Generate image embeddings from a list of base64 image strings or image URLs using batching. Returns a list of embeddings, with None f

Create initial state from request.

Load configuration from YAML file with optional override support. Args: config_path: Optional path to config file. If None, uses

Asynchronously retrieve relevant items from both text and image databases.

Generate text embeddings from a list of text strings, using chunking and batching.

Call the LLM to determine the appropriate agent for the query. Args: query: The user's query

Create the messages for the routing decision. Args: query: The user's query to route Return

Break all input texts into smaller chunks and return the chunks and their counts.

Embed all created chunks in efficient batches.

Use the LLM to determine relevant categories for the query using the search function.

Normalize agent names to match graph node names. Args: agent_name: Raw agent name from LLM

Reconstruct a single embedding for each original text from chunk embeddings.

Add timing information for a specific safety check.

Apply endpoint overrides to the RailsConfig if CONFIG_OVERRIDE is set. Args: config: RailsConfig object to modify config

Create the LangGraph for the shopping assistant. The graph orchestrates the flow between different specialized agents: - Memory retr

Embed a chunk of text.

Check if embeddings already exist in both text and image collections. Returns True if both collections have data, False otherwise.

Check if the state contains an image.

Converts an image to a base64 string.

Fetches an image from a URL, resizes and compresses it, then returns a base64-encoded string. Skips encoding if the base64 string would excee

Initialize all agent instances.

Check if the query is empty.

Simple check if a string is a path.

Simple check if a string is a URL.

Load configuration from YAML file with optional override support.

Load configuration from YAML file with optional override support. Args: base_config_path: Path to the base configuration file

Run all tests.

Fills the milvus database with the data from a CSV file. Only populates if embeddings don't already exist.

Resize a base64 image to fit within the size limits.

Generator function for streaming updates.

Initialize the PlannerAgent. Args: config: Configuration instance

Initialize the SummaryAgent with LLM configuration. Args: config: Configuration instance

Initialize the ChatterAgent with LLM configuration. Args: config: Configuration instance

Add timing information for a processing step.

Generate a response to user input using the LLM

Generate a response to user input using the LLM

Check if the user input is safe using guardrails.

Check if the generated response is safe using guardrails.

Process rail check results and update state timings.

Return the next agent to route to based on the state. This method is used by the LangGraph for conditional routing.

Route based on input safety check.