Test asynchronous streaming.

Test that when AUTH_SECRET is set, requests with correct token are allowed

Test that when AUTH_SECRET is set, requests with wrong token are rejected

Fixture to create a FastAPI test client.

Completely empty parts should also fail to construct an AIMessage, raising a pydantic ValidationError. LANGCHAIN V1 MIGRATION NOTE:

_create_ai_message should: - Drop unknown keys ("foobar") - Preserve keys that match AIMessage signature - Use the final value

AIMessage requires 'content'; if missing, _create_ai_message should raise a pydantic ValidationError from the constructor. LANGCHAIN V1

Test graph creation.

Test from_env when both STT and TTS are configured.

Test from_env returns None (and logs error) for invalid provider.

Test from_env returns None (and logs error) for invalid provider.

Test from_env when neither STT nor TTS are configured.

Test from_env when only TTS is configured.

Test from_env returns None when VOICE_STT_PROVIDER is not set.

Test from_env returns None when VOICE_TTS_PROVIDER is not set.

Test from_env creates STT instance with valid provider.

Test from_env creates TTS instance with valid provider.

Test that API errors are handled gracefully.

Test successful audio generation.

Test getting a lazy agent that has been loaded.

Test getting a lazy agent that hasn't been loaded.

Test getting a non-existent agent.

Test getting a static agent.

Test that API key is loaded from environment when not provided.

Test that API key is loaded from environment when not provided.

Test that explicit api_key parameter takes precedence over env var.

Test that explicit api_key parameter takes precedence over env var.

Test that get_graph works after load.

Test that get_graph raises error before load.

Test get_graph when agent is loaded.

Test that get_graph raises error if no graph was created.

Test get_graph when agent is not loaded.

Test chat history retrieval.

Test header generation with and without auth.

Test that /info returns the correct service metadata.

Test client initialization with different parameters.

Test creating OpenAISTT with API key.

Test creating VoiceManager with both STT and TTS.

Test that invalid model raises ValueError.

Test that invalid provider raises ValueError.

Test that invalid provider raises ValueError.

Test that invalid voice raises ValueError.

Test creating VoiceManager with only TTS (STT=None).

Test creating STT with openai provider and explicit API key.

Test creating TTS with openai provider and explicit API key.

Test that unimplemented provider raises NotImplementedError.

Test that unimplemented provider raises NotImplementedError.

Test creating OpenAITTS with valid parameters.

Test that agent initializes correctly.

Test that agent initializes correctly.

Test synchronous invocation.

Test that /invoke works with a custom agent_id path parameter.

Test that the agent_config parameter is correctly passed to the agent.

Test that the model parameter is correctly passed to the agent if specified.

Test that when no model is specified, UserInput defaults to None and isn't passed to the runnable config (not hardcoded gpt-5-nano).

Test that the lifespan sets up the database and store, loads the agents, and logs errors.

Test that load works correctly.

Test loading a lazy agent.

Test loading a non-existent agent.

Test loading a static agent (no-op).

Test load when GITHUB_PAT is set and configured.

Test load when MCP client creation fails.

Test load when GITHUB_PAT is not set.

Test LogLevel enum and its conversion to logging levels.

Verify that our list of messages is converted to the expected output.

Test that when AUTH_SECRET is not set, all requests are allowed

Test the service using the app. This test requires the service container to be running with USE_FAKE_MODEL=true

Test the service using the fake model. This test requires the service container to be running with USE_FAKE_MODEL=true

Test Azure OpenAI settings.

Test that LOG_LEVEL defaults to WARNING.

Test that LOG_LEVEL can be set from environment variable.