Return the assistant response to a user message if cached.

Get the current URL of the page.

Return the final recognition result, flushing the recognizer.

Get the title of the current page.

Return the current interim recognition result.

Get the answer from the last agent.

Get the answer from the last agent.

Use google gemini to generate text.

Use huggingface to generate text.

Provide feedback based on the output of the code execution

Provide feedback based on the output of the code execution.

Provide feedback based on the output of the bash interpreter

Feedback of web search to agent.

Not really needed for this tool (use return of execute() directly)

Provide feedback based on the output of the code execution

Provide feedback based on the output of the code execution

Check if a user msg is cached.

Return True if the audio stream is currently active.

Return a list of available audio input devices.

Use local lm-studio server to generate text.

Use local or remote Ollama server to generate text.

Use openai to generate text.

Use OpenRouter API to generate text.

Process the user prompt to conduct an autonomous web search. Start with a google search with searxng using web_search tool. T

Replace URL with domain name or empty string if IP address. Args: url (re.Match): Match object containing the URL pattern

Reset the recognizer state, clearing any partial result.

Save a JSON file.

Scroll to the bottom of the page.

Use a remote server with LLM to generate text.

Set up test fixtures before each test method.

Dynamically update the recognition grammar.

Set the query

Find long paragraph like **explanation**: <long text> by keeping only the first sentence. Args: sentence (str): The sente

Ask the LLM to process the prompt and return the answer and the reasoning.

Test that trailing slash is stripped since it's not alphanumeric.

Test that matching major versions are compatible.

Test folder creation.

Test folder creation when folder already exists.

Test with empty address