Clean up temporary test file.

Create a temporary test file for upload testing.

Encode audio file to base64 and determine format.

Use tiktoken to estimate token count of text

Generate random message ID using common LLM format

Get audio duration in seconds. Tries to use wave (for .wav), then falls back to mutagen (for mp3, etc).

Determine the appropriate audio file extension from URL or content type. Args: url: The URL of the audio file content_type:

Get default equal weights if generation failed

Unified LLM call and logging handling Returns: tuple[Optional[str], bool, Optional[object]]: (response_text, should_break, tool_ca

Parse new format tool name Args: tool_name: New format tool name, for example: - "tool-server_name

Check if we should block scraping of Hugging Face datasets/spaces. :param tool_name: The name of the tool being called :param

assume using default trace provider, which generates span_id in the format: input = f"span_{uuid.uuid4().hex[:24]}"

General key pattern-based parsing method

Convert an ISO 8601 timestamp to Unix nanoseconds. :param timestamp_in_iso: The timestamp in ISO 8601 format. :return: The timestamp in U

Adds a new trace processor. This processor will receive all traces/spans.

Analyze conversation flow, including tool calls

Configure only this logger, not the root logger

Call Claude vision API.

Clear all cache data

Define and return MCP server configuration list

Call LLM to generate response, supports tool calls - unified implementation

Creates and initializes the core components of the agent pipeline. Args: cfg: The Hydra configuration object. Returns:

Executes the full pipeline for a single task. Args: cfg: The Hydra configuration object. task_description: The description o

Extract JSON from LLM response

Extract text from message content

Format tool execution results to be fed back to LLM as user messages. Only includes necessary information (results or errors).

Generate a new span identifier.

Generate a new trace identifier.

Generate tasks for all experts. Args: experts: List of expert info Returns: List of tasks, each con

Get statistics about cache usage

Return the currently active span, if any.

Return the currently active trace, if any.

Get main agent message history

Get main agent message list

Parse MCP tool call

Process LLM response - implemented by subclass

Save data to a JSONL file. Args: data: List of dictionaries file_path: Target file path

Set the OpenAI API key for the exporter. Args: api_key: The OpenAI API key to use. This is the same key used by the OpenAI Python

Replace the list of processors with ``processors``.

Set the list of processors. This will replace the current list of processors.

Set the OpenAI API key for the backend exporter.

Clean up any resources used by the provider.

implementation from https://github.com/dimaqq/otlp-json/blob/e7490fe003309a1b3e8d238f6fc77362a57c220a/otlp_json/__init__.py#L112 definti

Apply cache control to the last user message and system message (if applicable)

Apply cache control to the last user message and system message (if applicable)

Build dictionary of all dimensions with their definitions

Calculate summary statistics across all queries

Conservative fallback strategy: only fix the most obvious issues

Convert message list to text format required by Response API

Convert Response API response to a serializable format

Create specific LLM client

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

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

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

Create a generator that yields delays for exponential backoff. :param max_retry_count: Maximum number of retries. :param base_delay: Base

Create provider-specific message - implemented by subclass

Helper to create a response using OpenAI's Response API.

Extract content from \\boxed{} patterns in the text. Uses safe regex patterns to avoid catastrophic backtracking. Returns the

Extract query ID from filename

Extract response content from result data

Filter message history, keep specified number of tool results

Find the end position of value

Format a single task for final output.

Get Anthropic API credentials, skip test if not available.

Get Serper API credentials, skip test if not available.