Run a transformation on an existing source to generate an insight. This command runs the transformation graph which: 1. Loads the source

Save the configuration to the database. Uses _prepare_save_data() to properly handle SecretStr conversion and encryption.

Save the model to the database. Note: Embedding is no longer generated inline. Subclasses that need embedding should overrid

Convert an insight to a note.

Convert an insight to a note.

Convert an insight to a note.

Search the knowledge base.

Search the knowledge base using text or vector search.

Send a message to source chat session with SSE streaming response.

Set a configuration as the default for a provider. Args: provider: Provider name config_id: Configuration ID

Start a background job to rebuild embeddings. - **mode**: "existing" (re-embed items with embeddings) or "all" (embed everything) - **in

Stream the audio file associated with a podcast episode

Sync models for all configured providers. Discovers and registers models from all providers that have valid API keys configured. This is

Sync models for a specific provider. Discovers available models from the provider's API and registers any new models in the database. Ex

POST /sources with type=link and async_processing=true persists Asset(url=...).

POST /sources with type=text and async_processing=true has asset=None.

POST /sources with type=upload and async_processing=true persists Asset(file_path=...).

Azure OpenAI endpoints should be validated.

Test that RuntimeError is raised after all retries are exhausted.

Test that a transient failure is retried and succeeds.

Test that large input is split into batches of EMBEDDING_BATCH_SIZE.

Deleting credential without options cascade-deletes linked models.

Test that long CJK text is chunked using token measurement.

Test convenience function for cleaning thinking content.

Test code file extensions are treated as plain text.

Test comparing equal versions.

Test comparing when first version is greater.

Test comparing when first version is less.

Test comparing versions with pre-release tags.

Test ContentSettings has proper defaults.

Test that content type parameter is passed through.

Test ContextBuilder initialization with various params.

Test ContextConfig default values.

Test that creating a duplicate model with different case returns 400.

Test that creating a duplicate model with same case returns 400.

Test that command_id is None when save returns None (no embedding).

Test that creating a note returns the embed command_id.

Test that creating a model with same name but different provider is allowed.

Test that creating a model with same name but different type is allowed.

Test connection using a credential's configuration. Returns dict with provider, success, message keys.

Test connection using this credential's configuration.

User-set title is NOT overwritten by content_state.title.

Deleting credential with no linked models works cleanly.

Create the directory on the real filesystem.

Header splitters can emit single-char chunks; they must be filtered.

Test that empty list returns empty list.

Test that empty list raises ValueError.

Test empty string input.

Test empty text defaults to plain.

Test chunking empty text.

Test that empty text raises ValueError.

Empty string title IS replaced by extracted title.

Empty URLs should not raise (handled elsewhere).

Test segment count validation (3-20).

Test chunking with explicit HTML content type.

Test chunking with explicit Markdown content type.

Test chunking with explicit plain content type.

Test HTML extension detection.

Test Markdown extension detection.

Test that file extension takes priority over heuristics.

Test chunking with file path for content type detection.

Even if every chunk would be dropped, at least one survives.

Test that OPENAI_COMPATIBLE_BASE_URL enables all 4 modes.

Test timestamp tool returns correct format.

Test that function is properly decorated as a tool.

Test timestamp represents valid datetime.

Test getting version of non-existent package.

Test getting installed package version.

Test GitHub version fetch with invalid URL.

Test that very high confidence heuristics can override plain extension.

Test mean pooling with high-dimensional embeddings.

Test HTML detection with DOCTYPE.

Test HTML detection with structural tags.

Test HTML file extensions.

Test mean pooling with identical embeddings.

Test individual mode-specific var (EMBEDDING only).

Test individual mode-specific var (LLM only).

Test individual mode-specific var (STT only).

Test individual mode-specific var (TTS only).

URLs with invalid schemes should be rejected.

Malformed URLs should be rejected.

IPv4-mapped IPv6 addresses pointing to link-local should be rejected.

IPv4-mapped IPv6 addresses pointing to private IPs should be allowed.

IPv6 localhost should be allowed for self-hosted scenarios.

Link-local addresses should be rejected (cloud metadata protection).

Localhost should be allowed for Ollama provider.

Localhost should be allowed for self-hosted services.