novelai-python 0.4.13__tar.gz → 0.4.15__tar.gz

{novelai_python-0.4.13 → novelai_python-0.4.15}/NOTICE.MD

@@ -4,7 +4,7 @@
 
 The MIT license applies to the files in:
 
-    file: "src/novelai_python/sdk/ai/generate_image.py" from https://github.com/HanaokaYuzu/NovelAI-API
+    file: "src/novelai_python/sdk/ai/generate-image.py" from https://github.com/HanaokaYuzu/NovelAI-API
     file: "src/novelai_python/tokenizer/novelai.model" from https://github.com/NovelAI/novelai-tokenizer
     file: "src/novelai_python/tokenizer/novelai_v2.model from https://github.com/NovelAI/novelai-tokenizer
     file: "src/novelai_python/tool/image_metadata/lsb_injector.py" from https://github.com/NovelAI/novelai-image-metadata

{novelai_python-0.4.13 → novelai_python-0.4.15}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: novelai-python
-Version: 0.4.13
+Version: 0.4.15
 Summary: NovelAI Python Binding With Pydantic
 Author-Email: sudoskys <coldlando@hotmail.com>
 License: Apache License 2.0
@@ -27,9 +27,9 @@ Requires-Dist: pynacl>=1.5.0
 Requires-Dist: ftfy>=6.2.0
 Requires-Dist: regex>=2023.12.25
 Requires-Dist: tokenizers>=0.15.2
+Provides-Extra: testing
 Requires-Dist: pytest>=7.4.4; extra == "testing"
 Requires-Dist: pytest-asyncio>=0.23.4; extra == "testing"
-Provides-Extra: testing
 Description-Content-Type: text/markdown
 
 ![banner](https://github.com/LlmKira/novelai-python/blob/dev/playground/banner-raw.png?raw=true)

{novelai_python-0.4.13 → novelai_python-0.4.15}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "novelai-python"
-version = "0.4.13"
+version = "0.4.15"
 description = "NovelAI Python Binding With Pydantic"
 authors = [
     { name = "sudoskys", email = "coldlando@hotmail.com" },

{novelai_python-0.4.13 → novelai_python-0.4.15}/src/novelai_python/credential/ApiToken.py

@@ -4,6 +4,9 @@
 # @File    : ApiToken.py
 # @Software: PyCharm
 
+import datetime
+
+import shortuuid
 from curl_cffi.requests import AsyncSession
 from loguru import logger
 from pydantic import SecretStr, Field, field_validator
@@ -16,6 +19,7 @@ class ApiCredential(CredentialBase):
     ApiCredential is the base class for all credential.
     """
     api_token: SecretStr = Field(None, description="api token")
+    _x_correlation_id: str = shortuuid.uuid()[0:6]
 
     async def get_session(self, timeout: int = 180, update_headers: dict = None):
         headers = {
@@ -26,6 +30,8 @@ class ApiCredential(CredentialBase):
             "Content-Type": "application/json",
             "Origin": "https://novelai.net",
             "Referer": "https://novelai.net/",
+            "x-correlation-id": self._x_correlation_id,
+            "x-initiated-at": f"{datetime.datetime.now(datetime.UTC).isoformat()}Z",
         }
 
         if update_headers:

{novelai_python-0.4.13 → novelai_python-0.4.15}/src/novelai_python/credential/JwtToken.py

@@ -3,6 +3,9 @@
 # @Author  : sudoskys
 # @File    : JwtToken.py
 # @Software: PyCharm
+import datetime
+
+import shortuuid
 from curl_cffi.requests import AsyncSession
 from loguru import logger
 from pydantic import SecretStr, Field, field_validator
@@ -15,6 +18,7 @@ class JwtCredential(CredentialBase):
     JwtCredential is the base class for all credential.
     """
     jwt_token: SecretStr = Field(None, description="jwt token")
+    _x_correlation_id: str = shortuuid.uuid()[0:6]
 
     async def get_session(self, timeout: int = 180, update_headers: dict = None):
         headers = {
@@ -25,6 +29,8 @@ class JwtCredential(CredentialBase):
             "Content-Type": "application/json",
             "Origin": "https://novelai.net",
             "Referer": "https://novelai.net/",
+            "x-correlation-id": self._x_correlation_id,
+            "x-initiated-at": f"{datetime.datetime.now(datetime.UTC).isoformat()}Z",
         }
 
         if update_headers:

