Feat: add gpustack image tools

api/core/tools/provider/_position.yaml

@@ -26,6 +26,7 @@
 - cogview
 - comfyui
 - getimgai
+- gpustack
 - siliconflow
 - spark
 - stepfun

api/core/tools/provider/builtin/gpustack/gpustack.py

@@ -0,0 +1,26 @@
+import requests
+
+from core.tools.errors import ToolProviderCredentialValidationError
+from core.tools.provider.builtin_tool_provider import BuiltinToolProviderController
+
+
+class GPUStackProvider(BuiltinToolProviderController):
+    def _validate_credentials(self, credentials: dict) -> None:
+        base_url = credentials.get("base_url", "").removesuffix("/").removesuffix("/v1-openai")
+        api_key = credentials.get("api_key", "")
+        tls_verify = credentials.get("tls_verify", True)
+
+        if not base_url:
+            raise ToolProviderCredentialValidationError("GPUStack base_url is required")
+        if not api_key:
+            raise ToolProviderCredentialValidationError("GPUStack api_key is required")
+        headers = {
+            "accept": "application/json",
+            "authorization": f"Bearer {api_key}",
+        }
+
+        response = requests.get(f"{base_url}/v1-openai/models", headers=headers, verify=tls_verify)
+        if response.status_code != 200:
+            raise ToolProviderCredentialValidationError(
+                f"Failed to validate GPUStack API key, status code: {response.status_code}-{response.text}"
+            )

api/core/tools/provider/builtin/gpustack/gpustack.yaml

@@ -0,0 +1,44 @@
+identity:
+  author: gpustack
+  name: gpustack
+  label:
+    en_US: GPUStack
+    zh_Hans: GPUStack
+  description:
+    en_US: GPUStack is an open-source GPU cluster manager for running AI models, providing efficient resource management and model deployment capabilities.
+    zh_Hans: GPUStack 是一款开源的 GPU 集群管理工具，专为 AI 模型部署和运行而设计，提供高效的资源管理和模型部署能力。
+  icon: icon.svg
+  tags:
+    - image
+credentials_for_provider:
+  base_url:
+    type: text-input
+    required: true
+    label:
+      en_US: Server URL
+      zh_Hans: 服务器 URL
+    placeholder:
+      en_US: http://your-server-address.com
+    help:
+      en_US: Please input GPUStack server's URL
+      zh_Hans: 请输入 GPUStack 服务器的 URL
+  api_key:
+    type: secret-input
+    required: true
+    label:
+      en_US: API Key
+      zh_Hans: API Key
+    placeholder:
+      en_US: Please input your GPUStack API Key
+      zh_Hans: 请输入你的 GPUStack API Key
+    url: https://docs.gpustack.ai/latest/user-guide/api-key-management/
+  tls_verify:
+    type: boolean
+    required: false
+    label:
+      en_US: TLS Verify
+      zh_Hans: 证书验证
+    help:
+      en_US: Whether to verify the TLS certificate of the GPUStack server.
+      zh_Hans: 是否验证 GPUStack 服务器的 TLS 证书。
+    default: true

api/core/tools/provider/builtin/gpustack/tools/image_edit.py

@@ -0,0 +1,47 @@
+import io
+from typing import Any, Union
+
+import requests
+
+from core.file.enums import FileType
+from core.file.file_manager import download
+from core.tools.entities.tool_entities import ToolInvokeMessage
+from core.tools.tool.builtin_tool import BuiltinTool
+
+from .utils import get_base_url, get_common_params, handle_api_error, handle_image_response
+
+
+class ImageEditTool(BuiltinTool):
+    def _invoke(
+        self, user_id: str, tool_parameters: dict[str, Any]
+    ) -> Union[ToolInvokeMessage, list[ToolInvokeMessage]]:
+        image = tool_parameters.get("image")
+        if image.type != FileType.IMAGE:
+            return [self.create_text_message("Not a valid image file")]
+
+        try:
+            params = get_common_params(tool_parameters)
+            params["strength"] = tool_parameters.get("strength", 0.75)
+
+            image_binary = io.BytesIO(download(image))
+            files = {"image": ("image.png", image_binary, "image/png")}
+
+            base_url = get_base_url(self.runtime.credentials["base_url"])
+            response = requests.post(
+                f"{base_url}/v1-openai/images/edits",
+                headers={"Authorization": f"Bearer {self.runtime.credentials['api_key']}"},
+                data=params,
+                files=files,
+                verify=self.runtime.credentials.get("tls_verify", True),
+            )
+
+            if not response.ok:
+                return self.create_text_message(handle_api_error(response))
+
+            result = []
+            return handle_image_response(result, response, self)
+
+        except ValueError as e:
+            return self.create_text_message(str(e))
+        except Exception as e:
+            return self.create_text_message(f"An error occurred: {str(e)}")

