Load the pipelines for the text classification used for routing. returns: Dict[str, Type[pipeline]]: The loaded pipelines

Generates a prompt for the agent based on the task and previous agents work information. Args: task (str): The task to be

Discard incoming audio without stopping the stream.

Notify all registered callbacks that TTS has started.

Notify all registered callbacks that TTS has stopped.

abstract method, implementation in child class. Process the prompt and return the answer of the agent.

Process the goal by dividing it into tasks and assigning them to agents. Args: goal (str): The goal to be achieved (user

Process a chunk of audio data. Args: audio_data: Raw PCM audio bytes (16-bit, mono). Returns: Dict

Return the answer with all the blocks inserted, as text.

Reads the content of a file with arbitrary encoding. Args: file_path (str): The path to the file to read Returns:

Reads the content of a file. Args: file_path (str): The path to the file to read Returns: str: The co

Read the input from the user.

Recursively searches for files in a directory and its subdirectories. Args: directory_path (str): The directory to search

Register a callback fired when TTS starts speaking.

Register a callback fired when TTS stops speaking.

Remove the reasoning block of reasoning model like deepseek.

Request the agent to stop.

Vote between the LLM router and BART model. Args: text: The input text labels: The labels to classify

Show the answer to the user.

Show the block in a pretty way.

Starts the agent process for a given task. Args: task (dict): The task to be performed. required_infos (dict

Stop capturing audio from the microphone.

Prompt for when the agent repeat itself, can happen when fail to extract a link.

Find and tick all checkboxes on the page. Returns True if successful, False if any issues occur.

Listen until the wake word is heard, then collect speech until a confirmation phrase is detected. Returns the spoken query, or None

Truncate a text to fit within the maximum context size of the model.

Resume processing of incoming audio.

Updates the plan with the results of the agents work. Args: goal (str): The goal to be achieved. agents_tasks

Wait for a submission outcome (e.g., URL change or new element).

Initialize the Speech2Text recognizer. Args: model_path: Path to local Vosk model folder. If None, downloads model for `

Initialize the LanguageUtility class args: supported_language: list of languages for translation, determine which Helsink

Initialize an agent with execution results. Args: block: The content or code block processed by the agent. f

Initialize the browser with optional AntiCaptcha installation.

A tool to perform a Google search and return information from the first result.

A tool to search for flight information using a flight number via SerpApi.

A tool for searching a SearxNG instance and extracting URLs and titles.

The file agent is a special agent for file operations.

Args: name (str): Name of the agent. prompt_path (str): Path to the prompt file for the agent. provider:

The mcp agent is a special agent for using MCPs. MCP agent will be disabled if the user does not explicitly set the MCP_FINDER_API_KE

The planner agent is a special agent that divides and conquers the task.

The casual agent is a special for casual talk to the user without specific tasks.

The Browser agent is an agent that navigate the web autonomously in search of answer

Handle generation using Ollama.

Handle generation using llama.cpp

Internal callback for the sounddevice audio stream.

Check if the AI's wake word appears in the transcript.

TTS start handler — mute mic to avoid self-echo.

TTS stop handler — resume listening.

Add a user/assistant pair to the cache if not present.

Use Anthropic to generate text.

Check all links, one by one.

Compress a text to fit within the maximum context size of the model.

Use deepseek api to generate text.

Use: xtekky/deepseek4free For free api. Api key should be set to DSK_DEEPSEEK_API_KEY This is an unofficial provider, you'll

Extract filename from path. Args: m (re.Match): Match object containing the path pattern match Returns: