pdf instructions update check
This commit is contained in:
Manish Tanwar
parent
dc770d65d3
commit
bec26a02be
5 changed files with 70 additions and 103 deletions
32
CLAUDE.md
32
CLAUDE.md
|
|
@ -17,6 +17,12 @@ A full-stack video query application using Google Gemini 2.0 Flash Exp AI for vi
|
|||
cd backend
|
||||
python3 -m venv venv
|
||||
source venv/bin/activate # On Windows: venv\Scripts\activate
|
||||
|
||||
# For Python 3.10 (use special requirements file)
|
||||
pip install -r requirements-py310.txt
|
||||
bash fix_jose.sh # Fix jose module conflict if needed
|
||||
|
||||
# For other Python versions
|
||||
pip install -r requirements.txt
|
||||
|
||||
# Install system dependencies (Ubuntu/Debian)
|
||||
|
|
@ -75,8 +81,8 @@ npm run build
|
|||
|
||||
### Frontend
|
||||
- **Development server**: `npm start` (port 3000)
|
||||
- **Production build**: `npm run build`
|
||||
- **Build with deploy script**: `./build.sh`
|
||||
- **Production build**: `./build.sh` (recommended - sets PUBLIC_URL automatically)
|
||||
- **Manual production build**: `PUBLIC_URL=/video_query npm run build` (if not using script)
|
||||
|
||||
### Video Processing
|
||||
- **Standalone script**: `python video_query.py <video_path> [--prompt "Your custom prompt"]`
|
||||
|
|
@ -97,6 +103,12 @@ npm run build
|
|||
cd /path/to/video-query/backend
|
||||
python3 -m venv venv
|
||||
source venv/bin/activate
|
||||
|
||||
# For Python 3.10
|
||||
pip install -r requirements-py310.txt
|
||||
bash fix_jose.sh
|
||||
|
||||
# For other Python versions
|
||||
pip install -r requirements.txt
|
||||
```
|
||||
|
||||
|
|
@ -150,7 +162,11 @@ EOF
|
|||
2. **Build for production**:
|
||||
```bash
|
||||
cd frontend
|
||||
npm run build
|
||||
# Option 1: Use build script (recommended - sets PUBLIC_URL automatically)
|
||||
./build.sh
|
||||
|
||||
# Option 2: Manual build with PUBLIC_URL
|
||||
PUBLIC_URL=/video_query npm run build
|
||||
```
|
||||
|
||||
3. **Deploy to web server**:
|
||||
|
|
@ -378,6 +394,11 @@ sudo systemctl restart video-query
|
|||
# Check ffmpeg installation
|
||||
which ffprobe
|
||||
ffprobe -version
|
||||
|
||||
# Python 3.10: Fix jose module SyntaxError
|
||||
# If you see: "SyntaxError: Missing parentheses in call to 'print'"
|
||||
cd backend
|
||||
bash fix_jose.sh
|
||||
```
|
||||
|
||||
### Frontend Issues
|
||||
|
|
@ -396,6 +417,11 @@ tail -f /var/log/apache2/error.log
|
|||
tail -f /var/log/nginx/error.log
|
||||
```
|
||||
|
||||
**Static Assets Not Loading** (JS/CSS 404 errors):
|
||||
- **Symptom**: `Loading failed for <script> with source "https://domain.com/static/js/main.xxx.js"`
|
||||
- **Cause**: Application built without correct PUBLIC_URL
|
||||
- **Solution**: Rebuild with `./build.sh` or `PUBLIC_URL=/video_query npm run build`
|
||||
|
||||
### Rate Limiting Issues
|
||||
- **Symptom**: 400 INVALID_ARGUMENT after 3-4 videos
|
||||
- **Check**: Gemini API rate limits (5 RPM for free tier)
|
||||
|
|
|
|||
14
backend/.env
14
backend/.env
|
|
@ -1,18 +1,18 @@
|
|||
GOOGLE_API_KEY=AIzaSyBF3Ia1nVS4PLuLpWt-85ct_heJ7FrlvkQ
|
||||
|
||||
# API Tier Configuration (IMPORTANT!)
|
||||
# Set to "free" for free tier (5 RPM) or "paid" for paid tier (60 RPM)
|
||||
# This prevents 503 errors by enforcing proper rate limits
|
||||
GEMINI_API_TIER=free
|
||||
# Video Splitting Configuration
|
||||
# Maximum duration per chunk in minutes (default: 53)
|
||||
# Videos longer than this will be split into multiple chunks
|
||||
CHUNK_DURATION_MINUTES=53
|
||||
|
||||
# Parallel Processing (auto-configured based on tier, uncomment to override)
|
||||
# MAX_PARALLEL_CHUNKS=2
|
||||
# Parallel Processing Configuration
|
||||
# Maximum number of chunks to process concurrently (default: 4)
|
||||
MAX_PARALLEL_CHUNKS=4
|
||||
|
||||
# Model Configuration
|
||||
VIDEO_PROCESSOR_MODEL=gemini-2.5-pro
|
||||
VIDEO_SYNTHESIS_MODEL=gemini-2.5-pro
|
||||
|
||||
|
||||
# Enable logging of prompts sent to AI for each video/chunk
|
||||
# Shows exactly what prompt was used for each video in batch
|
||||
BATCH_PROCESSING_LOG_PROMPTS=false
|
||||
|
|
|
|||
|
|
@ -1,26 +1,21 @@
|
|||
# Google Gemini API Key (REQUIRED)
|
||||
GOOGLE_API_KEY=your_api_key_here
|
||||
GOOGLE_API_KEY=AIzaSyBF3Ia1nVS4PLuLpWt-85ct_heJ7FrlvkQ
|
||||
|
||||
# =============================================================================
|
||||
# API TIER AND RATE LIMITING CONFIGURATION (IMPORTANT!)
|
||||
# VIDEO PROCESSING CONFIGURATION
|
||||
# =============================================================================
|
||||
# Set this based on your Gemini API subscription level
|
||||
# This prevents 503 UNAVAILABLE errors by enforcing proper rate limits
|
||||
|
||||
# API Tier: "free" or "paid"
|
||||
# - free: 5 requests per minute (RPM), 12 seconds between requests, max 2 parallel chunks
|
||||
# - paid: 60 requests per minute (RPM), 1 second between requests, max 4-10 parallel chunks
|
||||
# Default: free (conservative to prevent 503 overload errors)
|
||||
GEMINI_API_TIER=free
|
||||
# Video Splitting Configuration
|
||||
# Maximum duration per chunk in minutes (default: 53)
|
||||
# Videos longer than this will be split into multiple chunks for processing
|
||||
# Gemini API limits: ~55 min with audio, ~60 min without audio
|
||||
CHUNK_DURATION_MINUTES=53
|
||||
|
||||
# Parallel Processing Configuration
|
||||
# Maximum number of video chunks to process simultaneously
|
||||
# Recommended values:
|
||||
# - Free tier: 1-2 (safe, prevents overload)
|
||||
# - Paid tier: 4-10 (faster processing)
|
||||
# Default: Auto-configured based on GEMINI_API_TIER (2 for free, 4 for paid)
|
||||
# Uncomment to override:
|
||||
# MAX_PARALLEL_CHUNKS=2
|
||||
# Maximum number of video chunks to process concurrently (default: 4)
|
||||
# Higher values = faster processing but more API load
|
||||
# Recommended: 4-10 for paid tier with high rate limits
|
||||
MAX_PARALLEL_CHUNKS=4
|
||||
|
||||
# =============================================================================
|
||||
# MODEL CONFIGURATION (Optional)
|
||||
|
|
@ -30,7 +25,9 @@ GEMINI_API_TIER=free
|
|||
VIDEO_PROCESSOR_MODEL=gemini-2.5-pro
|
||||
VIDEO_SYNTHESIS_MODEL=gemini-2.5-pro
|
||||
|
||||
# Batch Processing Logging (Optional)
|
||||
# =============================================================================
|
||||
# BATCH PROCESSING LOGGING (Optional)
|
||||
# =============================================================================
|
||||
# Enable detailed logging for batch processing operations
|
||||
# Useful for debugging and understanding how videos are processed
|
||||
# Default: false (to reduce log volume)
|
