gpt

LunaTranslator/LunaTranslator/ocrengines/chatgptlike.py

@@ -0,0 +1,118 @@
+from ocrengines.baseocrclass import baseocr
+import base64
+
+
+class OCR(baseocr):
+
+    def langmap(self):
+        return {
+            "zh": "Simplified Chinese",
+            "ja": "Japanese",
+            "en": "English",
+            "ru": "Russian",
+            "es": "Spanish",
+            "ko": "Korean",
+            "fr": "French",
+            "cht": "Traditional Chinese",
+            "vi": "Vietnamese",
+            "tr": "Turkish",
+            "pl": "Polish",
+            "uk": "Ukrainian",
+            "it": "Italian",
+            "ar": "Arabic",
+            "th": "Thai",
+        }
+
+    def createdata(self, message):
+        try:
+            temperature = float(self.config["Temperature"])
+        except:
+            temperature = 0.3
+
+        data = dict(
+            model=self.config["model"],
+            messages=message,
+            # optional
+            max_tokens=self.config["max_tokens"],
+            n=1,
+            # stop=None,
+            top_p=self.config["top_p"],
+            temperature=temperature,
+            frequency_penalty=self.config["frequency_penalty"],
+        )
+        return data
+
+    def createparam(self):
+        return None
+
+    def createheaders(self):
+        return {"Authorization": "Bearer " + self.config["SECRET_KEY"]}
+
+    def checkv1(self, api_url: str):
+        # 傻逼豆包大模型是非要v3，不是v1
+        if api_url.endswith("/v3"):
+            return api_url
+        elif api_url.endswith("/v3/"):
+            return api_url[:-1]
+        # 智谱AI
+        elif api_url.endswith("/v4"):
+            return api_url
+        elif api_url.endswith("/v4/"):
+            return api_url[:-1]
+        # 正常的
+        elif api_url.endswith("/v1"):
+            return api_url
+        elif api_url.endswith("/v1/"):
+            return api_url[:-1]
+        elif api_url.endswith("/"):
+            return api_url + "v1"
+        else:
+            return api_url + "/v1"
+
+    def ocr(self, imagebinary):
+
+        if self.config["use_custom_prompt"]:
+            prompt = self.config["custom_prompt"]
+        else:
+            prompt = f"Recognize the {self.srclang} text in the picture."
+
+        base64_image = base64.b64encode(imagebinary).decode("utf-8")
+        message = [
+            {
+                "role": "user",
+                "content": [
+                    {"type": "text", "text": prompt},
+                    {
+                        "type": "image_url",
+                        "image_url": {
+                            "url": f"data:image/jpeg;base64,{base64_image}",
+                            "detail": "low",
+                        },
+                    },
+                ],
+            }
+        ]
+
+        response = self.proxysession.post(
+            self.createurl(),
+            headers=self.createheaders(),
+            params=self.createparam(),
+            json=self.createdata(message),
+        )
+        try:
+            message = (
+                response.json()["choices"][0]["message"]["content"]
+                .replace("\n\n", "\n")
+                .strip()
+            )
+            return message
+        except:
+            raise Exception(response.text)
+
+    def createurl(self):
+        url = self.config["apiurl"]
+        if url.endswith("/chat/completions"):
+            pass
+        else:
+            url = self.checkv1(url) + "/chat/completions"
+        return url

LunaTranslator/LunaTranslator/ocrengines/geminiocr.py

@@ -4,6 +4,25 @@ from ocrengines.baseocrclass import baseocr
 
 
 class OCR(baseocr):
+    def langmap(self):
+        return {
+            "zh": "Simplified Chinese",
+            "ja": "Japanese",
+            "en": "English",
+            "ru": "Russian",
+            "es": "Spanish",
+            "ko": "Korean",
+            "fr": "French",
+            "cht": "Traditional Chinese",
+            "vi": "Vietnamese",
+            "tr": "Turkish",
+            "pl": "Polish",
+            "uk": "Ukrainian",
+            "it": "Italian",
+            "ar": "Arabic",
+            "th": "Thai",
+        }
+
     def ocr(self, imagebinary):
         self.checkempty(["key"])
         self.checkempty(["url"])
@@ -13,12 +32,15 @@ class OCR(baseocr):
         model = self.config["model"]
         image_data = base64.b64encode(imagebinary).decode("utf-8")
 