{novelai_python-0.4.13 → novelai_python-0.4.15}/src/novelai_python/credential/UserAuth.py

@@ -2,9 +2,11 @@
 # @Time    : 2024/2/7 下午12:14
 # @Author  : sudoskys
 # @File    : UserAuth.py
+import datetime
 import time
 from typing import Optional
 
+import shortuuid
 from curl_cffi.requests import AsyncSession
 from pydantic import SecretStr, Field
 
@@ -19,6 +21,7 @@ class LoginCredential(CredentialBase):
     password: SecretStr = Field(None, description="password")
     _session_headers: dict = {}
     _update_at: Optional[int] = None
+    _x_correlation_id: str = shortuuid.uuid()[0:6]
 
     async def get_session(self, timeout: int = 180, update_headers: dict = None):
         headers = {
@@ -29,6 +32,8 @@ class LoginCredential(CredentialBase):
             "Content-Type": "application/json",
             "Origin": "https://novelai.net",
             "Referer": "https://novelai.net/",
+            "x-correlation-id": self._x_correlation_id,
+            "x-initiated-at": f"{datetime.datetime.now(datetime.UTC).isoformat()}Z",
         }
 
         # 30 天有效期

{novelai_python-0.4.13 → novelai_python-0.4.15}/src/novelai_python/sdk/ai/_enum.py

@@ -4,25 +4,27 @@
 # @File    : _enum.py
 # @Software: PyCharm
 from enum import Enum, IntEnum
+from typing import List
+
+from pydantic.dataclasses import dataclass
 
 
 class Sampler(Enum):
+    PLMS = "plms"
+    DDIM = "ddim"
     K_EULER = "k_euler"
     K_EULER_ANCESTRAL = "k_euler_ancestral"
-    K_DPMPP_2S_ANCESTRAL = "k_dpmpp_2s_ancestral"
-    K_DPMPP_2M = "k_dpmpp_2m"
-    K_DPMPP_SDE = "k_dpmpp_sde"
-    DDIM_V3 = "ddim_v3"
-
-    DDIM = "ddim"
-    PLMS = "plms"
     K_DPM_2 = "k_dpm_2"
     K_DPM_2_ANCESTRAL = "k_dpm_2_ancestral"
     K_LMS = "k_lms"
+    K_DPMPP_2S_ANCESTRAL = "k_dpmpp_2s_ancestral"
+    K_DPMPP_SDE = "k_dpmpp_sde"
+    K_DPMPP_2M = "k_dpmpp_2m"
     K_DPM_ADAPTIVE = "k_dpm_adaptive"
     K_DPM_FAST = "k_dpm_fast"
     K_DPMPP_2M_SDE = "k_dpmpp_2m_sde"
     K_DPMPP_3M_SDE = "k_dpmpp_3m_sde"
+    DDIM_V3 = "ddim_v3"
     NAI_SMEA = "nai_smea"
     NAI_SMEA_DYN = "nai_smea_dyn"
 
@@ -34,6 +36,56 @@ class NoiseSchedule(Enum):
     POLYEXPONENTIAL = "polyexponential"
 