||||
|
|
|
|||
|
|
@ -43,18 +43,14 @@ class VideoProcessor:
|
|||
|
||||
# Parallel processing configuration
|
||||
# Default max workers for parallel chunk processing
|
||||
# Free tier: 5 RPM (use 1-2 workers to be safe)
|
||||
# Paid tier: 60 RPM (can use 4-10 workers)
|
||||
DEFAULT_MAX_WORKERS = 2 # Conservative default for free tier (reduced from 4 to prevent 503 errors)
|
||||
DEFAULT_MAX_WORKERS = 4 # Default concurrent workers
|
||||
|
||||
# Model configuration
|
||||
DEFAULT_PROCESSING_MODEL = "gemini-2.5-pro" # Model for individual video processing
|
||||
DEFAULT_SYNTHESIS_MODEL = "gemini-2.5-pro" # Model for batch synthesis (updated for consistency)
|
||||
DEFAULT_SYNTHESIS_MODEL = "gemini-2.5-pro" # Model for batch synthesis
|
||||
|
||||
# Rate limiting and retry configuration
|
||||
MIN_REQUEST_INTERVAL_FREE = 12 # seconds (for 5 RPM free tier: 60/5 = 12s)
|
||||
MIN_REQUEST_INTERVAL_PAID = 1 # seconds (for 60 RPM paid tier: 60/60 = 1s)
|
||||
MAX_RETRY_ATTEMPTS = 5 # Maximum retry attempts (increased from 3)
|
||||
# Retry configuration
|
||||
MAX_RETRY_ATTEMPTS = 5 # Maximum retry attempts
|
||||
RETRY_DELAYS = [5, 10, 20, 40, 60] # Exponential backoff delays in seconds
|
||||
|
||||
def __init__(self, api_key: Optional[str] = None, max_parallel_chunks: int = None):
|
||||
|
|
@ -64,7 +60,7 @@ class VideoProcessor:
|
|||
Args:
|
||||
api_key: Google API key for Gemini
|
||||
max_parallel_chunks: Maximum number of chunks to process in parallel
|
||||
(default: 4, recommended 3-4 for free tier, up to 10+ for paid tier)
|
||||
(default: from MAX_PARALLEL_CHUNKS env var or 4)
|
||||
"""
|
||||
self.api_key = api_key or os.getenv("GOOGLE_API_KEY")
|
||||
if not self.api_key:
|
||||
|
|
@ -76,34 +72,22 @@ class VideoProcessor:
|
|||
self.client = genai.Client(api_key=self.api_key)
|
||||
logger.info("Gemini API client initialized successfully")
|
||||
|
||||
# Detect API tier (free or paid)
|
||||
self._api_tier = self._detect_api_tier()
|
||||
|
||||
# Set parallel processing configuration based on API tier
|
||||
# Set parallel processing configuration
|
||||
if max_parallel_chunks:
|
||||
self.max_parallel_chunks = max_parallel_chunks
|
||||
else:
|
||||
# Auto-configure based on API tier
|
||||
# Load from environment variable or use default
|
||||
env_max_workers = os.getenv("MAX_PARALLEL_CHUNKS")
|
||||
if env_max_workers:
|
||||
self.max_parallel_chunks = int(env_max_workers)
|
||||
else:
|
||||
# Default based on tier
|
||||
if self._api_tier == "paid":
|
||||
self.max_parallel_chunks = 4
|
||||
else:
|
||||
self.max_parallel_chunks = 2
|
||||
self.max_parallel_chunks = self.DEFAULT_MAX_WORKERS
|
||||
|
||||
logger.info(f"Parallel processing: max {self.max_parallel_chunks} concurrent chunks ({self._api_tier} tier)")
|
||||
logger.info(f"Parallel processing: max {self.max_parallel_chunks} concurrent chunks")
|
||||
|
||||
# Initialize video splitter
|
||||
self.video_splitter = VideoSplitter()
|
||||
|
||||
# Thread lock and tracking for rate limiting
|
||||
self._rate_limit_lock = threading.Lock()
|
||||
self._last_request_time = 0
|
||||
self._request_count = 0
|
||||
|
||||
# Load configuration from environment variables
|
||||
self.processing_model = os.getenv("VIDEO_PROCESSOR_MODEL", self.DEFAULT_PROCESSING_MODEL)
|
||||
self.synthesis_model = os.getenv("VIDEO_SYNTHESIS_MODEL", self.DEFAULT_SYNTHESIS_MODEL)
|
||||
|
|
@ -155,48 +139,6 @@ class VideoProcessor:
|
|||
logger.error(f"Error sending usage data to webhook: {str(e)}")
|
||||
# Don't raise the exception - webhook failure shouldn't block the main flow
|
||||
|
||||
def _detect_api_tier(self) -> str:
|
||||
"""
|
||||
Detect if using free or paid API tier.
