Wait for sandbox to become ready. Args: max_wait: Maximum seconds to wait Raises: TimeoutError: If

Apply Z-score normalization: z = (x - mean) / std.

Instantiate the quality-checker chain from config. Reads ``config.self_reflection.quality_check`` and returns a list of checker instances. C

Build ExaSearchRequest from tool call arguments.

Chunk the document using the `Hybrid Chunker`. Args: docs (Iterable[DoclingDocument]): An iterable of DoclingDocument ob

Clean up resources

Cleanup resources used by the tool manager.

Initialize LSP servers

Create a new session

Remove duplicate results based on URL. Args: results: List of search result dictionaries Returns: D

Dump a batch of search results to a local file.

Execute the skill DAG according to execution order. Execution order format: [skill1, skill2, [skill3, skill4], skill5, ...]

Execute JavaScript code via Node.js. Uses sandbox mode or local mode based on use_sandbox setting. Args: script

Execute a Python script file. Uses sandbox mode or local mode based on use_sandbox setting. Args: script_path:

Execute a shell command. Uses sandbox mode or local mode based on use_sandbox setting. Args: command: Shell com

Extract key information from URLs or files. Args: urls_or_files: List of URLs or file paths to process Returns:

Extract code blocks from the given text. <result>py:a.py </result> Args: text: The text to extract code blocks from. tar

Retrieve an analysis by ID.

Get the caption text for a given caption reference in the document.

Get current progress state safely.

Get deep research configuration

Get EdgeOne Pages configuration

Get edit_file_config configuration

Get environment variables for running agents

Get font using system font manager, same as CreateBackground agent

Get process-wide summarization usage totals (best-effort).

Get the reference key for a DocItem in the format of 'doc_file_name@binary_hash@self_ref'.

Get all messages for a session

Get the next available rId for relationships files.

Retrieve a note by ID.

Get search engine class by type. Args: engine_type: One of 'exa', 'serpapi', 'arxiv' Returns: SearchEngine class (not i

Get the corresponding search request generator Args: engine_type (SearchEngineType): The type of search engine user_prompt (

Return the path that stores the per-user FinResearch session snapshot.

Get the web search tool based on the configuration. Returns: SearchEngine: An instance of the SearchEngine class configured with the

Check if the Docker daemon is running.

Check if messages exceed context limit.

Check if user has an active task running.

Link output from upstream skill to downstream skill input. Args: upstream_skill_id: ID of the upstream skill.

List notes with optional filters. Args: task_id: Optional: Filter by task ID. tags: Optional: Filter by tags (notes m

Load key=value pairs from a .env file into ``os.environ``. Does not override variables already set in the process environment. If ``

Load search results from a JSON file.

Load image from URI as a PIL Image object and extract its format extension. URI format: data:[<mime>][;base64],<encoded> Args: u

Loads an image from a given URL and converts it into a PIL Image object in memory. Args: url: The URL of the image. Returns:

Load and return the full evidence index. Returns the complete index containing metadata for all notes, useful for understand

Load specific references by name, or all if names is None. Args: names: List of reference names to load, or None for all

Load specific resources by name, or all if names is None. Args: names: List of resource names to load, or None for all.

Load session data from file

Load requested specs documents and return their content. Returns: str: Markdown string (default) or JSON string mapping spec to c

Mount a skill directory for sandbox access. Args: skill_id: Unique identifier for the skill. skill_dir: Path

Open a document in the LSP server

Pack a directory into an Office file (.docx/.pptx/.xlsx). Args: input_dir: Path to unpacked Office document directory output_file

Initialize the LLM model from the configuration.

Reads configuration information and conversation history associated with the given task from the cache directory. This function attempts to

Removes all <resource_info>...</resource_info> tags and their enclosed content from the given text. Args: text (str): The original t

Resolve file_path against root_dir_abs. - if file_path starts with 'projects/', resolve from ms-agent base dir - if file_path is absolute

Public interface for running the research workflow with proper resource cleanup.

Run the agent, mainly contains a llm calling and tool calling loop. If skills are configured, skill-related queries will be automati

Saves the specified configuration and conversation history to a cache directory for later retrieval or restoration. This function saves the

Save current chat history to disk for future resuming. Args: messages (List[Message]): Current message history to save.

Save the execution spec to a markdown file in logs directory.