 
+def get_supported_noise_schedule(sample_type: Sampler) -> List[NoiseSchedule]:
+    """
+    Get supported noise schedule for a given sample type
+    :param sample_type: Sampler
+    :return: List[NoiseSchedule]
+    """
+    if sample_type in [
+        Sampler.K_EULER_ANCESTRAL,
+        Sampler.K_DPMPP_2S_ANCESTRAL,
+        Sampler.K_DPMPP_2M,
+        Sampler.K_DPMPP_2M_SDE,
+        Sampler.K_DPMPP_SDE,
+        Sampler.K_EULER
+    ]:
+        return [
+            NoiseSchedule.NATIVE,
+            NoiseSchedule.KARRAS,
+            NoiseSchedule.EXPONENTIAL,
+            NoiseSchedule.POLYEXPONENTIAL
+        ]
+    elif sample_type in [Sampler.K_DPM_2]:
+        return [
+            NoiseSchedule.EXPONENTIAL,
+            NoiseSchedule.POLYEXPONENTIAL
+        ]
+    else:
+        return []
+
+
+def get_default_noise_schedule(sample_type: Sampler) -> NoiseSchedule:
+    """
+    Get default noise schedule for a given sample type
+    :param sample_type: Sampler
+    :return: NoiseSchedule
+    """
+    if sample_type in [
+        Sampler.K_EULER_ANCESTRAL,
+        Sampler.K_DPMPP_2S_ANCESTRAL,
+        Sampler.K_DPMPP_2M,
+        Sampler.K_DPMPP_2M_SDE,
+        Sampler.K_DPMPP_SDE,
+        Sampler.K_EULER
+    ]:
+        return NoiseSchedule.KARRAS
+    elif sample_type in [Sampler.K_DPM_2]:
+        return NoiseSchedule.EXPONENTIAL
+    else:
+        return NoiseSchedule.NATIVE
+
+
 class UCPreset(IntEnum):
     TYPE0 = 0
     TYPE1 = 1
@@ -120,6 +172,8 @@ class ModelGroups(Enum):
 
 
 def get_model_group(model: str) -> ModelGroups:
+    if isinstance(model, Enum):
+        model = model.value
     mapping = {
         "stable-diffusion": ModelGroups.STABLE_DIFFUSION,
         "nai-diffusion": ModelGroups.STABLE_DIFFUSION,
@@ -162,8 +216,59 @@ PROMOTION = {
     "Stable Diffusion F1022D28": Model.NAI_DIFFUSION_2,
     "Stable Diffusion XL B0BDF6C1": Model.NAI_DIFFUSION_3,
     "Stable Diffusion XL C1E1DE52": Model.NAI_DIFFUSION_3,
+    "Stable Diffusion XL 7BCCAA2C": Model.NAI_DIFFUSION_3,
     "Stable Diffusion XL 8BA2AF87": Model.NAI_DIFFUSION_3,
     "Stable Diffusion XL 4BE8C60C": Model.NAI_DIFFUSION_FURRY_3,
     "Stable Diffusion XL C8704949": Model.NAI_DIFFUSION_FURRY_3,
+    "Stable Diffusion XL 37C2B166": Model.NAI_DIFFUSION_FURRY_3,
     "Stable Diffusion XL 9CC2F394": Model.NAI_DIFFUSION_FURRY_3,
 }
+
+
+def get_default_uc_preset(model: str, uc_preset: int) -> str:
+    if isinstance(model, Enum):
+        model = model.value
+    if isinstance(uc_preset, Enum):
+        uc_preset = uc_preset.value
+
+    @dataclass
+    class UcPrompt:
+        label: str
+        text: str
+
+    mapper = {
+        ModelGroups.STABLE_DIFFUSION: [
+            UcPrompt("lowQualityPlusBadAnatomy",
+                     text="lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry"),
+            UcPrompt("lowQuality",
+                     text="lowres, text, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry"),
+            UcPrompt("none", text="lowres")
+        ],
+        ModelGroups.STABLE_DIFFUSION_GROUP_2: [
+            UcPrompt("heavy",
+                     text="lowres, bad, text, error, missing, extra, fewer, cropped, jpeg artifacts, worst quality, bad quality, watermark, displeasing, unfinished, chromatic aberration, scan, scan artifacts"),
+            UcPrompt("light", text="lowres, jpeg artifacts, worst quality, watermark, blurry, very displeasing"),
+            UcPrompt("none", text="lowres")
+        ],
+        ModelGroups.STABLE_DIFFUSION_XL: [
+            UcPrompt("heavy",
+                     text="lowres, {bad}, error, fewer, extra, missing, worst quality, jpeg artifacts, bad quality, watermark, unfinished, displeasing, chromatic aberration, signature, extra digits, artistic error, username, scan, [abstract]"),
+            UcPrompt("light", text="lowres, jpeg artifacts, worst quality, watermark, blurry, very displeasing"),
+            UcPrompt("humanFocus",
+                     text="lowres, {bad}, error, fewer, extra, missing, worst quality, jpeg artifacts, bad quality, watermark, unfinished, displeasing, chromatic aberration, signature, extra digits, artistic error, username, scan, [abstract], bad anatomy, bad hands, @_@, mismatched pupils, heart-shaped pupils, glowing eyes"),
+            UcPrompt("none", text="lowres")
+        ],
+        ModelGroups.STABLE_DIFFUSION_XL_FURRY: [
+            UcPrompt("heavy",
+                     text="{{worst quality}}, [displeasing], {unusual pupils}, guide lines, {{unfinished}}, {bad}, url, artist name, {{tall image}}, mosaic, {sketch page}, comic panel, impact (font), [dated], {logo}, ych, {what}, {where is your god now}, {distorted text}, repeated text, {floating head}, {1994}, {widescreen}, absolutely everyone, sequence, {compression artifacts}, hard translated, {cropped}, {commissioner name}, unknown text, high contrast"),
+            UcPrompt("light",
+                     text="{worst quality}, guide lines, unfinished, bad, url, tall image, widescreen, compression artifacts, unknown text"),
+            UcPrompt("none", text="lowres")
+        ],
+    }
+    model_group = get_model_group(model)
+    prompts: List[UcPrompt] = mapper.get(model_group, [UcPrompt("none", "lowres")])
+    if 0 <= uc_preset < len(prompts):
+        return prompts[uc_preset].text
+    else:
+        return "lowres"

