Returns a string hash of the system configuration applicable to be a path. This is route results with different parameters to differe

Hovers the mouse over the specified coordinates. Args: page (Page): The Playwright page object. x (float): T

Initialize the browser, context, and page.

Check if obj is a valid JSON schema describing a tool compatible with OpenAI's tool calling. Example valid schema: { "name": "get_c

Press specified keys in sequence. Args: page (Page): The Playwright page object keys (List[str]): List of ke

Call ``client.create(messages=...)`` and parse the response as a JSON object containing every name in ``required_keys``. Retries up to ``max_

Uses an LLM to evaluate the candidate answer versus either the gold truth answer (if provided) or the task itself, returning a JSON object wi

Dynamically load a benchmark class based on the benchmark name.

Log the fn_call_template mapping as an MLflow tag.

CLI entry point for fara command.

Remove old screenshots from the chat history. Assuming we have not yet added the current screenshot message. Note: Original user messages (ma

Convert our LLMMessage to OpenAI API format

Lower text and remove punctuation, articles, and extra whitespace.

Parse events where thoughts and actions are embedded as text in the message field. Expected format: "Thought #X: <thought text> Acti

Clean an HTML element string, keeping only salient_attributes.

this is needed in the case of gpt_solver, where the action names are not consistent with those of trained models

Render the structured CP classification as a prompt-injection block.

Resize the image

Stub: the optional tool registry used for action-schema validation is not shipped with this package. ``MMRubricAgentConfig.tools`` is optiona

Run evaluation for a single example with retry logic. Args: example (dict): Example data. folder (str): Output folder.

Run and evaluate the benchmark. Args: benchmark_name (str): Benchmark name. system_name (str): System name. parallel

Return the decision-rule block for a given critical-point type. Args: cp_type: the ``critical_point_type`` from the classifier. ``None``

Set the download handler for the browser.

Rescales the image so that the following conditions are met: 1. Both dimensions (height and width) are divisible by 'factor'. 2. The to

Group events into steps based on action_id linkage. Args: with_tail_observations: If True (default), include a trailing step

Stop the Xvfb process if it's running.

Convert PIL image to base64 string

Truncate text keeping first and last halves with a marker between. Inlined from agento_next's ``agents/utils.truncate_observation`` so this m

Run Steps 9a, 9b, and 10 on top of a scored rubric. Args: rubric_dict: Scored rubric (output of :meth:`MMRubricAg

Wait for captcha to be resolved if one is being solved.

Wait for captcha to be solved with timeout