elevenlabs", etc.) api_key: API key

Initialize VoiceManager. Args: stt: SpeechToText instance (None to disable STT) tts: TextToSpeech instance (None to d

Initialize OpenAI STT. Args: api_key: OpenAI API key (uses env var if not provided) Raises: Exception: If Op

Initialize OpenAI TTS. Args: api_key: OpenAI API key (uses env var if not provided) voice: Voice name (alloy, echo, f

Initialize the agent.

Test graph creation.

Generate a response based on the retrieved documents.

Add two numbers.

Fixture for creating a test client with a clean environment.

This node is to demonstrate doing work before the interrupt

Calculates a math expression using numexpr. Useful for when you need to answer questions about math using numexpr. This tool is only for math

Generate idiomatic operation IDs for OpenAPI client generation.

Searches chroma_db for information in the company's handbook.

This node examines the conversation history to determine user's birthdate, checking store first.

Record feedback for a run to LangSmith. This is a simple wrapper for the LangSmith create_feedback API, so the credentials can be stored

Create TTS from environment variables. Reads VOICE_TTS_PROVIDER env var to determine which provider to use. Returns None if not confi

Create VoiceManager from environment variables. Reads VOICE_STT_PROVIDER and VOICE_TTS_PROVIDER to configure speech-to-text and text-

Generates the final response based on the user's query and the available birthdate.

Get audio format (MIME type). Returns: MIME type string for generated audio

Health check endpoint.

Get chat history.

Invoke an agent with user input to retrieve a final response. If agent_id is not provided, the default agent will be used. Use thread_id

Configurable lifespan that initializes the appropriate database checkpointer, store, and agents with async loading - for example for starting

Fixture to create a mock agent that can be configured for different test scenarios.

Fixture for creating a mock AgentClient with a clean environment.

BytesIO mock audio file for STT tests.

Fixture to ensure database settings are clean

Fixture to ensure environment is clean for each test.

Patch httpx.stream and httpx.get to use our test client.

Mock OpenAI client for TTS/STT tests.

Fixture to ensure settings are clean for each test.

Fixture providing reusable messages for multi-agent tests

Multiply two numbers.

Prepare a prompt augmented with retrieved document content.

Get the provider name. Returns: Provider name string

Get the provider name. Returns: Provider name string

Retrieve relevant documents from the knowledge base.

Stream an agent's response to a user input, including intermediate messages and tokens. If agent_id is not provided, the default agent will

Test asynchronous feedback creation.

Test that streaming from our static agent works correctly with token streaming.

Test asynchronous invocation.

TODO: Can't figure out how to interact with st.feedback

Test the app with an error in the agent initialization

Test the full app - happy path

Test the full app - happy path

Test the app with streaming enabled - including tool messages

Test nested sub-agents where agent B is a sub-agent of agent A

Test when the supervisor agent transfers to sub agent A, then back to supervisor, then transfers to sub agent C, and back again

Test a single sub-agent with multiple tool calls to verify popover functionality

Test the thread_id is generated