Simulate AI response in conversation.

Start the server

Translate a natural language question to a semantic query JSON.

Print text with typewriter effect

Build a single training example with all rubric features.

Extract human-readable text from schema JSON, avoiding noise like URLs and JSON syntax.

Find items that match the requested item using parallel LLM calls.

Extract inputs from batch and run forward pass.

Get the Azure AI Search index client for managing indices

Parse schema.org JSON, handling lists and errors.

Resolve relative paths to absolute paths. Args: path: The path to resolve Returns: str: Absolute pa

Send an error message to the user.

Send message when items cannot be found for comparison.

Send message when no matching accompaniments are found.

Send message when no matching conversations are found.

Generate a hashable representation for potentially nested data structures.

Build a complete URL to the NLWeb gateway with query parameters. Args: site_url: The site domain (will be cleaned of https:// if pre

Check if a package is installed.

Create the OpenSearch index with proper kNN vector mapping if it doesn't exist. Args: index_name: Optional index name (d

Log a debug message asynchronously.

Delete the OpenSearch index. Args: index_name: Optional index name (defaults to configured index name) Returns:

Expand test cases that have llm_provider: "all" to test all available providers. Args: test_cases: Original test cases

* Extracts an image URL from various image formats * * @param {*} image - The image data * @returns {string|null} - The image URL or null

Extract @type from schema.org JSON.

Create Message from dictionary.

Generate improved description using specified strategy.

* Get all messages for a conversation

Get all environment variables and their current values

Get a list of LLM providers that have their required API keys available. Returns: List of provider names that are available for use.

Get or create the global handler instance

Get or create a logger instance with caching. Args: name: Logger name level_env_var: Environment variable name for log level

Get the model task type for the configured inference endpoint (self._model).

Get next work item using round-robin across models for fairness. Round-robin ensures we cycle through models fairly when selecting, b

Get latency percentiles for operation.

Get handler statistics

Get tools for a specific schema type, including inherited tools from parent types.

Log an info message asynchronously.

Initialize the Qdrant client and create collection if needed.

Check if query analysis is enabled.

Check if decontextualization is enabled.

Join a conversation.

Load data from a file with precomputed embeddings into the database. The file should have 3 columns per row, separated by tabs: 1. URL f

Log an unknown provider configuration

Log an unknown provider configuration

Normalize scalar features to zero mean, unit variance. Returns normalized features and stats dict (for inference).

Parse an RSS 2.0 feed into Schema.org format. Args: root: XML root element feed_url: URL of the feed Returns: L

Extract handcrafted features from examples as tensors.

Print test summary in a formatted way.

Print overall summary of all queries

Process a work item with semaphore control.

Read lines from a file, handling different encodings. Args: file_path: Path to the file Returns: List of lines from the

* Registers all type renderers with a JSON renderer * * @param {JsonRenderer} jsonRenderer - The JSON renderer to register with

* Registers a renderer for a specific type * * @param {string} type - The type to register for * @param {Function} renderer - The renderer f

* Renders JSON as HTML for display * * @param {Object|string} json - The JSON to render * @returns {string} - HTML representation of the JSO

* Sanitize a URL to prevent dangerous protocols

Send a search request to a Cortex Search Service which has the columns URL and SCHEMA. See: https://docs.snowflake.com/developer-guide/s

Search across all sites using a simplified interface. Args: query: The search query top_n: Number of results to return (defa

Search across all sites. Args: query: Search query string num_results: Maximum number of results to return

Send message through authenticated WebSocket.

Send typing indicator.

Set the logging verbosity level.

Check if a site supports standard vector database retrieval

Summarize query results into natural language.

Upload documents to the database. Args: documents: List of document objects **kwargs: Additional parameters

Wait for a specific type of message.

Wait for a specific type of message.

Write a named SSE event (v0.55 format): event: X\\ndata: {...}\\n\\n