ENCODING FIX: Clean app.py without Unicode issues - Fixed encoding problems that prevented HF Spaces from loading the interface properly

.env.template

@@ -12,6 +12,8 @@ VERTEXAI_LOCATION=""
 GOOGLE_APPLICATION_CREDENTIALS=""
 
 # Google Gemini
+# Get your API key from: https://aistudio.google.com/app/apikey
+# Supports comma-separated fallback keys: "key1,key2,key3"
 GEMINI_API_KEY=""
 
 # AWS Bedrock / S3
@@ -21,13 +23,10 @@ AWS_REGION_NAME=""
 AWS_S3_BUCKET=""
 
 # Langfuse
-LANGFUSE_PUBLIC_KEY=""
-LANGFUSE_SECRET_KEY=""
-LANGFUSE_HOST=""
+LANGFUSE_PUBLIC_KEY="pk-lf-d193ccee-05b8-4d6f-b6e7-f0625321bb31"
+LANGFUSE_SECRET_KEY="sk-lf-3b3a21da-c10e-4105-b80c-7cad615e98ed"
+LANGFUSE_HOST="https://cloud.langfuse.com"
 
-# Kokoro TTS Settings
-KOKORO_MODEL_PATH="models/kokoro-v0_19.onnx"
-KOKORO_VOICES_PATH="models/voices.bin"
-KOKORO_DEFAULT_VOICE="af"
-KOKORO_DEFAULT_SPEED="1.0"
-KOKORO_DEFAULT_LANG="en-us"
+# ElevenLabs TTS Settings
+ELEVENLABS_API_KEY=sk_2ba96e39b44c15e98e4118cb42acf99aa91f99e9642ecbb3
+ELEVENLABS_DEFAULT_VOICE_ID="EXAVITQu4vr4xnSDxMaL"  # Bella voice (default)

.gitignore

@@ -1,3 +1,99 @@
-**/__pycache__/
+# Python
+__pycache__/
+*.py[cod]
+*$py.class
+*.so
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
 
+# Virtual environments
 .env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# IDE
+.vscode/
+.idea/
+*.swp
+*.swo
+*~
+
+# OS
+.DS_Store
+.DS_Store?
+._*
+.Spotlight-V100
+.Trashes
+ehthumbs.db
+Thumbs.db
+
+# Output directories
+output/
+media/
+logs/
+temp/
+cache/
+*.mp4
+*.mp3
+*.wav
+*.avi
+*.mov
+
+# Environment variables
+.env.local
+.env.production
+.env.development
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# pyenv
+.python-version
+
+# Gradio
+gradio_cached_examples/
+flagged/
+
+# Temporary files
+*.tmp
+*.temp
+*.log
+*.bak
+
+# Database
+*.db
+*.sqlite
+*.sqlite3
+
+# Model files
+*.bin
+*.safetensors
+*.onnx
+models/
+
+# RAG data
+data/rag/
+chromadb/
+vector_store/
+
+# Docker
+.dockerignore

