Get the global auth manager.

Handle OAuth2 token requests.

List available models from the OpenAI API. Returns: Dict containing the models data from the OpenAI API response

Convert a PDF file to a list of PIL Image objects. Args: pdf_data: Path to the PDF file or PDF data as bytes dpi: Resolution

Track a task eviction.

Update stored tasks count.

Wait for task completion with configurable polling.

Initialize the MLX LM client. Args: model_name: The name or path of the model to use (default: "mlx-community/Llama-3.2-

Extract text from a client response item which may be a str or a dict.

Format jobs list for display.

Format workspace state for inclusion in prompts.

Get local client configuration.

Save API keys to file.

Cancel a task.

Handle chat completions using the OpenAI client, but route to NotDiamond Args: messages: List of message dictionaries wi

Handle synchronous chat completions. If you pass a list of message dicts, we do one call for that entire conversation. If you pass a

Clear the conversation history but keep the secure connection

Generate text completions for the given prompts. Args: prompts: Prompt text(s) to complete **kwargs:

Convert Minions result to A2A artifact format.

Create client configuration dictionaries for local and remote clients.

Create and return a new minions instance with configured clients.

Create A2A streaming event from Minions callback.

End the secure session and clear all sensitive data

Execute a tool with the given parameters synchronously

Extract task, context, and file paths from A2A message. Legacy method for backwards compatibility. Returns:

Extract text content from a PDF file

Extract text content from a .txt file

Get the default public agent card.

Get or create the singleton instance of SGLangClient.

Get OpenAI API key from environment or session state.

Get statistics about the client pool.

Get task status and results.

Get task by ID, optionally checking ownership.

Initialize the global auth manager.

List all OAuth2 clients (without secrets).

Query the teacher model (GPT-4o).

Run all servers.

Set up streaming for a task.

Gracefully shutdown task manager.

Stop all servers.

Update active streaming sessions count.

Validate character data structure. Args: character_data: Dictionary containing character information Returns:

Validate that the story structure is complete and properly formatted. Args: story_data: Story data dictionary Retur

Wait for task completion with configurable polling.

Handle async chat with multiple messages in parallel.

Handle async chat with multiple messages in parallel.

Check if the model is available in Docker Model Runner.

Decode a base64-encoded int8 embedding to floats.

Execute Minion (singular) query.

Extract code from ```repl``` and ```python``` fenced blocks.

Extract cost information from response if available.

Extract JSON from text that may be wrapped in markdown code blocks.

Extract JSON from text that may be wrapped in markdown code blocks.

Internal PDF extraction method.

Extract text content from a PDF file

Format messages for Gemini API using the types module.

Format messages for Gemini API using the types module.

Format messages for OpenAI API format.

Format the story written so far for context.

Format namespace variables for display to the remote model.

Get API key from environment variables.

Get a FirecrawlApp client instance.

Get remote client configuration.

Initialize HTTP client for MLX Omni Server.

Initialize HTTP client for MLX Omni Server.

Check if a file should be ignored (e.g., temporary files, cache). Args: file_path: Path to check

Common generation config for both sync and async calls.

Common generation config for both sync and async calls.

Common chat options for both sync and async calls.

Common chat options for both sync and async calls.

Prepare parameters for generation. Args: messages: List of message dictionaries with 'role' and 'content' keys

Prepare tools list for generation.

Process a folder containing text, PDF, and image files. Returns: Tuple containing: - Concatenated text content from a

Process an image file and return a data URL or None if processing fails

Process Ollama tool calls and execute MCP tools.

Save OAuth2 clients to storage.

Send an encrypted message to the secure endpoint

Add entries to the Qdrant collection. :param embeddings: list of document embeddings to add, must be either TextEmbedding, ImageEmbe

Add a turn to the conversation history, removing oldest if at capacity. If summarize_older_turns is enabled, will periodically summar

Apply file changes to the workspace. Args: files: Dictionary mapping file paths to their content

Assesses the need for more information and generates a new query if needed Args: query: The original user query

Calculate a simple performance score based on metrics.

Handle chat completions using the Sarvam AI API. Args: messages: List of message dictionaries with 'role' and 'content'

Handle chat completions using the OpenRouter API. Args: messages: List of message dicts with 'role' and 'content' keys.

Handle chat completions using the OpenAI client, but route to perplexity Args: messages: List of message dictionaries w

Generate a response for a chat conversation using the MLX PARALLM model. Args: messages: List of message dictionaries, e

Handle chat completions using the MLX LM client. Args: messages: List of message dictionaries with 'role' and 'content'

Handle chat completions using the MLX Omni Server. Args: messages: List of message dictionaries with 'role' and 'content