Manually converts a Dataplex Entry object to a dictionary that is compatible with the existing simulation logic.

Evaluate all reference glasses against the generated VTO image. Uses the first glasses image as the primary reference. Returns: dict

Runs a sub-agent by spinning up a temporary async loop in a SEPARATE THREAD. Args: agent (LlmAgent): The sub-agent to run. pr

Export the spans to Google Cloud Logging and Cloud Trace. :param spans: A sequence of spans to export :return: The result of

Extracts structured guideline data from a document using Gemini.

Extract a specific rule (e.g., "4.1", "33.1", "12.1"). Handles formats: - #### [Rule 4.1: WAF Exceptions] - ###

Retrieves official historical data points from Anthropic's Economic Index Reports (from August 2025, November 2025, and February 2026).

Fetches live labor statistics from the BLS (Bureau of Labor Statistics) API v2.

Fetches list of URLs from the API.

Fetches real-time average electricity prices per kWh from the EIA Open Data API. Crucial for calculating the operational ROI of data centers

Filter a list of (mask, score) tuples, keeping only those in the top 30% of scores. Ensures a minimum number of masks are always returned, even i

Find the best matching frame for each reference using SSIM. Then pick 3 frames: best match, frame at -offset, and frame at +offset. Args:

Consolidated tool that generates a presentation outline AND saves it to session state. Use this instead of calling outline_specialist and sav

Generates a detailed radar chart for the asset evaluation.

Generate a virtual try-on image using 2-step approach (always full body). Args: client: Gemini client instance scenario: Descript

Get the GCS blob path for a generated asset.

Generate a background mask using the rembg library.

Recursively gets the code for a Pydantic type and its dependencies. Args: base_type: The Pydantic type to start from. Returns:

Recursively gets the code for a Pydantic type and its dependencies. Args: base_type: The Pydantic type to start from. Returns:

Get or create the shared evaluation thread pool.

Get events from the calendar within a time range.

Return mapping of scope -> [rule_id, ...].

Generate analytics query for BigQuery using ADK tools.

Return an MCP toolset connected to Google Maps Grounding Lite. The toolset includes a tool called `search_places` that can be used to search

Get all file paths for a given run ID. Args: run_id: The unique identifier for the run. base_dir: The base directory where run files

Intelligently maps a requested layout name to the best available layout in the current presentation.

Get MCP Toolbox client.

Simulates a web search. Use it get information on weather. Args: query: A string containing the location to get weather information for.

Initializes and returns the global Vertex AI GenAI client.

Converts LaTeX-style math blocks frequently returned by Deep Research into plain Unicode text for better readability in PowerPoint.

Return sorted list of available case IDs.

Load all agent processing outputs for a case. Note: The general agent does NOT consume any preprocessing/metadata files. It only processes

Load evaluation set configuration.

Load and return the current rule_base.json.

maskSecret replaces most of the string with asterisks for logging.

Extract markdown from a PDF file. Splits into pages and sends to Gemini in parallel.

Extract markdown from an HTML file. 1. Read and clean HTML with BeautifulSoup 2. Check body has enough content (50+ words, 2+ sentences)

Render a JSON/JSONL file as markdown for downstream chunking.

Parse key-value pairs from a comma-separated KEY=VALUE string.

Parse key-value pairs from a comma-separated KEY=VALUE string.

Parse key-value pairs from a comma-separated KEY=VALUE string.

Parse key-value pairs from a comma-separated KEY=VALUE string.

Preprocess face image: crop face, upscale, then remove background. Args: client: Gemini client instance for background removal up

Preprocess face image: crop face, upscale, then remove background. Returns: tuple: (reference_face, preprocessed_face) - refe

Intelligently populates a slide based on its layout and the spec's content.

resolveTripID looks up the database PK from the ADK Session ID.

Revise a rule based on SME feedback, then run safety loop. 1. Revises the rule via LLM using SME feedback 2. Runs the same safety loo

Write a rule to rule_base.json. Args: rule_dict: Complete rule dict matching ALF schema. mode: "add" (n

Animate a ready model image into catwalk-style videos. Takes an image of a model already wearing garments and generates videos using Veo 3.1

Runs econometric analysis (OLS, Correlation Matrix, or ADF Stationarity Test) on numerical data. Args: dependent_values: List of

Run the full image VTO pipeline. Yields results as each variation completes. Yield order: 1. A single VTOResult with status="reference_fac

Run the clothes video VTO R2V pipeline. Uses Veo 3.1 R2V mode with three reference images at different framings (lower body, upper body, face

Embed a query and perform vector similarity search. Args: query: The query string. k: Number of results to return. filter

Helper to send a message and get the final text response.

setupTestDB connects to the test database and returns the connection. It skips the test if the connection fails.

Surgically updates a slide in the session STATE. Automatically grows the deck if needed.

Validate that the rule base is internally consistent. Checks: 1. No duplicate rule IDs 2. No conflicting rules

CreateArtifact saves artifact metadata to the database.

DegreesToDirection converts a compass bearing in degrees to a cardinal direction string (e.g. "N", "NNE").

GetArtifact retrieves artifact metadata by filename.

GetTrip retrieves the trip details.

LoadConfig loads configuration from environment variables.