-        # Prepare the request payload
+        if self.config["use_custom_prompt"]:
+            prompt = self.config["custom_prompt"]
+        else:
+            prompt = f"Recognize the {self.srclang} text in the picture."
         payload = {
             "contents": [
                 {
                     "parts": [
-                        {"text": "Ocr this picture"},
+                        {"text": prompt},
                         {"inlineData": {"mimeType": "image/png", "data": image_data}},
                     ]
                 }

LunaTranslator/files/defaultconfig/config.json

@@ -1471,6 +1471,10 @@
         "geminiocr": {
             "use": false,
             "name": "GeminiOCR"
+        },
+        "chatgptlike": {
+            "use": false,
+            "name": "ChatGPT_兼容接口_OCR"
         }
     },
     "fanyi": {

LunaTranslator/files/defaultconfig/ocrsetting.json

@@ -272,7 +272,83 @@
         "args": {
             "key": "",
             "url": "https://generativelanguage.googleapis.com/v1",
-            "model": "gemini-1.5-flash"
+            "model": "gemini-1.5-flash",
+            "use_custom_prompt": false,
+            "custom_prompt": ""
+        },
+        "argstype": {
+            "custom_prompt": {
+                "name": "自定义promt",
+                "type": "multiline",
+                "refswitch": "use_custom_prompt"
+            },
+            "use_custom_prompt": {
+                "type": "switch_ref",
+                "name": "使用自定义promt"
+            }
+        }
+    },
+    "chatgptlike": {
+        "args": {
+            "model": "gpt-4",
+            "apiurl": "https://api.openai.com/v1",
+            "SECRET_KEY": "",
+            "use_custom_prompt": false,
+            "custom_prompt": "",
+            "Temperature": 0.3,
+            "top_p": 0.3,
+            "max_tokens": 128,
+            "frequency_penalty": 0,
+            "s": ""
+        },
+        "argstype": {
+            "SECRET_KEY": {
+                "rank": 1.1
+            },
+            "s": {
+                "type": "split",
+                "rank": 1.5
+            },
+            "apiurl": {
+                "name": "API接口地址",
+                "rank": 0
+            },
+            "model": {
+                "rank": 1
+            },
+            "top_p": {
+                "type": "spin",
+                "min": 0,
+                "max": 1,
+                "step": 0.01
+            },
+            "frequency_penalty": {
+                "type": "spin",
+                "min": 0,
+                "max": 2,
+                "step": 0.05
+            },
+            "max_tokens": {
+                "type": "intspin",
+                "min": 1,
+                "max": 4096,
+                "step": 1
+            },
+            "custom_prompt": {
+                "name": "自定义promt",
+                "type": "multiline",
+                "refswitch": "use_custom_prompt"
+            },
+            "use_custom_prompt": {
+                "type": "switch_ref",
+                "name": "使用自定义promt"
+            },
+            "Temperature": {
+                "type": "spin",
+                "min": 0,
+                "max": 1,
+                "step": 0.1
+            }
         }
     },
     "xunfei": {
@@ -319,4 +395,4 @@
             }
         }
     }
-}
+}

LunaTranslator/files/defaultconfig/translatorsetting.json

@@ -363,7 +363,7 @@
 				"rank": 1
 			},
 			"API_KEY": {
-				"rank": 2
+				"rank": 0.5
 			},
 			"流式输出": {
 				"type": "switch",

plugins/CMakeLists.txt

@@ -29,7 +29,7 @@ include(generate_product_version)
 
 set(VERSION_MAJOR 5)
 set(VERSION_MINOR 15)
-set(VERSION_PATCH 10)
+set(VERSION_PATCH 11)
 
 add_library(pch pch.cpp)
 target_precompile_headers(pch PUBLIC pch.h)