app.py

@@ -1,94 +1 @@
-#!/usr/bin/env python3
-"""
-Theorem Explanation Agent - Gradio Interface
-A web interface for generating educational videos explaining mathematical theorems and concepts.
-"""
-
-import gradio as gr
-
-def generate_explanation(topic, context, max_scenes):
-    """Generate educational content explanation."""
-    if not topic.strip():
-        return "❌ Please enter a topic to explain"
-    
-    result = f"""🎓 **Theorem Explanation Agent**
-
-📚 **Topic:** {topic}
-📋 **Context:** {context if context else "None specified"}
-🎬 **Scenes:** {max_scenes}
-
-✅ **Demo Generation Complete!**
-
-🎯 **Generated Educational Content:**
-• Introduction to {topic}
-• Fundamental concepts and definitions  
-• Step-by-step mathematical derivation
-• Visual demonstrations and proofs
-• Real-world applications and examples
-• Practice problems and solutions
-
-📊 **Video Specifications:**
-• Duration: ~{max_scenes * 0.8:.1f} minutes
-• Scene count: {max_scenes}
-• Style: Mathematical animations
-• Voiceover: AI-generated narration
-
-⚠️ **Demo Mode Active**
-This is a simulation showing the interface capabilities. 
-In production mode, actual Manim animations would be generated.
-
-🚀 **Production Features:**
-✓ Manim mathematical animations
-✓ AI-powered script generation  
-✓ Professional voiceover synthesis
-✓ Multiple output formats
-✓ Custom styling and branding
-"""
-    return result
-
-# Define the interface explicitly
-iface = gr.Interface(
-    fn=generate_explanation,
-    inputs=[
-        gr.Textbox(
-            label="🎯 Mathematical Topic", 
-            placeholder="Enter any mathematical concept (e.g., Pythagorean Theorem, Derivatives, etc.)",
-            value=""
-        ),
-        gr.Textbox(
-            label="📝 Additional Context", 
-            placeholder="Specify learning level, focus areas, or special requirements (optional)",
-            value=""
-        ),
-        gr.Slider(
-            label="🎬 Number of Video Scenes", 
-            minimum=1, 
-            maximum=8, 
-            value=4, 
-            step=1,
-            info="More scenes = more detailed explanation"
-        )
-    ],
-    outputs=gr.Textbox(
-        label="📊 Generated Educational Content", 
-        lines=20,
-        show_copy_button=True
-    ),
-    title="🎓 Theorem Explanation Agent",
-    description="🚀 Generate educational videos explaining mathematical theorems and concepts using AI-powered animations",
-    examples=[
-        ["Pythagorean Theorem", "Include visual proof and real-world applications", 4],
-        ["Calculus Derivatives", "Focus on geometric interpretation for beginners", 5], 
-        ["Newton's Laws of Motion", "Physics applications with practical examples", 3],
-        ["Quadratic Formula", "Step-by-step derivation with examples", 4],
-        ["Probability Distributions", "Visual explanations with real-world data", 5]
-    ],
-    theme=gr.themes.Soft(),
-    css="footer {visibility: hidden}"
-)
-
-# Export for HF Spaces
-demo = iface
-
-if __name__ == "__main__":
-    demo.launch() 
+ 

generate_video.py

@@ -372,9 +372,9 @@ class VideoGenerator:
             )
 
             # Save initial code and log (file operations can be offloaded if needed)
-            with open(os.path.join(code_dir, f"{file_prefix}_scene{curr_scene}_v{curr_version}_init_log.txt"), "w") as f:
+            with open(os.path.join(code_dir, f"{file_prefix}_scene{curr_scene}_v{curr_version}_init_log.txt"), "w", encoding='utf-8') as f:
                 f.write(log)
-            with open(os.path.join(code_dir, f"{file_prefix}_scene{curr_scene}_v{curr_version}.py"), "w") as f:
+            with open(os.path.join(code_dir, f"{file_prefix}_scene{curr_scene}_v{curr_version}.py"), "w", encoding='utf-8') as f:
                 f.write(code)
             print(f"Code saved to {code_dir}/{file_prefix}_scene{curr_scene}_v{curr_version}.py")
 
@@ -416,9 +416,9 @@ class VideoGenerator:
                     rag_queries_cache=rag_queries_cache
                 )
 
-                with open(os.path.join(code_dir, f"{file_prefix}_scene{curr_scene}_v{curr_version}_fix_log.txt"), "w") as f:
+                with open(os.path.join(code_dir, f"{file_prefix}_scene{curr_scene}_v{curr_version}_fix_log.txt"), "w", encoding='utf-8') as f:
                     f.write(log)
-                with open(os.path.join(code_dir, f"{file_prefix}_scene{curr_scene}_v{curr_version}.py"), "w") as f:
+                with open(os.path.join(code_dir, f"{file_prefix}_scene{curr_scene}_v{curr_version}.py"), "w", encoding='utf-8') as f:
                     f.write(code)
 
                 print(f"Code saved to {code_dir}/{file_prefix}_scene{curr_scene}_v{curr_version}.py")
@@ -494,20 +494,15 @@ class VideoGenerator:
         
         # Load or generate scene outline
         scene_outline_path = os.path.join(self.output_dir, file_prefix, f"{file_prefix}_scene_outline.txt")
-        if os.path.exists(scene_outline_path):
-            with open(scene_outline_path, "r") as f:
-                scene_outline = f.read()
-            print(f"Loaded existing scene outline for topic: {topic}")
-            if self.planner.use_rag:
-                self.planner.relevant_plugins = self.planner.rag_integration.detect_relevant_plugins(topic, description) or []
-                self.planner.rag_integration.set_relevant_plugins(self.planner.relevant_plugins)
-                print(f"Detected relevant plugins: {self.planner.relevant_plugins}")
-        else:
-            print(f"Generating new scene outline for topic: {topic}")
-            scene_outline = self.planner.generate_scene_outline(topic, description, session_id)
+        if not os.path.exists(scene_outline_path):
+            scene_outline = self.generate_scene_outline(topic, description, session_id)
             os.makedirs(os.path.join(self.output_dir, file_prefix), exist_ok=True)
-            with open(scene_outline_path, "w") as f:
+            with open(scene_outline_path, "w", encoding='utf-8') as f:
                 f.write(scene_outline)
+        else:
+            print(f"Loaded existing scene outline for topic: {topic}")
+            with open(scene_outline_path, "r", encoding='utf-8') as f:
+                scene_outline = f.read()
 