api/core/tools/provider/builtin/gpustack/tools/image_edit.yaml

@@ -0,0 +1,181 @@
+identity:
+  name: image_edit
+  author: gpustack
+  label:
+    en_US: Image Edit
+    zh_Hans: 图片编辑
+  icon: icon.svg
+description:
+  human:
+    en_US: Edit images with GPUStack's image editing model.
+    zh_Hans: 使用 GPUStack 的图像编辑模型编辑图片。
+  llm: This tool is used to edit image.
+parameters:
+  - name: image
+    type: file
+    required: true
+    label:
+      en_US: Image
+      zh_Hans: 图片
+    human_description:
+      en_US: The image to be edited.
+      zh_Hans: 要编辑的图片。
+    llm_description: The image to be edited.
+    form: llm
+  - name: prompt
+    type: string
+    required: true
+    label:
+      en_US: prompt
+      zh_Hans: 提示词
+    human_description:
+      en_US: The text prompt used to edit the image.
+      zh_Hans: 用于编辑图片的文字提示词
+    llm_description: this prompt text will be used to edit image.
+    form: llm
+  - name: model
+    type: string
+    required: true
+    label:
+      en_US: Model
+      zh_Hans: 模型
+    human_description:
+      en_US: image model name that running in GPUStack.
+      zh_Hans: 在 GPUStack 上运行的图像模型名称。
+    form: form
+  - name: cfg_scale
+    type: number
+    required: false
+    default: 4.5
+    label:
+      en_US: CFG Scale
+    human_description:
+      en_US: Classifier-free guidance scale, affecting the image's adherence to the prompt.
+      zh_Hans: 无分类器引导比例，影响图片的对 Prompt 的贴合度。
+    form: form
+  - name: n
+    type: number
+    required: false
+    default: 1
+    label:
+      en_US: Number
+      zh_Hans: 数量
+    human_description:
+      en_US: Number of images to generate.
+      zh_Hans: 生成图片数量。
+    form: form
+  - name: size
+    type: string
+    required: true
+    default: "512x512"
+    label:
+      en_US: Image Size
+      zh_Hans: 图片尺寸
+    human_description:
+      en_US: The maximum size of the generated image is controlled by the deployment parameters of the model.
+      zh_Hans: 图片生成的最大尺寸受控于模型的部署参数。
+    form: form
+  - name: sample_method
+    type: select
+    required: true
+    default: euler
+    options:
+      - value: euler_a
+        label:
+          en_US: euler_a
+      - value: euler
+        label:
+          en_US: euler
+      - value: heun
+        label:
+          en_US: heun
+      - value: dpm2
+        label:
+          en_US: dpm2
+      - value: dpm++2s_a
+        label:
+          en_US: dpm++2s_a
+      - value: dpm++2m
+        label:
+          en_US: dpm++2m
+      - value: dpm++2mv2
+        label:
+          en_US: dpm++2mv2
+      - value: ipndm
+        label:
+          en_US: ipndm
+      - value: ipndm_v
+        label:
+          en_US: ipndm_v
+      - value: icm
+        label:
+          en_US: icm
+    label:
+      en_US: Sample Method
+      zh_Hans: 采样方法
+    human_description:
+      en_US: The sample method for the image generation model.
+      zh_Hans: 图像生成模型的采样方法。
+    form: form
+  - name: sampling_steps
+    type: number
+    required: false
+    default: 20
+    label:
+      en_US: Sampling Steps
+      zh_Hans: 采样步数
+    human_description:
+      en_US: Number of sampling steps to generate the image.
+      zh_Hans: 生成图片所需的采样步数。
+    form: form
+  - name: guidance
+    type: number
+    required: false
+    default: 4.5
+    label:
+      en_US: Guidance
+    human_description:
+      en_US: Guidance scale, affecting the quality and diversity of the image.
+      zh_Hans: 引导比例，影响图片的质量和多样性
+    form: form
+  - name: schedule_method
+    type: select
+    required: true
+    default: discrete
+    options:
+      - value: discrete
+        label:
+          en_US: discrete
+      - value: karras
+        label:
+          en_US: karras
+      - value: exponential
+        label:
+          en_US: exponential
+      - value: ays
+        label:
+          en_US: ays
+      - value: gits
+        label:
+          en_US: gits
+    label:
+      en_US: Schedule Method
+      zh_Hans: 调度方法
+    form: form
+  - name: strength
+    type: number
+    required: false
+    default: 0.75
+    label:
+      en_US: Strength
+      zh_Hans: 强度
+    human_description:
+      en_US: The higher the value, the greater the modification to the original image.
+      zh_Hans: 值越高，它对原图的修改越大。
+    form: form
+  - name: seed
+    type: number
+    required: false
+    label:
+      en_US: Seed
+    form: form

api/core/tools/provider/builtin/gpustack/tools/text2image.py

@@ -0,0 +1,34 @@
+from typing import Any, Union
+
+import requests
+
+from core.tools.entities.tool_entities import ToolInvokeMessage
+from core.tools.tool.builtin_tool import BuiltinTool
+
+from .utils import get_base_url, get_common_params, handle_api_error, handle_image_response
+
+
+class TextToImageTool(BuiltinTool):
+    def _invoke(
+        self, user_id: str, tool_parameters: dict[str, Any]
+    ) -> Union[ToolInvokeMessage, list[ToolInvokeMessage]]:
+        try:
+            params = get_common_params(tool_parameters)
+            base_url = get_base_url(self.runtime.credentials["base_url"])
+            response = requests.post(
+                f"{base_url}/v1-openai/images/generations",
+                headers={"Authorization": f"Bearer {self.runtime.credentials['api_key']}"},
+                json=params,
+                verify=self.runtime.credentials.get("tls_verify", True),
+            )
+
+            if not response.ok:
+                return self.create_text_message(handle_api_error(response))
+
+            result = []
+            return handle_image_response(result, response, self)
+
+        except ValueError as e:
+            return self.create_text_message(str(e))
+        except Exception as e:
+            return self.create_text_message(f"An error occurred: {str(e)}")