getimg_client 0.1.2 → 0.1.4

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: bebe5b9162ab45505ab6f74e4b816f9c14cee51734a12a8baf8e273824acae23
-  data.tar.gz: 506ab1f2ff553a225dc93fd0e5bd039c68a574d7c6f6c2a2a7da819940358cf3
+  metadata.gz: df68c244073fbc74c6cb913c4b9208c40a2d136d226a1974dec6addae2bc4682
+  data.tar.gz: 641715c9f551b30eb44f06252be3ded05df990676e5260cff8e89893d3b0ca77
 SHA512:
-  metadata.gz: 251e4fef81fbaa5a0c6bcb5d9b2bd5b26717db6b7af29df3208461041ad02d07b74539256c24dc2bb080afd354e5ba58f48e253896781d8e186e5cc34f8e7dcd
-  data.tar.gz: 972aa553495c690c9e30d3eb45a1d86de972639235b3c4fa1e8916c635540b99ed662ac1579d8cb6e60a69326684541745e63184344647a1274475337bfaf62b
+  metadata.gz: 5249e227172db1f03aeced82d4a8f6429b1ad75c6eaa2c51d0285000e0f324b9b14ef833bc638949dd35e3aef03d1ab99c93049a0de65bebe57c58d8f9d35f97
+  data.tar.gz: dec0e93875f2d227ac3a7128b57d7e65f7f3d8b334f7f4c379f7a4cb87b1c6aa0672789979b950a65cd5f96bc6fabb2fb0cf9f613d8596147b8449dffbb7b9dd

data/README.md

@@ -46,14 +46,14 @@ Fetch and cache the list of available models from the API.
 ### `models`
 Retrieve cached models or fetch them if not already retrieved.
 
-### `generate_image(prompt, model:, **options)`
-Generate an image based on a text prompt and specified model. Supports various image manipulation methods such as image-to-image, inpainting, controlnet, face fixing, instruct and upscaling.
+### `generate_image(prompt, model: :essential, **options)`
+Generate an image based on a text prompt and specified model. Supports various image manipulation methods such as image-to-image, inpainting, controlnet, face fixing, instruct, and upscaling. Specifying a model is optional; by default, it will invoke the Essential V1 endpoint.
 
 ### `get_balance`
 Retrieve the current account balance from the Getimg.ai API.
 
 ## Request Routing
-The `generate_image` method routes requests based on the provided options and the model's supported pipelines, sending it to its respective pipeline whether that's SD1.5, SDXL or LCM based:
+The `generate_image` method routes requests based on the provided options and the model's supported pipelines, sending it to its respective pipeline whether that's SD1.5, SDXL, Essential, or LCM based:
 - **text-to-image**: Default if no images are provided.
 - **image-to-image**: Triggered if a base image is provided.
 - **inpaint**: Triggered if a mask image is provided.
@@ -82,8 +82,12 @@ If an HTTP error occurs, the error message will include both the HTTP status and
 ## Models
 The provided `model` option will determine the model use, and contribute to the client's inference of the desired endpoint. Models can be the string `id` listed online at [the GetImg dashboard](https://dashboard.getimg.ai/models) or retrieved using the `GetimgClient.models` method. Equally, you can use symbols instead. For example, `:realistic_vision_v5_1` will translate to `'realistic-vision-v5-1'`
 
+## Essential
+Getimg offers the "essential" checkpoints, which perform more abstract Stable Diffusion operations based on the provided prompt. Note that *:essential* and *:essential_v2* will *not* be listed in the model listing, as these are not in fact actual models, nor valid "model" value at the API's end.
+In order to route a request to Essential or Essential V2 however, you can provide *:essential* or *:essential_v2* as a model argument in *#generate_image*.
+
 ## Latent Consistency Models (LCM)
-Latent consistency models are designed to speed up image generation by circumventing the iterative generation process of diffusion-based methods. Building on direct consistency models that operate on image pixels, latent consistency models operate in the latent (lower-dimensional) space. The client will automatically route requests to these models as needed.
+Latent Consistency Models (LCM) are optimized for faster image generation and lower costs by avoiding the repetitive steps of traditional diffusion methods. They work in a lower-dimensional space, resulting in quicker outputs but with slightly less detail compared to standard models.
 
 ## Base Image Options
 The `base_image` option can be provided, which will automatically set the "image" property for image-to-image, controlnet, face-fix, upscale, inpaint and instruct calls, and can be either a file path or a base64 encoded string of the file contents.
@@ -98,7 +102,13 @@ If all went well and no errors were reported, the response will be equal to the
 GetimgClient.set_api_key('your_api_key')
 ```
 
-### Generate Text-to-Image
+### Generate Text-to-Image using Essential V1 with default options
+```ruby
+result = GetimgClient.generate_image("A city skyline at night")
+puts result["image"]
+```
+
+### Generate Text-to-Image using Stable Diffusion 1.5
 ```ruby
 result = GetimgClient.generate_image("A scenic landscape", model: :stable_diffusion_v1_5, width: 512, height: 512)
 puts result["image"]
@@ -114,7 +124,7 @@ puts result["image"]
 ### Generate an image using Controlnet
 ```ruby
 base_image = "path/to/base_image.jpg"
-result = GetimgClient.generate_image("Enhance the scene", model: :stable_diffusion_v1_5, base_image:, strength: 0.7, controlnet: canny-1.1)
+result = GetimgClient.generate_image("Enhance the scene", model: :stable_diffusion_v1_5, base_image:, strength: 0.7, controlnet: 'canny-1.1')
 puts result["image"]
 ```
 