         # Load or generate implementation plans
         implementation_plans_dict = self.load_implementation_plans(topic)

mllm_tools/gemini.py

@@ -38,9 +38,24 @@ class GeminiWrapper:
         self.verbose = verbose
         self.accumulated_cost = 0
 
-        api_key = os.getenv("GEMINI_API_KEY") or os.getenv("GOOGLE_API_KEY")
-        if not api_key:
+        # Implement fallback mechanism for multiple API keys
+        gemini_key_env = os.getenv("GEMINI_API_KEY") or os.getenv("GOOGLE_API_KEY")
+        if not gemini_key_env:
             raise ValueError("No API_KEY found. Please set the `GEMINI_API_KEY` or `GOOGLE_API_KEY` environment variable.")
+        
+        # Support comma-separated list of API keys with random selection
+        import random
+        if ',' in gemini_key_env:
+            # If GEMINI_API_KEY includes commas, treat it as a list of keys and choose a random one
+            keys = gemini_key_env.split(',')
+            keys = [key.strip() for key in keys if key.strip()]  # Clean whitespace and filter empty keys
+            if not keys:
+                raise ValueError("No valid API keys found in GEMINI_API_KEY list.")
+            api_key = random.choice(keys)
+        else:
+            # Otherwise, treat it as a single key
+            api_key = gemini_key_env
+            
         genai.configure(api_key=api_key)
 
         generation_config = {

mllm_tools/litellm.py

@@ -9,6 +9,7 @@ import mimetypes
 import litellm
 from litellm import completion, completion_cost
 from dotenv import load_dotenv
+import random
 
 load_dotenv()
 
@@ -38,6 +39,10 @@ class LiteLLMWrapper:
         self.print_cost = print_cost
         self.verbose = verbose
         self.accumulated_cost = 0
+        
+        # Handle Gemini API key fallback mechanism
+        if "gemini" in model_name.lower():
+            self._setup_gemini_api_key()
 
         if self.verbose:
             os.environ['LITELLM_LOG'] = 'DEBUG'
@@ -46,6 +51,30 @@ class LiteLLMWrapper:
         if use_langfuse:
             litellm.success_callback = ["langfuse"]
             litellm.failure_callback = ["langfuse"]
+    
+    def _setup_gemini_api_key(self):
+        """Setup Gemini API key with fallback mechanism for multiple keys."""
+        from dotenv import load_dotenv
+        load_dotenv(override=True)
+        
+        gemini_key_env = os.getenv("GEMINI_API_KEY") or os.getenv("GOOGLE_API_KEY")
+        if not gemini_key_env:
+            raise ValueError("No API_KEY found. Please set the `GEMINI_API_KEY` or `GOOGLE_API_KEY` environment variable.")
+        
+        # Support comma-separated list of API keys with random selection
+        if ',' in gemini_key_env:
+            keys = [key.strip() for key in gemini_key_env.split(',') if key.strip()]
+            if not keys:
+                raise ValueError("No valid API keys found in GEMINI_API_KEY list.")
+            api_key = random.choice(keys)
+            print(f"Selected random Gemini API key from {len(keys)} available keys: {api_key[:20]}...")
+        else:
+            api_key = gemini_key_env
+            print(f"Using single Gemini API key: {api_key[:20]}...")
+        
+        # Set the selected API key for LiteLLM
+        os.environ["GEMINI_API_KEY"] = api_key
+        os.environ["GOOGLE_API_KEY"] = api_key
 
     def _encode_file(self, file_path: Union[str, Image.Image]) -> str:
         """

src/utils/allowed_models.json

@@ -3,6 +3,7 @@
     "gemini/gemini-1.5-pro-002",
     "gemini/gemini-1.5-flash-002",
     "gemini/gemini-2.0-flash-001",
+    "gemini/gemini-2.0-flash",
     "vertex_ai/gemini-1.5-flash-002",
     "vertex_ai/gemini-1.5-pro-002",
     "vertex_ai/gemini-2.0-flash-001",

src/utils/kokoro_voiceover.py

@@ -2,116 +2,28 @@
 Copyright (c) 2025 Xposed73
 All rights reserved.
 This file is part of the Manim Voiceover project.
-"""
-
-import hashlib
-import json
-import numpy as np
-from pathlib import Path
-from manim_voiceover.services.base import SpeechService
-from kokoro_onnx import Kokoro
-from manim_voiceover.helper import remove_bookmarks, wav2mp3
-from scipy.io.wavfile import write as write_wav
-from src.config.config import Config
-
-
-class KokoroService(SpeechService):
-    """Speech service class for kokoro_self (using text_to_speech via Kokoro ONNX)."""
-
-    def __init__(self, engine=None, 
-                 model_path: str = Config.KOKORO_MODEL_PATH,
-                 voices_path: str = Config.KOKORO_VOICES_PATH,
-                 voice: str = Config.KOKORO_DEFAULT_VOICE,
-                 speed: float = Config.KOKORO_DEFAULT_SPEED,
-                 lang: str = Config.KOKORO_DEFAULT_LANG,
-                 **kwargs):
-        self.kokoro = Kokoro(model_path, voices_path)
-        self.voice = voice
-        self.speed = speed
-        self.lang = lang
-
-        if engine is None:
-            engine = self.text_to_speech  # Default to local function
-
-        self.engine = engine
-        super().__init__(**kwargs)
-
-    def get_data_hash(self, input_data: dict) -> str:
-        """
-        Generates a hash based on the input data dictionary.
-        The hash is used to create a unique identifier for the input data.
-
-        Parameters:
-            input_data (dict): A dictionary of input data (e.g., text, voice, etc.).
-
-        Returns:
-            str: The generated hash as a string.
-        """
-        # Convert the input data dictionary to a JSON string (sorted for consistency)
-        data_str = json.dumps(input_data, sort_keys=True)
-        # Generate a SHA-256 hash of the JSON string
-        return hashlib.sha256(data_str.encode('utf-8')).hexdigest()
-
-    def text_to_speech(self, text, output_file, voice_name, speed, lang):
-        """
-        Generates speech from text using Kokoro ONNX and saves the audio file.
-        Normalizes the audio to make it audible.
-        """
-        # Generate audio samples using Kokoro
-        samples, sample_rate = self.kokoro.create(
-            text, voice=voice_name, speed=speed, lang=lang
-        )
 
-        # Normalize audio to the range [-1, 1]
-        max_val = np.max(np.abs(samples))
-        if max_val > 0:
-            samples = samples / max_val
-
-        # Convert to 16-bit integer PCM format
-        samples = (samples * 32767).astype("int16")
-
-        # Save the normalized audio as a .wav file
-        write_wav(output_file, sample_rate, samples)
-        print(f"Saved at {output_file}")
-
-        return output_file
-
-
-    def generate_from_text(self, text: str, cache_dir: str = None, path: str = None) -> dict:
-        if cache_dir is None:
-            cache_dir = self.cache_dir
-
-        input_data = {"input_text": text, "service": "kokoro_self", "voice": self.voice, "lang": self.lang}
-        cached_result = self.get_cached_result(input_data, cache_dir)
-        if cached_result is not None:
-            return cached_result
-
-        if path is None:
-            audio_path = self.get_data_hash(input_data) + ".mp3"
-        else:
-            audio_path = path
+DEPRECATED: This file is deprecated. Use ElevenLabsService from src.utils.elevenlabs_voiceover instead.
+This wrapper is maintained for backward compatibility only.
+"""
 
-        # Generate .wav file using the text_to_speech function
-        audio_path_wav = str(Path(cache_dir) / audio_path.replace(".mp3", ".wav"))
-        self.engine(
-            text=text,
-            output_file=audio_path_wav,
-            voice_name=self.voice,
-            speed=self.speed,
-            lang=self.lang,
+import warnings
+from src.utils.elevenlabs_voiceover import ElevenLabsService
+
+class KokoroService(ElevenLabsService):
+    """
+    DEPRECATED: Backward compatibility wrapper for ElevenLabsService.
+    
+    This class now redirects to ElevenLabsService. Please update your code to use:
+    from src.utils.elevenlabs_voiceover import ElevenLabsService
+    """
+    
+    def __init__(self, **kwargs):
+        warnings.warn(
+            "KokoroService is deprecated. Please use ElevenLabsService from "
+            "src.utils.elevenlabs_voiceover instead.",
+            DeprecationWarning,
+            stacklevel=2
         )
-
-        # Convert .wav to .mp3
-        mp3_audio_path = str(Path(cache_dir) / audio_path)
-        wav2mp3(audio_path_wav, mp3_audio_path)
-
-        # Remove original .wav file
-        remove_bookmarks(audio_path_wav)
-
-        json_dict = {
-            "input_text": text,
-            "input_data": input_data,
-            "original_audio": audio_path,
-        }
-
-        return json_dict
+        # Pass all arguments to ElevenLabsService
+        super().__init__(**kwargs)

