ask in prompt to send 200 dialogues; calc wpm using length of the audio

backend/app/prompts.py

@@ -19,7 +19,7 @@
 Discuss in the podcast, the main components of the system (e.g., frontend, backend, database, building, external services).
 Discuss the relationships and interactions between these components.
 Sometimes the answers can also be single word or very small so that it seems natural. Long answers all the time makes it monotonous.
-Make it a 10 minute long or longer podcast if possible.  Give atleast 80 voice tags for the host which is probably Ava here (160 including closing tag). + Same amount of voice tags for guest. Slowly count them and re-write the ssml if its falling short and then return the ssml."""
+Make it a 20 minute long or longer podcast if possible.  Give atleast 200 voice tags for the host + Same amount of voice tags for guest. Slowly count them and re-write the ssml if its falling short and then return the ssml."""
 
 SYSTEM_FIRST_PROMPT = """
 You are tasked with explaining to a principal software engineer how to draw the best and most accurate system design diagram / architecture of a given project. This explanation should be tailored to the specific project's purpose and structure. To accomplish this, you will be provided with two key pieces of information:

backend/app/routers/generate.py

@@ -13,6 +13,8 @@
 from tempfile import NamedTemporaryFile
 import base64
 import xml.etree.ElementTree as ET
+from pydub import AudioSegment
+import io
 
 load_dotenv()
 
@@ -37,7 +39,8 @@ def get_cached_github_data(username: str, repo: str):
         file_list = openai_service.get_important_files(file_tree)
         for fpath in file_list:
             content = github_service.get_github_file_content(username, repo, fpath)
-            file_content += f"FPATH: fpath CONTENT:{content}"
+            discuss_or_not = "- discuss this file." if '.md' not in fpath else ""
+            file_content += f"FPATH: {fpath} {discuss_or_not} \n CONTENT:{content}"
     except Exception as e:
         print(f"Some error in getting github file content {e}. Proceeding.")
 
@@ -90,7 +93,15 @@ def calculate_duration(text_line, wpm=135):
     seconds = minutes * 60
     return seconds
 
-def ssml_to_webvtt(ssml_content, max_line_length=45, max_words_per_cue=30):
+def no_of_words(text_lines):
+    if isinstance(text_lines, str):  # If it's a single string
+        return len([word for word in text_lines.split() if word])
+    elif isinstance(text_lines, list):  # If it's a list of strings
+        return sum(len([word for word in line.split() if word]) for line in text_lines)
+    else:
+        return 0
+
+def ssml_to_webvtt(ssml_content, duration_in_seconds, max_line_length=45, max_words_per_cue=30):
     # Helper function to insert line breaks at appropriate places
     def add_line_breaks(text, max_length):
         words = text.split()
@@ -110,13 +121,14 @@ def add_line_breaks(text, max_length):
     text_content = re.sub(r'<speak[^>]*>|</speak>|<break[^>]*>', '', ssml_content)
     text_content = re.sub(r'<voice[^>]*>', '\n\n', text_content)
     text_content = re.sub(r'</voice>', '', text_content)
-    text_lines = filter(None, [line.strip() for line in text_content.splitlines()])
+    text_lines = list(filter(None, [line.strip() for line in text_content.splitlines()]))
 
     # Step 2: Generate WebVTT content with sequential timestamps
     vtt_content = "WEBVTT\n\n"
     cumulative_time = 0.0
     cue_index = 0
-
+    wpm = int(no_of_words(text_lines) / duration_in_seconds * 60)
+    print(wpm, " Words per minute")
     for i, line in enumerate(text_lines):
 
         # Break the line if it's too long into sub-lines based on word count
@@ -128,7 +140,7 @@ def add_line_breaks(text, max_length):
 
         # Generate VTT for each sub-line
         for sub_line in sub_lines:
-            duration = calculate_duration(sub_line)
+            duration = calculate_duration(sub_line, wpm=wpm)
             start_time = cumulative_time
             end_time = start_time + duration
             cumulative_time = end_time  # Update cumulative time for next line
@@ -170,7 +182,7 @@ async def generate(request: Request, body: ApiRequest):
         file_content = github_data["file_content"]
 
         # Check combined token count
-        combined_content = f"{file_tree}\n{readme}\n{file_content}"
+        combined_content = f"FILE TREE: {file_tree}\n README: {readme}\n IMPORTANT FILES: {file_content}"
         combined_content = combined_content[:250000]
         print(combined_content)
         try:
@@ -218,7 +230,11 @@ async def generate(request: Request, body: ApiRequest):
             # mp3_bytes = convert_wav_to_mp3(audio_bytes)
             if audio_bytes:
                 response = Response(content=audio_bytes, media_type="audio/mpeg", headers={"Content-Disposition": "attachment; filename=explanation.mp3"})
-                vtt_content = ssml_to_webvtt(ssml_response)
+                # Assuming audio_bytes contains the audio data
+                audio = AudioSegment.from_file(io.BytesIO(audio_bytes), format="mp3")
+                duration_in_seconds = len(audio) / 1000.0
+                print("duration in sec", duration_in_seconds)
+                vtt_content = ssml_to_webvtt(ssml_response, duration_in_seconds)
                 encoded_vtt_content = base64.b64encode(vtt_content.encode('utf-8')).decode('utf-8')
                 response.headers["X-VTT-Content"] = encoded_vtt_content
                 # Add CORS headers

backend/app/services/openai_service.py

@@ -29,7 +29,7 @@ def call_openai_for_response(self, files_path, ssml_prompt_text):
         """
         # Read the content of the file specified by files_path
         with open(files_path[0], 'r') as file:
-            file_content = file.read()
+            file_content = file.read()  # this has everything readme + tree + other files
         # Send the prompt to Azure OpenAI for processing
         response = openai.chat.completions.create(
             model=self.model_name,