data/lib/endpoints.json

@@ -10,6 +10,7 @@
     "lcm_text_to_image": "https://api.getimg.ai/v1/latent-consistency/text-to-image",
     "lcm_image_to_image": "https://api.getimg.ai/v1/latent-consistency/image-to-image",
     "essential_text_to_image": "https://api.getimg.ai/v1/essential/text-to-image",
+    "essentialv2_text_to_image": "https://api.getimg.ai/v1/essential-v2/text-to-image",
     "face_fix": "https://api.getimg.ai/v1/enhancements/face-fix",
     "upscale": "https://api.getimg.ai/v1/enhancements/upscale",
     "models": "https://api.getimg.ai/v1/models",

data/lib/getimg_client.rb

@@ -32,7 +32,7 @@ class GetimgClient
   end
 
   # Generate an image based on the prompt and model
-  def self.generate_image(prompt, model:, **options)
+  def self.generate_image(prompt, model: :essential, **options)
     # Validate the prompt
     raise ArgumentError, 'Prompt is required' unless prompt.is_a?(String) && !prompt.strip.empty?
 
@@ -43,7 +43,11 @@ class GetimgClient
 
     # Convert model to appropriate ID format
     model_id = model.to_s.gsub('_', '-')
-    model_info = models[model_id] || raise(ArgumentError, "Unknown model: #{model}")
+    if model_id === 'essential' || model_id === 'essential-v2'
+      model_info = models[model_id] || { name: model_id, pipelines: ['text-to-image'] }
+    else
+      model_info = models[model_id] || raise(ArgumentError, "Unknown model: #{model}")
+    end
 
     # Determine the requested pipeline
     requested_pipeline = method_requested(base_image, mask_image_path, controlnet, model_info).to_s.gsub('_', '-')
@@ -54,7 +58,7 @@ class GetimgClient
     end
 
     # Determine the endpoint key based on the model family and pipeline
-    endpoint_key = determine_endpoint_key(model_info, requested_pipeline)
+    endpoint_key = determine_endpoint_key(model_info, requested_pipeline, model_id)
     uri = URI(API_ENDPOINTS[endpoint_key.to_s])
 
     # Handle image to image, controlnet, instruct, inpaint, face-fix, and upscale pipelines
@@ -113,8 +117,12 @@ class GetimgClient
   end
 
   # Determine the appropriate endpoint key based on the model family and pipeline
-  def self.determine_endpoint_key(model_info, requested_pipeline)
-    if model_info[:family] == 'stable-diffusion-xl'
+  def self.determine_endpoint_key(model_info, requested_pipeline, model_id)
+    if model_id == 'essential'
+      return :essential_text_to_image
+    elsif model_id == 'essential-v2'
+      return :essentialv2_text_to_image
+    elsif model_info[:family] == 'stable-diffusion-xl'
       return :sdxl_image_to_image if requested_pipeline == 'image-to-image'
       return :sdxl_inpaint if requested_pipeline == 'inpaint'
       return :sdxl_text_to_image
@@ -133,10 +141,8 @@ class GetimgClient
     request['accept'] = 'application/json'
     request['content-type'] = 'application/json'
 
-    body = {
-      prompt: prompt,
-      model: model
-    }.merge(options)
+    body = { prompt: prompt }.merge(options)
+    body[:model] = model unless %w[essential essential-v2].include?(model)
 
     request.body = body.to_json
     request

data/test/integration/getimg_client_integration_test.rb

@@ -3,9 +3,7 @@ require_relative "../test_helper"
 class GetimgClientIntegrationTest < Minitest::Test
   def setup
     @base_image = Base64.strict_encode64(File.read(File.join(__dir__, "sample_image.jpeg")))
-
     @small_image = Base64.strict_encode64(File.read(File.join(__dir__, "small_sample_image.jpeg")))
-
     @mask_image = Base64.strict_encode64(File.read(File.join(__dir__, "mask_image.jpeg")))
   end
 
@@ -58,6 +56,13 @@ class GetimgClientIntegrationTest < Minitest::Test
     save_image(result["image"], "test_upscale")
   end
 
+  def test_generate_text_to_image_essential
+    puts "-- Running test_generate_text_to_image_essential..."
+    result = GetimgClient.generate_image("A modern cityscape")
+    assert result["image"]
+    save_image(result["image"], "test_generate_text_to_image_essential")
+  end
+
   private
 
   def save_image(base64_image, filename)

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: getimg_client
 version: !ruby/object:Gem::Version
-  version: 0.1.2
+  version: 0.1.4
 platform: ruby
 authors:
 - Melvin Sommer
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2024-06-15 00:00:00.000000000 Z
+date: 2024-06-18 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: minitest
@@ -43,7 +43,8 @@ files:
 homepage: https://gitlab.com/coeusit/getimg_client
 licenses:
 - MIT
-metadata: {}
+metadata:
+  source_code_uri: https://gitlab.com/coeusit/getimg_client
 post_install_message: 
 rdoc_options: []
 require_paths: