A clear description of the parameters extracted from API response

[dinusha94]

Here are the extracted parameters and their descriptions from the provided code snippet:

Saving images/grids

• samples_save: Always save all generated images
• samples_format: File format for images
• samples_filename_pattern: Images filename pattern
• save_images_add_number: Add number to filename when saving
• save_images_replace_action: Saving the image to an existing file
• grid_save: Always save all generated image grids
• grid_format: File format for grids
• grid_extended_filename: Add extended info (seed, prompt) to filename when saving grid
• grid_only_if_multiple: Do not save grids consisting of one picture
• grid_prevent_empty_spots: Prevent empty spots in grid (when set to autodetect)
• grid_zip_filename_pattern: Archive filename pattern
• n_rows: Grid row count; use -1 for autodetect and 0 for it to be same as batch size
• font: Font for image grids that have text
• grid_text_active_color: Text color for image grids
• grid_text_inactive_color: Inactive text color for image grids
• grid_background_color: Background color for image grids
• save_images_before_face_restoration: Save a copy of image before doing face restoration
• save_images_before_highres_fix: Save a copy of image before applying highres fix
• save_images_before_color_correction: Save a copy of image before applying color correction to img2img results
• save_mask: For inpainting, save a copy of the greyscale mask
• save_mask_composite: For inpainting, save a masked composite
• jpeg_quality: Quality for saved jpeg images
• webp_lossless: Use lossless compression for webp images
• export_for_4chan: Save copy of large images as JPG if the file size is above the limit, or either width or height are above the limit
• img_downscale_threshold: File size limit for the above option, MB
• target_side_length: Width/height limit for the above option, in pixels
• img_max_size_mp: Maximum image size in megapixels
• use_original_name_batch: Use original name for output filename during batch process in extras tab
• use_upscaler_name_as_suffix: Use upscaler name as filename suffix in the extras tab
• save_selected_only: When using 'Save' button, only save a single selected image
• save_init_img: Save init images when using img2img
• temp_dir: Directory for temporary images; leave empty for default
• clean_temp_dir_at_start: Cleanup non-default temporary directory when starting webui
• save_incomplete_images: Save incomplete images that have been interrupted in mid-generation; even if not saved, they will still show up in webui output
• notification_audio: Play notification sound after image generation (notification.mp3 should be present in the root directory)
• notification_volume: Notification sound volume in %

Paths for saving

• outdir_samples: Output directory for images; if empty, defaults to three directories below
• outdir_txt2img_samples: Output directory for txt2img images
• outdir_img2img_samples: Output directory for img2img images
• outdir_extras_samples: Output directory for images from extras tab
• outdir_grids: Output directory for grids; if empty, defaults to two directories below
• outdir_txt2img_grids: Output directory for txt2img grids
• outdir_img2img_grids: Output directory for img2img grids
• outdir_save: Directory for saving images using the Save button
• outdir_init_images: Directory for saving init images when using img2img

Saving to a directory

• save_to_dirs: Save images to a subdirectory
• grid_save_to_dirs: Save grids to a subdirectory
• use_save_to_dirs_for_ui: When using "Save" button, save images to a subdirectory
• directories_filename_pattern: Directory name pattern
• directories_max_prompt_words: Max prompt words for [prompt_words] pattern

Upscaling

• ESRGAN_tile: Tile size for ESRGAN upscalers
• ESRGAN_tile_overlap: Tile overlap for ESRGAN upscalers
• realesrgan_enabled_models: Select which Real-ESRGAN models to show in the web UI
• dat_enabled_models: Select which DAT models to show in the web UI
• DAT_tile: Tile size for DAT upscalers
• DAT_tile_overlap: Tile overlap for DAT upscalers
• upscaler_for_img2img: Upscaler for img2img
• set_scale_by_when_changing_upscaler: Automatically set the Scale by factor based on the name of the selected Upscaler

Face restoration

• face_restoration: Restore faces using a third-party model on generation result to reconstruct faces
• face_restoration_model: Face restoration model
• code_former_weight: CodeFormer weight (0 = maximum effect; 1 = minimum effect)
• face_restoration_unload: Move face restoration model from VRAM into RAM after processing

System

• auto_launch_browser: Automatically open webui in browser on startup
• enable_console_prompts: Print prompts to console when generating with txt2img and img2img
• show_warnings: Show warnings in console
• show_gradio_deprecation_warnings: Show gradio deprecation warnings in console
• memmon_poll_rate: VRAM usage polls per second during generation (0 = disable)
• samples_log_stdout: Always print all generation info to standard output
• multiple_tqdm: Add a second progress bar to the console that shows progress for an entire job
• enable_upscale_progressbar: Show a progress bar in the console for tiled upscaling
• print_hypernet_extra: Print extra hypernetwork information to console
• list_hidden_files: Load models/files in hidden directories (directory is hidden if its name starts with ".")
• disable_mmap_load_safetensors: Disable memmapping for loading .safetensors files (fixes very slow loading speed in some cases)
• hide_ldm_prints: Prevent Stability-AI's ldm/sgm modules from printing noise to console
• dump_stacks_on_signal: Print stack traces before exiting the program with ctrl+c

API

• api_enable_requests: Allow http:// and https:// URLs for input images in API
• api_forbid_local_requests: Forbid URLs to local resources
• api_useragent: User agent for requests

Training

• unload_models_when_training: Move VAE and CLIP to RAM when training if possible (Saves VRAM)
• pin_memory: Turn on pin_memory for DataLoader (Makes training slightly faster but can increase memory usage)
• save_optimizer_state: Saves Optimizer state as separate *.optim file (Training of embedding or HN can be resumed with the matching optim file)
• save_training_settings_to_txt: Save textual inversion and hypernet settings to a text file whenever training starts
• dataset_filename_word_regex: Filename word regex
• dataset_filename_join_string: Filename join string
• training_image_repeats_per_epoch: Number of repeats for a single input image per epoch (used only for displaying epoch number)
• training_write_csv_every: Save a csv containing the loss to log directory every N steps (0 to disable)
• training_xattention_optimizations: Use cross attention optimizations while training
• training_enable_tensorboard: Enable tensorboard logging
• training_tensorboard_save_images: Save generated images within tensorboard
• training_tensorboard_flush_every: How often, in seconds, to flush the pending tensorboard events and summaries to disk

Stable Diffusion

• sd_model_checkpoint: Stable Diffusion checkpoint
• sd_checkpoints_limit: Maximum number of checkpoints loaded at the same time
• sd_checkpoints_keep_in_cpu: Only keep one model on device (will keep models other than the currently used one in RAM rather than VRAM)
• sd_checkpoint_cache: Checkpoints to cache in RAM (obsolete; set to 0 and use the two settings above instead)
• sd_unet: SD Unet (choose Unet model: Automatic = use one with same filename as checkpoint; None = use Unet from checkpoint)
• enable_quantization: Enable quantization in K samplers for sharper and cleaner results (This may change existing seeds)
• emphasis: Emphasis mode (makes it possible to make model to pay more or less attention to text when you use the syntax in prompt)
• enable_batch_seeds: Make K-diffusion samplers produce same images in a batch as when making a single image
• comma_padding_backtrack: Prompt word wrap length limit (in tokens - for texts shorter than specified, if they don't fit into 75 token limit, move them to the next 75 token chunk)
• CLIP_stop_at_last_layers: Clip skip (ignore last layers of CLIP network; 1 ignores none,

Sure, here is the section describing the VAE:

VAE

• sd_vae_explanation:

  VAE (Variational autoencoder) is a neural network that transforms a standard RGB image into latent space representation and back. Latent space representation is what stable diffusion is working on during sampling (i.e. when the progress bar is between empty and full). For txt2img, VAE is used to create a resulting image after the sampling is finished. For img2img, VAE is used to process user's input image before the sampling, and to create an image after sampling.
  

• sd_vae_checkpoint_cache: VAE Checkpoints to cache in RAM
• sd_vae: SD VAE (choose VAE model: Automatic = use one with same filename as checkpoint; None = use VAE from checkpoint)
• sd_vae_overrides_per_model_preferences: Selected VAE overrides per-model preferences (you can set per-model VAE either by editing user metadata for checkpoints, or by making the VAE have same name as checkpoint)
• auto_vae_precision_bfloat16: Automatically convert VAE to bfloat16 (triggers when a tensor with NaNs is produced in VAE; disabling the option in this case will result in a black square image; if enabled, overrides the option below)
• auto_vae_precision: Automatically revert VAE to 32-bit floats (triggers when a tensor with NaNs is produced in VAE; disabling the option in this case will result in a black square image)
• sd_vae_encode_method: VAE type for encode (method to encode image to latent (use in img2img, hires-fix or inpaint mask))
• sd_vae_decode_method: VAE type for decode (method to decode latent to image)

[asmrtfm]

What does " Forbid URLs to local resources" actually mean though?

[w-e-w]

api image input allows you to use a url the webui will fetch as the input image
becuase it is the server that is doing the fetching of the image this could lead to a Server-side request forgery (SSRF) attack

is is possible you could be hosting a web server in your local network
this server is only meant to be accessed by someone inside the local network

now let's say if you launch your webui with --share makeing it accessible from outside of your local network
then some one can connect to your webui and make an API request with an URL to the internal server, awebui will then make a request to internal server, since it also in the local network they request is accepted returning with the image, would then process the image depending on the API you're using then returning the results to the external user

even though the risk is generally quite low as the person who's making a request has to know that there exist an internal web server, and know the valid URL that allows them to retring the image file, and image file only, it is still potentially undesirable

the Forbid URLs to local resources option toggles the behavior whether or not a request to internal server is allowed from API