claude-3 instead of open AI

app.py

@@ -678,19 +678,44 @@ def generate_reading_passage(df_string):
         敘述中，請把數學或是專業術語，用 Latex 包覆（$...$），並且不要去改原本的文章
         加減乘除、根號、次方等等的運算式口語也換成 LATEX 數學符號
     """
-    messages = [
-        {"role": "system", "content": sys_content},
-        {"role": "user", "content": user_content}
-    ]
 
-    request_payload = {
-        "model": "gpt-4-turbo",
-        "messages": messages,
-        "max_tokens": 4000,
-    }
+    try:
+        # 使用 OPEN AI 生成 Reading Passage
+        messages = [
+            {"role": "system", "content": sys_content},
+            {"role": "user", "content": user_content}
+        ]
 
-    response = OPEN_AI_CLIENT.chat.completions.create(**request_payload)
-    reading_passage = response.choices[0].message.content.strip()
+        request_payload = {
+            "model": "gpt-4-turbo",
+            "messages": messages,
+            "max_tokens": 4000,
+        }
+
+        response = OPEN_AI_CLIENT.chat.completions.create(**request_payload)
+        reading_passage = response.choices[0].message.content.strip()
+    except:
+        # 使用 REDROCK 生成 Reading Passage
+        messages = [
+            {"role": "user", "content": user_content}
+        ]
+        model_id = "anthropic.claude-3-sonnet-20240229-v1:0"
+        # model_id = "anthropic.claude-3-haiku-20240307-v1:0"
+        kwargs = {
+            "modelId": model_id,
+            "contentType": "application/json",
+            "accept": "application/json",
+            "body": json.dumps({
+                "anthropic_version": "bedrock-2023-05-31",
+                "max_tokens": 4000,
+                "system": sys_content,
+                "messages": messages
+            })
+        }
+        response = BEDROCK_CLIENT.invoke_model(**kwargs)
+        response_body = json.loads(response.get('body').read())
+        reading_passage = response_body.get('content')[0].get('text')
+    
     print("=====reading_passage=====")
     print(reading_passage)
     print("=====reading_passage=====")
@@ -755,19 +780,43 @@ def generate_mind_map(df_string):
         注意：不需要前後文敘述，直接給出 markdown 文本即可
         這對我很重要
     """
-    messages = [
-        {"role": "system", "content": sys_content},
-        {"role": "user", "content": user_content}
-    ]
 
-    request_payload = {
-        "model": "gpt-4-turbo",
-        "messages": messages,
-        "max_tokens": 4000,
-    }
+    try:
+        # 使用 OPEN AI 生成
+        messages = [
+            {"role": "system", "content": sys_content},
+            {"role": "user", "content": user_content}
+        ]
 
-    response = OPEN_AI_CLIENT.chat.completions.create(**request_payload)
-    mind_map = response.choices[0].message.content.strip()
+        request_payload = {
+            "model": "gpt-4-turbo",
+            "messages": messages,
+            "max_tokens": 4000,
+        }
+
+        response = OPEN_AI_CLIENT.chat.completions.create(**request_payload)
+        mind_map = response.choices[0].message.content.strip()
+    except:
+        # 使用 REDROCK 生成
+        messages = [
+            {"role": "user", "content": user_content}
+        ]
+        model_id = "anthropic.claude-3-sonnet-20240229-v1:0"
+        # model_id = "anthropic.claude-3-haiku-20240307-v1:0"
+        kwargs = {
+            "modelId": model_id,
+            "contentType": "application/json",
+            "accept": "application/json",
+            "body": json.dumps({
+                "anthropic_version": "bedrock-2023-05-31",
+                "max_tokens": 4000,
+                "system": sys_content,
+                "messages": messages
+            })
+        }
+        response = BEDROCK_CLIENT.invoke_model(**kwargs)
+        response_body = json.loads(response.get('body').read())
+        mind_map = response_body.get('content')[0].get('text')
     print("=====mind_map=====")
     print(mind_map)
     print("=====mind_map=====")
@@ -891,19 +940,44 @@ def generate_summarise(df_string, metadata=None):
     #     💡 5. 結論反思（為什麼我們要學這個？）
     #     ❓ 6. 延伸小問題
 
