Tokenize a batch of responses.

If we see subprocess.run(..., shell=True), or similar calls, flag it as especially dangerous.

Pad responses for non-active examples with pad tokens.

Extract frames at uniform intervals

Convert image to base64 string

Wrapper for generation that handles multi-GPU padding requirements. if num_gpus <= 1, return self.actor_rollout_wg.generate_s

Extract the 'mode' argument if the user calls open(..., mode='w'), etc.

Get video duration in seconds

Check if file type is supported

获取文件锁

获取文件锁

Process responses to stop at search operation or answer operation.

Retry with exponential backoff

读取缓存文件

读取缓存文件

We discovered something like <object>.to_csv(...), so we mark it as potentially dangerous since it writes to disk.

Return one response generator based on the received messages. The workflow for an agent to generate a reply. Each agent subclass need

释放文件锁

释放文件锁

Read webpage content using Aidata service. Args: url: The URL to read Returns:

LLM chat interface. Args: messages: Inputted messages. functions: Inputted functions for function calling. OpenAI for

Scans Python code for certain banned keywords (exit, yield, requests, url, pip, install, conda) as real code tokens, ignoring strings/comment

Parse the given Python AST node, and detect known dangerous function calls or file operations. Returns (is_safe, message).

Args: res (str): markdown content Returns: str: cleaned markdown content

Cut tensors to their effective length based on attention mask.

Execute predictions across multiple environments. NOTE: the function is the actual `step` function in the environment NOTE pe

Args: html (str): html content Returns: str: clickable buttons

Args: start_tag (str): start_tag end_tag (str): end_tag text (str): complete sentence Returns: str: the cont

Args: url (str): url screentshot (bool): whether to take a screenshot Returns: str: html content and cleaned mar

Process (text-based) predictions from llm into actions and validity flags. Args: predictions: List of raw predic

Attempt to read webpage content by alternating between jina and aidata services. Args: url: The URL to read

Attempt to read webpage content by alternating between jina and aidata services. Args: url: The URL to read