{novelai_python-0.4.13 → novelai_python-0.4.15}/src/novelai_python/sdk/ai/generate_image/__init__.py

@@ -23,8 +23,9 @@ from pydantic import BaseModel, ConfigDict, PrivateAttr, field_validator, model_
 from tenacity import retry, stop_after_attempt, wait_random, retry_if_exception
 from typing_extensions import override
 
-from novelai_python.sdk.ai._enum import Model, Sampler, NoiseSchedule, ControlNetModel, Action, UCPreset, INPAINTING_MODEL_LIST
 from novelai_python.sdk.ai._cost import CostCalculator
+from novelai_python.sdk.ai._enum import Model, Sampler, NoiseSchedule, ControlNetModel, Action, UCPreset, \
+    INPAINTING_MODEL_LIST, get_default_noise_schedule, get_supported_noise_schedule, get_default_uc_preset
 from ...schema import ApiBaseModel
 from ...._exceptions import APIError, AuthError, ConcurrentGenerationError, SessionHttpError
 from ...._response.ai.generate_image import ImageGenerateResp, RequestParams
@@ -52,7 +53,7 @@ class Params(BaseModel):
     controlnet_strength: Optional[float] = Field(1.0, ge=0.1, le=2, multiple_of=0.1)
     """ControlNet Strength"""
     dynamic_thresholding: Optional[bool] = False
-    """Reduce artifacts caused by high prompt guidance values"""
+    """Decrisp:Reduce artifacts caused by high prompt guidance values"""
     height: Optional[int] = Field(1216, ge=64, le=49152)
     """Height For Generate Image"""
     image: Optional[Union[str, bytes]] = None
@@ -69,7 +70,7 @@ class Params(BaseModel):
     """Number of samples"""
     negative_prompt: Optional[str] = ''
     """Negative Prompt"""
-    noise_schedule: Optional[NoiseSchedule] = NoiseSchedule.NATIVE
+    noise_schedule: Optional[NoiseSchedule] = None
     """Noise Schedule"""
     # Misc
     params_version: Optional[int] = 1
@@ -112,7 +113,7 @@ class Params(BaseModel):
     # TODO: find out the usage
     qualityToggle: Optional[bool] = True
     """Whether to add the quality prompt"""
-    sampler: Optional[Sampler] = Sampler.K_EULER
+    sampler: Optional[Sampler] = Sampler.K_EULER_ANCESTRAL
     """Sampler For Generate Image"""
     scale: Optional[float] = Field(6.0, ge=0, le=10, multiple_of=0.1)
     """Prompt Guidance"""
@@ -123,6 +124,8 @@ class Params(BaseModel):
         le=4294967295 - 7,
     )
     """Seed"""
+    skip_cfg_above_sigma: Optional[int] = None
+    """Variety Boost, a new feature to improve the diversity of samples."""
     extra_noise_seed: Optional[int] = Field(
         default_factory=lambda: random.randint(0, 4294967295 - 7),
         gt=0,
@@ -133,7 +136,7 @@ class Params(BaseModel):
     # TODO: find out the usage
     sm_dyn: Optional[bool] = False
     # TODO: find out the usage
-    steps: Optional[int] = Field(28, ge=1, le=50)
+    steps: Optional[int] = Field(23, ge=1, le=50)
     """Steps"""
     ucPreset: Optional[UCPreset] = 0
     """The Negative Prompt Preset"""
@@ -212,6 +215,16 @@ class Params(BaseModel):
     # Validators
     @model_validator(mode="after")
     def image_validator(self):
+        # Noise schedule check
+        if self.noise_schedule is None:
+            self.noise_schedule = get_default_noise_schedule(self.sampler)
+        supported_noise_schedule = get_supported_noise_schedule(self.sampler)
+        if supported_noise_schedule:
+            if self.noise_schedule not in supported_noise_schedule:
+                raise ValueError(f"Invalid noise_schedule, must be one of {supported_noise_schedule}")
+        else:
+            logger.warning(f"Inactive sampler {self.sampler} does not support noise_schedule.")
+        # Check the noise value
         if self.sampler:
             if self.sampler in [Sampler.DDIM, Sampler.DDIM_V3]:
                 self.sm = False
@@ -223,6 +236,7 @@ class Params(BaseModel):
                 self.sm_dyn = True
                 if not self.sm_dyn:
                     logger.warning("sm and sm_dyn is enabled when using nai_smea_dyn sampler.")
+        # Make sure the image is base64 encoded
         if isinstance(self.image, str) and self.image.startswith("data:"):
             raise ValueError("Invalid `image` format, must be base64 encoded directly.")
         if isinstance(self.mask, str) and self.mask.startswith("data:"):
@@ -310,7 +324,7 @@ class GenerateImageInfer(ApiBaseModel):
 
     action: Union[str, Action] = Field(Action.GENERATE, description="Mode for img generate")
     input: str = "1girl, best quality, amazing quality, very aesthetic, absurdres"
-    model: Optional[Model] = "nai-diffusion-3"
+    model: Optional[Union[Model, str]] = "nai-diffusion-3"
     parameters: Params = Params()
     model_config = ConfigDict(extra="ignore")
 
@@ -324,30 +338,10 @@ class GenerateImageInfer(ApiBaseModel):
         """
         if self.parameters.negative_prompt is None:
             self.parameters.negative_prompt = ""
-        if self.parameters.ucPreset == 0:
-            # 0: 重型
-            self.parameters.negative_prompt = (
-                "lowres, {bad}, error, fewer, extra, missing, worst quality, jpeg artifacts, bad quality, "
-                "watermark, unfinished, displeasing, chromatic aberration, signature, extra digits, artistic error, "
-                "username, scan, [abstract], bad anatomy, bad hands, @_@, mismatched pupils, heart-shaped pupils, "
-                "glowing eyes, lowres, bad anatomy, bad hands, text, error, missing fingers, "
-                "extra digit, fewer digits, cropped, worst quality, low quality, normal quality, "
-                f"jpeg artifacts, signature, watermark, username, blurry "
-                f",{self.parameters.negative_prompt}"
-            )
-        elif self.parameters.ucPreset == 1:
-            # 1: 轻型
-            self.parameters.negative_prompt = (f"lowres, jpeg artifacts, worst quality, watermark, blurry, "
-                                               f"very displeasing"
-                                               f",{self.parameters.negative_prompt}")
-        elif self.parameters.ucPreset == 2:
-            # 2: 人物
-            self.parameters.negative_prompt = ("lowres, {bad}, error, fewer, extra, missing, worst quality, "
-                                               "jpeg artifacts, bad quality, watermark, unfinished, displeasing, "
-                                               "chromatic aberration, signature, extra digits, artistic error, "
-                                               "username, scan, [abstract], bad anatomy, bad hands, @_@, mismatched "
-                                               f"pupils, heart-shaped pupils, glowing eyes "
-                                               f",{self.parameters.negative_prompt}")
+        default_negative_prompt = get_default_uc_preset(self.model, self.parameters.ucPreset)
+        self.parameters.negative_prompt = ", ".join(
+            filter(None, [default_negative_prompt, self.parameters.negative_prompt])
+        )
         if self.parameters.qualityToggle:
             self.input += ", best quality, amazing quality, very aesthetic, absurdres"
 