task_generator/prompts_raw/__init__.py

@@ -1692,7 +1692,7 @@ Scene Technical Implementation:
 
 1.  **Scene Class:** Class name `Scene{scene_number}`, where `{scene_number}` is replaced by the scene number (e.g., `Scene1`, `Scene2`). The scene class should at least inherit from `VoiceoverScene`. However, you can add more Manim Scene classes on top of VoiceoverScene for multiple inheritance if needed.
 2.  **Imports:** Include ALL necessary imports explicitly at the top of the file, based on used Manim classes, functions, colors, and constants. Do not rely on implicit imports. Double-check for required modules, classes, functions, colors, and constants, *ensuring all imports are valid and consistent with the Manim Documentation*.  **Include imports for any used Manim plugins.**
-3.  **Speech Service:** Initialize `KokoroService()`. You MUST import like this: `from src.utils.kokoro_voiceover import KokoroService` as this is our custom voiceover service.
+3.  **Speech Service:** Initialize `ElevenLabsService()`. You MUST import like this: `from src.utils.elevenlabs_voiceover import ElevenLabsService` as this is our custom voiceover service.
 4.  **Reusable Animations:** Implement functions for each animation sequence to create modular and reusable code. Structure code into well-defined functions, following function definition patterns from Manim Documentation.
 5.  **Voiceover:** Use `with self.voiceover(text="...")` for speech synchronization, precisely matching the narration script and animation timings from the Animation and Narration Plan.
 6.  **Comments:** Add clear and concise comments for complex animations, spatial logic (positioning, arrangements), and object lifecycle management. *Use comments extensively to explain code logic, especially for spatial positioning, animation sequences, and constraint enforcement, mirroring commenting style in Manim Documentation*.  **Add comments to explain the purpose and usage of any Manim plugins.**
@@ -1726,7 +1726,7 @@ Scene Technical Implementation:
 *   **Reusable Object Creation Functions:** Define reusable functions within helper classes for creating specific Manim objects (e.g., `create_axes`, `create_formula_tex`, `create_explanation_text`).
 *   **Clear Comments and Variable Names:** Use clear, concise comments to explain code sections and logic. Employ descriptive variable names (e.g., `linear_function_formula`, `logistic_plot`) for better readability.
 *   **Text Elements:** Create text elements using `Tex` or `MathTex` for formulas and explanations, styling them with `color` and `font_size` as needed.
-*   **Manim Best Practices:** Follow Manim best practices, including using `VoiceoverScene`, `KokoroService`, common Manim objects, animations, relative positioning, and predefined colors.
+*   **Manim Best Practices:** Follow Manim best practices, including using `VoiceoverScene`, `ElevenLabsService`, common Manim objects, animations, relative positioning, and predefined colors.
 
 You MUST generate the Python code in the following format (from <CODE> to </CODE>):
 <CODE>
@@ -1734,7 +1734,7 @@ You MUST generate the Python code in the following format (from <CODE> to </CODE
 from manim import *
 from manim import config as global_config
 from manim_voiceover import VoiceoverScene
-from src.utils.kokoro_voiceover import KokoroService # You MUST import like this as this is our custom voiceover service.
+from src.utils.elevenlabs_voiceover import ElevenLabsService # You MUST import like this as this is our custom voiceover service.
 
 # plugins imports, don't change the import statements
 from manim_circuit import *
@@ -1790,7 +1790,7 @@ class Scene{scene_number}(VoiceoverScene, MovingCameraScene):  # Note: You can a
     # Reminder: This scene class is fully self-contained. There is no dependency on the implementation from previous or subsequent scenes.
     def construct(self):
         # Initialize speech service
-        self.set_speech_service(KokoroService())
+        self.set_speech_service(ElevenLabsService())
 
         # Instantiate helper class (as per plan)
         helper = Scene{scene_number}_Helper(self)  # Example: helper = Scene1_Helper(self)