update

llm/deepinfra_api.py

@@ -25,7 +25,7 @@ class DeepInfraApi(LlmApi):
         """
         try:
             async with httpx.AsyncClient() as client:
-                response = await client.get(f"{super().params.url}/v1/openai/models", super().create_headers())
+                response = await client.get(f"{self.params.url}/v1/openai/models", super().create_headers())
                 if response.status_code == 200:
                     json_data = response.json()
                     return [item['id'] for item in json_data.get('data', [])]
@@ -45,8 +45,8 @@ class DeepInfraApi(LlmApi):
         """
         actual_prompt = self.apply_llm_template_to_prompt(prompt)
         messages = []
-        if super().params.predict_params and super().params.predict_params.system_prompt:
-            messages.append({"role": "system", "content": super().params.predict_params.system_prompt})
+        if self.params.predict_params and self.params.predict_params.system_prompt:
+            messages.append({"role": "system", "content": self.params.predict_params.system_prompt})
         messages.append({"role": "user", "content": actual_prompt})
         return messages
 
@@ -61,8 +61,8 @@ class DeepInfraApi(LlmApi):
             str: Промпт с примененным шаблоном (или оригинальный, если шаблон отсутствует).
         """
         actual_prompt = prompt
-        if super().params.template is not None:
-            actual_prompt = super().params.template.replace("{{PROMPT}}", actual_prompt)
+        if self.params.template is not None:
+            actual_prompt = self.params.template.replace("{{PROMPT}}", actual_prompt)
         return actual_prompt
 
     async def tokenize(self, prompt: str) -> Optional[dict]:
@@ -81,15 +81,15 @@ class DeepInfraApi(LlmApi):
         Returns:
             dict: Словарь с параметрами для выполнения запроса.
         """
-        print(super().params)
+        print(self.params)
         print(self.params)
         
         request = {
             "stream": False,
-            "model": super().params.model,
+            "model": self.params.model,
         }
 
-        predict_params = super().params.predict_params
+        predict_params = self.params.predict_params
         if predict_params:
             if predict_params.stop:
                 non_empty_stop = list(filter(lambda o: o != "", predict_params.stop))
@@ -150,6 +150,6 @@ class DeepInfraApi(LlmApi):
             request = await self.create_request(prompt)
 
             async with httpx.AsyncClient() as client:
-                response = client.post(f"{super().params.url}/v1/openai/chat/completions", super().create_headers(), json=request)
+                response = client.post(f"{self.params.url}/v1/openai/chat/completions", super().create_headers(), json=request)
                 if response.status_code == 200:
                     return response.json()["choices"][0]["message"]["content"]

llm/vllm_api.py

@@ -27,7 +27,7 @@ class LlmApi(LlmApi):
         """
         try:
             async with httpx.AsyncClient() as client:
-                response = await client.get(f"{super().params.url}/v1/models", super().create_headers())
+                response = await client.get(f"{self.params.url}/v1/models", super().create_headers())
                 if response.status_code == 200:
                     json_data = response.json()
                     return [item['id'] for item in json_data.get('data', [])]
@@ -37,8 +37,8 @@ class LlmApi(LlmApi):
     
     async def get_model(self) -> str:
         model = None
-        if super().params.model is not None:
-            model = super().params.model
+        if self.params.model is not None:
+            model = self.params.model
         else:
             models = await self.get_models()
             model = models[0] if models else None
@@ -60,8 +60,8 @@ class LlmApi(LlmApi):
         """
         actual_prompt = self.apply_llm_template_to_prompt(prompt)
         messages = []
-        if super().params.predict_params and super().params.predict_params.system_prompt:
-            messages.append({"role": "system", "content": super().params.predict_params.system_prompt})
+        if self.params.predict_params and self.params.predict_params.system_prompt:
+            messages.append({"role": "system", "content": self.params.predict_params.system_prompt})
         messages.append({"role": "user", "content": actual_prompt})
         return messages
 
@@ -76,8 +76,8 @@ class LlmApi(LlmApi):
             str: Промпт с примененным шаблоном (или оригинальный, если шаблон отсутствует).
         """
         actual_prompt = prompt
-        if super().params.template is not None:
-            actual_prompt = super().params.template.replace("{{PROMPT}}", actual_prompt)
+        if self.params.template is not None:
+            actual_prompt = self.params.template.replace("{{PROMPT}}", actual_prompt)
         return actual_prompt
 
     async def tokenize(self, prompt: str) -> Optional[dict]:
@@ -102,7 +102,7 @@ class LlmApi(LlmApi):
         try:
             async with httpx.AsyncClient() as client:
                 response = await client.post(
-                    f"{super().params.url}/tokenize",
+                    f"{self.params.url}/tokenize",
                     json=request_data,
                     headers=super().create_headers(),
                 )
@@ -136,7 +136,7 @@ class LlmApi(LlmApi):
         try:
             async with httpx.AsyncClient() as client:
                 response = await client.post(
-                    f"{super().params.url}/detokenize",
+                    f"{self.params.url}/detokenize",
                     json=request_data,
                     headers=super().create_headers(),
                 )
@@ -170,7 +170,7 @@ class LlmApi(LlmApi):
             "model": model,
         }
 
-        predict_params = super().params.predict_params
+        predict_params = self.params.predict_params
         if predict_params:
             if predict_params.stop:
                 non_empty_stop = list(filter(lambda o: o != "", predict_params.stop))
@@ -251,7 +251,7 @@ class LlmApi(LlmApi):
         # Максимально допустимое количество токенов для источников
         max_length = (
             max_token_count
-            - (super().params.predict_params.n_predict or 0)
+            - (self.params.predict_params.n_predict or 0)
             - aux_token_count
             - system_prompt_token_count
         )
@@ -290,7 +290,7 @@ class LlmApi(LlmApi):
             request = await self.create_request(prompt)
 
             # Начинаем потоковый запрос
-            async with client.stream("POST", f"{super().params.url}/v1/chat/completions", json=request) as response:
+            async with client.stream("POST", f"{self.params.url}/v1/chat/completions", json=request) as response:
                 if response.status_code != 200:
                     # Если ошибка, читаем ответ для получения подробностей
                     error_content = await response.aread()