bob_flux2.flux2_node
====================

.. py:module:: bob_flux2.flux2_node


Classes
-------

.. autoapisummary::

   bob_flux2.flux2_node.TTInode


Functions
---------

.. autoapisummary::

   bob_flux2.flux2_node.main


Module Contents
---------------

.. py:class:: TTInode

   Bases: :py:obj:`rclpy.node.Node`


   A ROS node that interfaces with a FLUX.2 model for text-to-image generation.

   This node manages the entire text-to-image pipeline. It is optimized for
   low-VRAM environments by sequentially loading and unloading the text encoder
   and the image generation models for each request, ensuring they are never
   in memory simultaneously. It can be configured to run once for an initial
   prompt or continuously listen for prompts on a topic. It also supports
   an optional input image to guide the generation process.


   .. py:attribute:: bridge


   .. py:attribute:: repo_id


   .. py:attribute:: model_dir


   .. py:attribute:: device


   .. py:attribute:: image_path


   .. py:attribute:: prompt


   .. py:attribute:: seed


   .. py:attribute:: input_image


   .. py:attribute:: once


   .. py:attribute:: image_counter


   .. py:attribute:: torch_dtype


   .. py:attribute:: hf_kwargs


   .. py:attribute:: subscription


   .. py:attribute:: image_pub


   .. py:method:: _get_image_output_path()

      Determines the output path for the generated image.

      If the 'image_path' parameter ends with "auto", it generates a unique
      filename with a random 10-character string in the specified directory.
      Otherwise, it returns the 'image_path' parameter directly.

      Returns:
          str: The final path to save the image file.



   .. py:method:: prompt_callback(msg)

      Processes an incoming prompt to generate and save an image.

      This is the core callback function that executes the full text-to-image
      pipeline in a memory-efficient manner. It performs the following steps:
      1. Loads the text encoder model.
      2. Encodes the incoming prompt string into tensor embeddings.
      3. Unloads the text encoder to free VRAM.
      4. Loads the image generation pipeline.
      5. If an input image is specified, loads it.
      6. Generates an image from the prompt embeddings (and optional input image).
      7. Saves the image to the configured file path.
      8. Unloads the image generation pipeline to free VRAM.
      9. If the 'once' parameter is set, it shuts down the node.

      Args:
          msg (std_msgs.msg.String): The message containing the text prompt.



.. py:function:: main(args=None)

   The main entry point for the ROS node.