-    messages = [
-        {"role": "system", "content": sys_content},
-        {"role": "user", "content": user_content}
-    ]
+    try: 
+        #OPEN AI
+        messages = [
+            {"role": "system", "content": sys_content},
+            {"role": "user", "content": user_content}
+        ]
+
+        request_payload = {
+            "model": "gpt-4-turbo",
+            "messages": messages,
+            "max_tokens": 4000,
+        }
+
+        response = OPEN_AI_CLIENT.chat.completions.create(**request_payload)
+        df_summarise = response.choices[0].message.content.strip()
+    except:
+        #REDROCK
+        messages = [
+            {"role": "user", "content": user_content}
+        ]
+        model_id = "anthropic.claude-3-sonnet-20240229-v1:0"
+        # model_id = "anthropic.claude-3-haiku-20240307-v1:0"
+        kwargs = {
+            "modelId": model_id,
+            "contentType": "application/json",
+            "accept": "application/json",
+            "body": json.dumps({
+                "anthropic_version": "bedrock-2023-05-31",
+                "max_tokens": 4000,
+                "system": sys_content,
+                "messages": messages
+            })
+        }
+        response = BEDROCK_CLIENT.invoke_model(**kwargs)
+        response_body = json.loads(response.get('body').read())
+        df_summarise = response_body.get('content')[0].get('text')
 
-    request_payload = {
-        "model": "gpt-4-turbo",
-        "messages": messages,
-        "max_tokens": 4000,
-    }
 
-    response = OPEN_AI_CLIENT.chat.completions.create(**request_payload)
-    df_summarise = response.choices[0].message.content.strip()
     print("=====df_summarise=====")
     print(df_summarise)
     print("=====df_summarise=====")
@@ -975,26 +1049,50 @@ def generate_questions(df_string):
 
     sys_content = "你是一個擅長資料分析跟影片教學的老師，user 為學生，請精讀資料文本，自行判斷資料的種類，並用既有資料為本質猜測用戶可能會問的問題，使用 zh-TW"
     user_content = f"請根據 {content_text} 生成三個問題，並用 JSON 格式返回 questions:[q1的敘述text, q2的敘述text, q3的敘述text]"
-    messages = [
-        {"role": "system", "content": sys_content},
-        {"role": "user", "content": user_content}
-    ]
-    response_format = { "type": "json_object" }
+    
+    try:
+        messages = [
+            {"role": "system", "content": sys_content},
+            {"role": "user", "content": user_content}
+        ]
+        response_format = { "type": "json_object" }
 
-    print("=====messages=====")
-    print(messages)
-    print("=====messages=====")
+        print("=====messages=====")
+        print(messages)
+        print("=====messages=====")
 
 
-    request_payload = {
-        "model": "gpt-4-turbo",
-        "messages": messages,
-        "max_tokens": 4000,
-        "response_format": response_format
-    }
+        request_payload = {
+            "model": "gpt-4-turbo",
+            "messages": messages,
+            "max_tokens": 4000,
+            "response_format": response_format
+        }
+
+        response = OPEN_AI_CLIENT.chat.completions.create(**request_payload)
+        questions = json.loads(response.choices[0].message.content)["questions"]
+    except:
+        messages = [
+            {"role": "user", "content": user_content}
+        ]
+        model_id = "anthropic.claude-3-sonnet-20240229-v1:0"
+        # model_id = "anthropic.claude-3-haiku-20240307-v1:0"
+        kwargs = {
+            "modelId": model_id,
+            "contentType": "application/json",
+            "accept": "application/json",
+            "body": json.dumps({
+                "anthropic_version": "bedrock-2023-05-31",
+                "max_tokens": 4000,
+                "system": sys_content,
+                "messages": messages
+            })
+        }
+        response = BEDROCK_CLIENT.invoke_model(**kwargs)
+        response_body = json.loads(response.get('body').read())
+        response_completion = response_body.get('content')[0].get('text')
+        questions = json.loads(response_completion)["questions"]
 
-    response = OPEN_AI_CLIENT.chat.completions.create(**request_payload)
-    questions = json.loads(response.choices[0].message.content)["questions"]
     print("=====json_response=====")
     print(questions)
     print("=====json_response=====")
@@ -1182,30 +1280,53 @@ def generate_key_moments(formatted_simple_transcript, formatted_transcript):
             "keywords": ["關鍵字", "關鍵字"]
         }}]
     """
-    messages = [
-        {"role": "system", "content": sys_content},
-        {"role": "user", "content": user_content}
-    ]
-    response_format = { "type": "json_object" }
-
-    request_payload = {
-        "model": "gpt-4-turbo",
-        "messages": messages,
-        "max_tokens": 4096,
-        "response_format": response_format
-    }
 
