Get value from cache Args: key: Cache key default: Default value if key not found

Execute a single tool call. :param server_name: Server name :param tool_name: Tool name :param arguments: Tool argume

Get all cache items

Get all cache keys

Get the global trace provider used by tracing utilities.

Record execution step

Configure only this logger, not the root logger

Create a configured ToolManager instance.

Create a configured ToolManager instance.

Create a minimal agent span for testing processors.

Render a prompt template by replacing {{variable}} placeholders with actual values. Args: template: Template string containing {{var

Create a new custom span, to which you can add your own metadata. The span will not be started automatically, you should either do `with custom_sp

Get or create a cache instance Args: cache_name: Name of the cache Returns: Fil

Create a new trace. The trace will not be started automatically; you should either use it as a context manager (`with trace(...):`) or call `

Create a new span.

Create a configured ToolManager instance.

Call a tool while maintaining the session.

Create a minimal trace.

Load domain list. Returns: List of domain names Raises: FileNotFoundError: File does not exist

Robust JSON parsing function, first try standard json, fallback to json5 if that fails Args: json_str (str): The JSON string to pars

Set value in cache Args: key: Cache key value: Value to cache

Create a configured ToolManager instance.

Create a configured ToolManager instance.

Connect to all configured servers and get their tool definitions. Returns a list suitable for passing to Prompt generators.

Parse tool_calls or <use_mcp_tool> tags from LLM response text. Returns a list containing tool call information.

Remove key from cache Args: key: Cache key Returns: True if key was removed, Fa

Create a configured ToolManager instance.

Close the underlying HTTP client.

Connect to the MCP server and initialize the session.

Finish the span. Args: reset_current: If true, the span will be reset as the current span.

Check if response is an error sentinel.

Safely parse JSON from text. First extracts JSON string, then attempts to parse as dictionary. Args: text: Text that may contai

Save data to a JSON file. Args: data: Data to save file_path: Target file path indent: Indentation spaces

Called when the application stops. We signal our thread to stop, then join it.

Load file if it exists, otherwise return default value. Args: file_path: Path to the JSON file default: Default return value

Call LLM to generate response. Args: messages: List of messages in OpenAI format max_tokens: Maximum tokens

Create a new generation span. The span will not be started automatically, you should either do `with generation_span() ...` or call `span.start()`

Generic get method for any cache Args: cache_name: Name of the cache key: Key to retrieve

Persist TaskTracer to disk. used in a finally block, thus never raise Exception.

Generic set method for any cache Args: cache_name: Name of the cache key: Key to store v

Return the current time in ISO 8601 format.

Assert that result indicates an error and contains expected error keywords.

Assert that a result indicates success and contains expected keywords.

Drains the queue and exports in batches. If force=True, export everything. Otherwise, export up to `max_batch_size` repeatedly until the queue

Save cache to file

Create a new agent span. The span will not be started automatically, you should either do `with agent_span() ...` or call `span.start()` + `span.f

Analyze browser session conversation flow

Extract JSON string from text. Finds content between the first '{' and last '}'. Args: text: Text that may contain JSON Re

Create a new function span. The span will not be started automatically, you should either do `with function_span() ...` or call `span.start()` + `

Generate response from LLM with error handling

Helper to create a completion, handling async and sync calls.

Get default criteria if generation failed

Update cumulative token usage - OpenAI implementation

Close the session and connection.

Ensure the directory for a file exists, creating it if necessary. Args: file_path: Path to the file

Extract JSON from analysis output with specific tags

Get all browser agent sessions

Guess the MIME type based on the file extension.

Load a JSON file. Args: file_path: Path to the JSON file Returns: Parsed Python object Raises: FileNotFoun

Set the global trace provider used by tracing utilities.

Set whether tracing is globally disabled.

create LLMClientProvider from hydra configuration.

Calculate and accumulate API call cost

Remove content between \\n\\nUser: and <use_mcp_tool> in assistant response (if no <use_mcp_tool>, remove to end)