@@ -430,7 +424,10 @@ class GenerateImageInfer(ApiBaseModel):
               height: int = 1216,
               qualityToggle: bool = None,
               image: Union[str, bytes] = None,
+              decrisp_mode: bool = False,
+              variety_boost: bool = False,
               noise: float = 0,
+              noise_schedule: Union[NoiseSchedule, str] = None,
               reference_image_multiple: List[Union[str, bytes]] = None,
               reference_strength_multiple: List[float] = None,
               reference_information_extracted_multiple: List[float] = None,
@@ -450,6 +447,11 @@ class GenerateImageInfer(ApiBaseModel):
         """
         The build function, more convenient to create a GenerateImageInfer instance.
         构建函数，更方便地创建一个GenerateImageInfer实例。
+        :param decrisp_mode: Reduce artifacts caused by high prompt guidance values.
+                            减少由高提示引导值引起的伪影。
+        :param variety_boost: A new feature to improve the diversity of samples.
+                            Variety Boost means your negative prompt will only be used after the body shape has been decided.
+                            一项新功能，用于提高样本的多样性。但是，这意味着您的负面提示只会在身体形状确定后使用。
         :param reference_information_extracted_multiple:  The reference information extracted for Vibe mode.
                                                             Vibe模式的提取的参考信息。
         :param reference_image_multiple: The reference image for Vibe mode.
@@ -459,52 +461,53 @@ class GenerateImageInfer(ApiBaseModel):
         :param ucPreset: 0: Heavy, 1: Light, 2: Character
                         0: 重量级, 1: 轻量级, 2: 角色
         :param qualityToggle: add the quality prompt or not.
-                              是否添加质量提示。
+                            是否添加质量提示。
         :param prompt: input prompt for generation.
-                       生成的输入提示。
+                    生成的输入提示。
         :param model: select a model.
-                      选择一个模型。
+                    选择一个模型。
         :param action: mode for image generation [generate, img2img, infill].
-                       图像生成模式 [生成, 图像到图像, 填充]。
+                    图像生成模式 [生成, 图像到图像, 填充]。
         :param negative_prompt: the content of negative prompt.
                                 负面提示的内容。
         :param seed: the seed for generate image.
-                     生成图像的种子。
+                    生成图像的种子。
         :param steps: the steps for generate image.
-                      生成图像的步骤。
+                    生成图像的步骤。
         :param scale: the scale for generate image.
-                      生成图像的比例。
+                    生成图像的比例。
         :param cfg_rescale: prompt guidance rescale 0-1 lower is more creative.
                             提示引导的重新缩放0-1，越低则越富有创造性。
         :param sampler: the sampler for generate image.
                         生成图像的采样器。
         :param width: the width of the image.
-                      图像的宽度。
+                    图像的宽度。
         :param height: the height of the image.
-                       图像的高度。
+                    图像的高度。
         :param image: the input image.
-                      输入图像。
+                    输入图像。
         :param noise: the noise to be added in the image.
-                      要添加到图像中的噪声。
+                    要添加到图像中的噪声。
+        :param noise_schedule: the noise schedule for generate image.
+                    生成图像的噪声计划。
         :param add_original_image: Overlay Original Image. Prevents the existing image from changing, only for IMG2IMG mode.
-                                   叠加原始图像。防止现有图像发生更改，仅适用于IMG2IMG模式。
+                                叠加原始图像。防止现有图像发生更改，仅适用于IMG2IMG模式。
         :param strength: the strength of IMG2IMG mode.
-                         IMG2IMG模式的强度。
+                        IMG2IMG模式的强度。
         :param mask: the inpainting mask.
-                     纹理绘制掩码。
+                    纹理绘制掩码。
         :param controlnet_model: the model used for the control network.
-                                 用于控制网络的模型。
+                                用于控制网络的模型。
         :param controlnet_condition: the condition used for the control network.
-                                     用于控制网络的条件。
+                                    用于控制网络的条件。
         :param sm: whether to use sm.
-                    是否使用sm。
+                是否使用sm。
         :param sm_dyn: whether to use dynamic sm.
-                       是否使用动态sm。
+                    是否使用动态sm。
         :param uncond_scale: the strength of the unrelated content.
-                             无关内容的强度。
+                            无关内容的强度。
         :param kwargs: any additional parameters.
-                       其他任何参数。
-
+                    其他任何参数。
         :return: self
         返回：自身
         """
@@ -528,6 +531,7 @@ class GenerateImageInfer(ApiBaseModel):
             "sm_dyn": sm_dyn,
             "sm": sm,
             "uncond_scale": uncond_scale,
+            "noise_schedule": noise_schedule,
         })
 
