ADDED: NoteGem

README.md

@@ -212,6 +212,16 @@ percentage, missing keywords, and profile summary.
   - **Image Quality Note:** Indicates if certain aspects are 'Unable to be determined based on the provided image.'
   - **Disclaimer:** Includes the disclaimer: "Consult with a Doctor before making any decisions."
 
+### 24. NoteGem: Automated Note-Taking Assistant
+
+- **Route:** `/NoteGem`
+- **Description:** This API endpoint leverages the Google Gemini AI Model to generate comprehensive notes from YouTube video transcripts.
+- **Feature:**
+  - **Input Video URL:** Users can provide a YouTube video URL for processing.
+  - **Transcript Extraction:** The API extracts the transcript from the provided YouTube video.
+  - **Error Handling for Transcripts:** If the transcript is not available, it returns a message indicating that the transcript is not available for transcription.
+  - **AI Summary Generation:** The AI model generates a structured summary of the transcript focusing on main points, critical information, key takeaways, examples or case studies, quotes, and actionable steps.
+
 ## Usage
 
 Each endpoint accepts specific parameters as described in the respective endpoint documentation. Users can make POST requests to these endpoints with the required parameters to perform the desired tasks.

api.py

@@ -15,7 +15,7 @@
 from mongo import MongoDB
 from helper_functions import get_qa_chain,get_gemini_response,get_url_doc_qa,extract_transcript_details,\
     get_gemini_response_health,get_gemini_pdf,read_sql_query,remove_substrings,questions_generator,groq_pdf,\
-    summarize_audio,chatbot_send_message,extraxt_pdf_text,advance_rag_llama_index,parse_sql_response
+    summarize_audio,chatbot_send_message,extraxt_pdf_text,advance_rag_llama_index,parse_sql_response, extract_video_id
 from langchain_groq import ChatGroq
 from langchain.chains.conversation.base import ConversationChain
 from langchain.chains.conversation.memory import ConversationBufferWindowMemory
@@ -38,6 +38,7 @@
 from langchain_community.agent_toolkits import SQLDatabaseToolkit
 import tempfile
 import shutil
+from youtube_transcript_api import YouTubeTranscriptApi, TranscriptsDisabled, NoTranscriptFound
 
 
 os.environ["LANGCHAIN_TRACING_V2"]="true"
@@ -737,6 +738,37 @@ async def medigem(image_file: UploadFile = File(...)):
     result = db.insert_data(mongo_data)
     print(result)
     return ResponseText(response=remove_substrings(response.text))
+
+@app.post("/NoteGem", description="This API endpoint leverages the Google Gemini AI Model to generate comprehensive notes from YouTube video transcripts")
+def process_video(video_url: str = Form(...)):
+    video_id = extract_video_id(video_url)
+    if not video_id:
+        raise HTTPException(status_code=400, detail="Invalid YouTube URL")
+
+    try:
+        transcript_text = YouTubeTranscriptApi.get_transcript(video_id)
+        transcript = " ".join([i["text"] for i in transcript_text])
+    except (TranscriptsDisabled, NoTranscriptFound):
+        return {"transcript": "Transcript not available", "error": True}
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=str(e))
+
+    try:
+        model = genai.GenerativeModel(settings.GEMINI_PRO_1_5)
+        response = model.generate_content(settings.NOTE_GEN_PROMPT + transcript)
+        summary = response.text
+        db = MongoDB()
+        payload = {
+                "endpoint" : "/NoteGem",
+                "video_url": video_url,
+                "output" : response.text
+            }
+        mongo_data = {"Document": payload}
+        result = db.insert_data(mongo_data)
+        print(result)
+        return ResponseText(response=summary)
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=str(e))
 
 if __name__ == '__main__':
     import uvicorn

helper_functions.py

@@ -338,6 +338,21 @@ def parse_sql_response(response):
     formatted_response = '\n'.join(formatted_sql_statements)
     return formatted_response
 
+def extract_video_id(url):
+    video_id = None
+    regex_patterns = [
+        r"(?<=v=)[^#\&\?]*",
+        r"(?<=be/)[^#\&\?]*",
+        r"(?<=embed/)[^#\&\?]*",
+        r"(?<=youtu.be/)[^#\&\?]*"
+    ]
+    for pattern in regex_patterns:
+        match = re.search(pattern, url)
+        if match:
+            video_id = match.group(0)
+            break
+    return video_id
+
 if __name__ == "__main__":
     create_vector_db()
     chain = get_qa_chain()

settings.py

@@ -141,4 +141,17 @@
 
     Please provide me an output with these 4 headings Detailed Analysis, Findings Report, Recommendations and Next Steps and Treatment Suggestions.
 
-    """
+    """
+
+NOTE_GEN_PROMPT = """
+You are a professional note-taker with expertise in distilling key insights from video content. Your task is to generate a comprehensive, yet concise set of notes from the provided video transcript. Focus on the following:
+
+1. Main points
+2. Critical information
+3. Key takeaways
+4. Examples or case studies
+5. Quotes or important statements
+6. Actionable steps or recommendations
+
+Make sure the notes are well-structured and formatted as bullet points. The total length should not exceed 1000 words. Please summarize the following text:
+"""