     try:
+        #OPEN AI
+        messages = [
+            {"role": "system", "content": sys_content},
+            {"role": "user", "content": user_content}
+        ]
+        response_format = { "type": "json_object" }
+
+        request_payload = {
+            "model": "gpt-4-turbo",
+            "messages": messages,
+            "max_tokens": 4096,
+            "response_format": response_format
+        }
+
         response = OPEN_AI_CLIENT.chat.completions.create(**request_payload)
         print("===response===")
         print(dict(response))
         key_moments = json.loads(response.choices[0].message.content)["key_moments"]
     except Exception as e:
-        error_msg = f" {video_id} 關鍵時刻錯誤: {str(e)}"
+        error_msg = f" {video_id} OPEN AI 關鍵時刻錯誤: {str(e)}"
         print("===generate_key_moments error===")
         print(error_msg)
         print("===generate_key_moments error===")
-        raise Exception(error_msg)
+        
+        #REDROCK
+        messages = [
+            {"role": "user", "content": user_content}
+        ]
+        model_id = "anthropic.claude-3-sonnet-20240229-v1:0"
+        # model_id = "anthropic.claude-3-haiku-20240307-v1:0"
+        kwargs = {
+            "modelId": model_id,
+            "contentType": "application/json",
+            "accept": "application/json",
+            "body": json.dumps({
+                "anthropic_version": "bedrock-2023-05-31",
+                "max_tokens": 4096,
+                "system": sys_content,
+                "messages": messages
+            })
+        }
+        response = BEDROCK_CLIENT.invoke_model(**kwargs)
+        response_body = json.loads(response.get('body').read())
+        response_completion = response_body.get('content')[0].get('text')
+        key_moments = json.loads(response_completion)["key_moments"]
 
     print("=====key_moments=====")
     print(key_moments)
@@ -1229,18 +1350,43 @@ def generate_key_moments_keywords(transcript):
         不用給上下文，直接給出關鍵字，使用 zh-TW，用逗號分隔， example: 關鍵字1, 關鍵字2
         transcript：{transcript}
     """
-    messages = [
-        {"role": "system", "content": system_content},
-        {"role": "user", "content": user_content}
-    ]
-    request_payload = {
-        "model": "gpt-4-turbo",
-        "messages": messages,
-        "max_tokens": 100,
-    }
 
-    response = OPEN_AI_CLIENT.chat.completions.create(**request_payload)
-    keywords = response.choices[0].message.content.strip().split(", ")
+    try:
+        # OPEN AI
+        messages = [
+            {"role": "system", "content": system_content},
+            {"role": "user", "content": user_content}
+        ]
+        request_payload = {
+            "model": "gpt-4-turbo",
+            "messages": messages,
+            "max_tokens": 100,
+        }
+
+        response = OPEN_AI_CLIENT.chat.completions.create(**request_payload)
+        keywords = response.choices[0].message.content.strip().split(", ")
+    except:
+        # REDROCK
+        messages = [
+            {"role": "user", "content": user_content}
+        ]
+        model_id = "anthropic.claude-3-sonnet-20240229-v1:0"
+        # model_id = "anthropic.claude-3-haiku-20240307-v1:0"
+        kwargs = {
+            "modelId": model_id,
+            "contentType": "application/json",
+            "accept": "application/json",
+            "body": json.dumps({
+                "anthropic_version": "bedrock-2023-05-31",
+                "max_tokens": 100,
+                "system": system_content,
+                "messages": messages
+            })
+        }
+        response = BEDROCK_CLIENT.invoke_model(**kwargs)
+        response_body = json.loads(response.get('body').read())
+        response_completion = response_body.get('content')[0].get('text')
+        keywords = response_completion.strip().split(", ")
     
     return keywords
 
@@ -1599,7 +1745,6 @@ def create_LLM_content(video_id, df_string, kind):
 def reading_passage_add_latex_version(video_id):
     # 確認 GCS 是否有 reading_passage.json
     print("===reading_passage_convert_to_latex===")
-    gcs_client = GCS_CLIENT
     bucket_name = 'video_ai_assistant'
     file_name = f'{video_id}_reading_passage.json'
     blob_name = f"{video_id}/{file_name}"
@@ -1652,7 +1797,6 @@ def reading_passage_add_latex_version(video_id):
 def summary_add_markdown_version(video_id):
     # 確認 GCS 是否有 summary.json
     print("===summary_convert_to_markdown===")
-    gcs_client = GCS_CLIENT
     bucket_name = 'video_ai_assistant'
     file_name = f'{video_id}_summary.json'
     blob_name = f"{video_id}/{file_name}"
@@ -1826,7 +1970,7 @@ def generate_ai_content(password, df_string, topic, grade, level, specific_featu
     request_payload = {
         "model": ai_model_name,
         "messages": messages,
-        "max_tokens": 4000  # 举例，实际上您可能需要更详细的配置
+        "max_tokens": 4000
     }
     ai_content = material.send_ai_request(OPEN_AI_CLIENT, request_payload)
     return ai_content, prompt