         def _merge_param(v1, v2):
@@ -579,6 +583,10 @@ class GenerateImageInfer(ApiBaseModel):
                 "noise": noise,
             }
         )
+        if decrisp_mode:
+            kwargs.update({"dynamic_thresholding": True})
+        if variety_boost:
+            kwargs.update({"skip_cfg_above_sigma": 19})
         # 清理空值
         param = {k: v for k, v in kwargs.items() if v is not None}
         _build_prop = Params(**param)

{novelai_python-0.4.13 → novelai_python-0.4.15}/src/novelai_python/server.py

@@ -15,6 +15,7 @@ from loguru import logger
 from starlette.responses import JSONResponse, StreamingResponse
 
 from novelai_python import LLMStream, LLMStreamResp
+from novelai_python.sdk.ai.augment_image import ReqType, Moods, AugmentImageInfer
 from .credential import JwtCredential, SecretStr
 from .sdk.ai.generate import LLM
 from .sdk.ai.generate_image import GenerateImageInfer, Model
@@ -123,7 +124,7 @@ async def upscale(
         return JSONResponse(status_code=500, content=e.__dict__)
 
 
-@app.get("/ai/generate_image/suggest_tags")
+@app.get("/ai/generate-image/suggest_tags")
 async def suggest_tags(
         model: Model,
         prompt: str,
@@ -145,7 +146,7 @@ async def suggest_tags(
         return JSONResponse(status_code=500, content=e.__dict__)
 
 
-@app.post("/ai/generate_image")
+@app.post("/ai/generate-image")
 async def generate_image(
         req: GenerateImageInfer,
         current_token: str = Depends(get_current_token)
@@ -244,6 +245,48 @@ async def generate_voice(
         return JSONResponse(status_code=500, content=e.__dict__)
 
 
+@app.post("/ai/augment-image")
+async def augment_image(
+        req_type: ReqType,
+        image: bytes,
+        mood: Optional[Moods] = None,
+        prompt: Optional[str] = None,
+        defry: Optional[int] = None,
+        current_token: str = Depends(get_current_token)
+):
+    """
+    生成图片
+    :param current_token: Authorization
+    :param req_type: ReqType
+    :param image: bytes
+    :param mood: Optional[Moods]
+    :param prompt: Optional[str]
+    :param defry: Optional[int]
+    :return:
+    """
+    try:
+        req = AugmentImageInfer.build(
+            req_type=req_type,
+            image=image,
+            mood=mood,
+            prompt=prompt,
+            defry=defry,
+        )
+        _result = await req.request(session=get_session(current_token))
+        zip_file_bytes = io.BytesIO()
+        with zipfile.ZipFile(zip_file_bytes, mode="w", compression=zipfile.ZIP_DEFLATED) as zip_file:
+            for file in _result.files:
+                zip_file.writestr(zinfo_or_arcname=file[0], data=file[1])
+        # return the zip file
+        zip_file_bytes.seek(0)
+        return StreamingResponse(zip_file_bytes, media_type='application/zip', headers={
+            'Content-Disposition': 'attachment;filename=image.zip'
+        })
+    except Exception as e:
+        logger.exception(e)
+        return JSONResponse(status_code=500, content=e.__dict__)
+
+
 # 获取输入参数
 def usage():
     print("Usage: python -m novelai_python.server -h <host> -p <port>")

{novelai_python-0.4.13 → novelai_python-0.4.15}/tests/test_server_run.py

@@ -29,7 +29,7 @@ def test_subscription_without_api_key(mock_subscription):
 @patch('src.novelai_python.server.GenerateImageInfer')
 def test_generate_image_without_api_key(mock_generate_image):
     mock_generate_image.return_value.request.return_value = {"status": "image generated"}
-    response = client.post("/ai/generate_image")
+    response = client.post("/ai/generate-image")
     assert response.status_code == 403