gpt

LunaTranslator/LunaTranslator/translator/gptcommon.py

@@ -1,5 +1,5 @@
 from translator.basetranslator import basetrans
-import json
+import json, requests
 from traceback import print_exc
 
 
@@ -28,10 +28,25 @@ class gptcommon(basetrans):
         self.context = []
         super().__init__(typename)
 
-    def createurl(self):
+    def createdata(self, message):
-        if self.config["API接口地址"].endswith("/chat/completions"):
+        try:
-            return self.config["API接口地址"]
+            temperature = float(self.config["Temperature"])
-        return self.checkv1(self.config["API接口地址"]) + "/chat/completions"
+        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"],
+            stream=self.config["流式输出"],
+        )
+        return data
 
     def createparam(self):
         return None
@@ -54,14 +69,88 @@ class gptcommon(basetrans):
             api_url += "/v1"
         return api_url
 
+    def alicreateheaders(self):
+        h = self.createheaders()
+        if self.config["流式输出"]:
+            h.update({"Accept": "text/event-stream"})
+        return h
+
+    def alicreatedata(self, message):
+
+        data = dict(model=self.config["model"], input=dict(messages=message))
+        if self.config["流式输出"]:
+            data.update(dict(parameters=dict(incremental_output=True)))
+
+        return data
+
+    def aliparseresponse(self, query, response: requests.ResponseBase, usingstream):
+        if usingstream:
+            message = ""
+            for chunk in response.iter_lines():
+                response_data = chunk.decode("utf-8").strip()
+                if not response_data:
+                    continue
+                if response_data.startswith("data:") == False:
+                    continue
+                try:
+                    json_data = json.loads(response_data[5:])
+                    msg = json_data["output"]["text"]
+                    yield msg
+                    message += msg
+                except:
+                    print_exc()
+                    raise Exception(response_data)
+
+                if json_data["output"]["finish_reason"] == "stop":
+                    break
+
+        else:
+            try:
+                message = (
+                    response.json()["output"]["text"].replace("\n\n", "\n").strip()
+                )
+                yield message
+            except:
+                raise Exception(response.text)
+        self.context.append({"role": "user", "content": query})
+        self.context.append({"role": "assistant", "content": message})
+
+    def commonparseresponse(self, query, response: requests.ResponseBase, usingstream):
+        if usingstream:
+            message = ""
+            for chunk in response.iter_lines():
+                response_data = chunk.decode("utf-8").strip()
+                if not response_data:
+                    continue
+                try:
+                    json_data = json.loads(response_data[6:])
+                    rs = json_data["choices"][0].get("finish_reason")
+                    if rs and rs != "null":
+                        break
+                    msg = json_data["choices"][0]["delta"]["content"]
+                    yield msg
+                    message += msg
+
+                except:
+                    print_exc()
+                    raise Exception(response_data)
+        else:
+            try:
+
+                message = (
+                    response.json()["choices"][0]["message"]["content"]
+                    .replace("\n\n", "\n")
+                    .strip()
+                )
+                yield message
+            except:
+                raise Exception(response.text)
+        self.context.append({"role": "user", "content": query})
+        self.context.append({"role": "assistant", "content": message})
+
     def translate(self, query):
         self.contextnum = int(self.config["附带上下文个数"])
 
-        try:
-            temperature = float(self.config["Temperature"])
-        except:
-            temperature = 0.3
-
         if self.config["使用自定义promt"]:
             message = [{"role": "user", "content": self.config["自定义promt"]}]
         else:
@@ -85,52 +174,25 @@ class gptcommon(basetrans):
         message.append({"role": "user", "content": query})
 
         usingstream = self.config["流式输出"]
-        data = dict(
+        url = self.config["API接口地址"]
-            model=self.config["model"],
+        parseresponse = self.commonparseresponse
-            messages=message,
+        createheaders = self.createheaders
-            # optional
+        createdata = self.createdata
-            max_tokens=self.config["max_tokens"],
+        if url.endswith("/chat/completions"):
-            n=1,
+            pass
-            # stop=None,
+        elif url.endswith("/text-generation/generation") or 'dashscope.aliyuncs.com' in url:
-            top_p=self.config["top_p"],
+            # https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation
-            temperature=temperature,
+            # 阿里云百炼
-            frequency_penalty=self.config["frequency_penalty"],
+            parseresponse = self.aliparseresponse
-            stream=usingstream,
+            createheaders = self.alicreateheaders
-        )
+            createdata = self.alicreatedata
-        response = self.proxysession.post(
-            self.createurl(),
-            headers=self.createheaders(),
-            params=self.createparam(),
-            json=data,
-            stream=usingstream,
-        )
-        if usingstream:
-            message = ""
-            for chunk in response.iter_lines():
-                response_data = chunk.decode("utf-8").strip()
-                if not response_data:
-                    continue
-                try:
-                    json_data = json.loads(response_data[6:])
-                    rs = json_data["choices"][0]["finish_reason"]
-                    if rs and rs != "null":
-                        break
-                    msg = json_data["choices"][0]["delta"]["content"]
-                    yield msg
-                    message += msg
-                except:
-                    print_exc()
-                    raise Exception(response_data)
-
         else:
-            try:
+            url = self.checkv1(url) + "/chat/completions"
-                message = (
+        response = self.proxysession.post(
-                    response.json()["choices"][0]["message"]["content"]
+            url,
-                    .replace("\n\n", "\n")
+            headers=createheaders(),
-                    .strip()
+            params=self.createparam(),
-                )
+            json=createdata(message),
-                yield message
+            stream=usingstream,
-            except:
+        )
-                raise Exception(response.text)
+        return parseresponse(query, response, usingstream)
-        self.context.append({"role": "user", "content": query})
-        self.context.append({"role": "assistant", "content": message})

docs/zh/guochandamoxing.md

@@ -1,6 +1,6 @@
 ## 国产大模型如何使用ChatGPT兼容接口
 
-### 火山引擎（豆包大模型等）
+### 字节跳动豆包大模型等
 
 **API接口地址** `https://ark.cn-beijing.volces.com/api/v3`
 
@@ -10,3 +10,8 @@
 
 ![img](https://image.lunatranslator.xyz/zh/damoxing/doubao.png)
 
+
+### 阿里云百炼大模型
+
+**API接口地址** `https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation`
+