|
||||
Can be overridden with env var: GEMINI_API_TIER=free or GEMINI_API_TIER=paid
|
||||
|
||||
Returns:
|
||||
"free" or "paid"
|
||||
"""
|
||||
tier = os.getenv("GEMINI_API_TIER", "free").lower()
|
||||
if tier in ["free", "paid"]:
|
||||
logger.info(f"Using {tier} tier API configuration")
|
||||
return tier
|
||||
logger.warning(f"Unknown API tier '{tier}', defaulting to 'free' for safety")
|
||||
return "free"
|
||||
|
||||
def _wait_for_rate_limit(self) -> None:
|
||||
"""
|
||||
Smart rate limiting that respects API tier limits.
|
||||
Free tier: 5 RPM = 12 seconds between requests
|
||||
Paid tier: 60 RPM = 1 second between requests
|
||||
|
||||
This method ensures we don't overwhelm the API with parallel requests.
|
||||
"""
|
||||
with self._rate_limit_lock:
|
||||
current_time = time.time()
|
||||
time_since_last = current_time - self._last_request_time
|
||||
|
||||
# Determine minimum interval based on API tier
|
||||
if self._api_tier == "paid":
|
||||
min_interval = self.MIN_REQUEST_INTERVAL_PAID
|
||||
else:
|
||||
min_interval = self.MIN_REQUEST_INTERVAL_FREE
|
||||
|
||||
if time_since_last < min_interval:
|
||||
wait_time = min_interval - time_since_last
|
||||
logger.info(f"Rate limiting: waiting {wait_time:.1f}s before next API call")
|
||||
time.sleep(wait_time)
|
||||
|
||||
self._last_request_time = time.time()
|
||||
self._request_count += 1
|
||||
logger.debug(f"API request #{self._request_count} at {self._last_request_time:.2f}")
|
||||
|
||||
def _extract_error_code(self, error_message: str) -> str:
|
||||
"""
|
||||
Extract HTTP error code from error message.
|
||||
|
|
@ -285,9 +227,6 @@ class VideoProcessor:
|
|||
|
||||
for attempt in range(self.MAX_RETRY_ATTEMPTS):
|
||||
try:
|
||||
# Apply rate limiting before each attempt
|
||||
self._wait_for_rate_limit()
|
||||
|
||||
# Make the API call
|
||||
if attempt == 0:
|
||||
logger.info(f"{context} Sending request to Gemini API")
|
||||
|
|
|
|||
|
|
@ -21,17 +21,22 @@ class VideoSplitter:
|
|||
Handles video duration detection and splitting operations.
|
||||
"""
|
||||
|
||||
# Default chunk duration in minutes (43 min to stay under 45 min Gemini API limit for videos with audio)
|
||||
# Google Gemini 2.5 Pro limits: ~45 min with audio, ~60 min without audio
|
||||
DEFAULT_CHUNK_DURATION = 43
|
||||
# Default chunk duration in minutes (53 min to stay under 55 min Gemini API limit for videos with audio)
|
||||
# Google Gemini 2.5 Pro limits: ~55 min with audio, ~60 min without audio
|
||||
DEFAULT_CHUNK_DURATION = 53
|
||||
|
||||
def __init__(self, chunk_duration_minutes: int = DEFAULT_CHUNK_DURATION):
|
||||
def __init__(self, chunk_duration_minutes: int = None):
|
||||
"""
|
||||
Initialize VideoSplitter with specified chunk duration.
|
||||
|
||||
Args:
|
||||
chunk_duration_minutes: Duration of each chunk in minutes (default: 54)
|
||||
chunk_duration_minutes: Duration of each chunk in minutes
|
||||
(default: from CHUNK_DURATION_MINUTES env var or 53)
|
||||
"""
|
||||
if chunk_duration_minutes is None:
|
||||
# Load from environment variable or use default
|
||||
chunk_duration_minutes = int(os.getenv("CHUNK_DURATION_MINUTES", self.DEFAULT_CHUNK_DURATION))
|
||||
|
||||
self.chunk_duration_minutes = chunk_duration_minutes
|
||||
self.chunk_duration_seconds = chunk_duration_minutes * 60
|
||||
logger.info(f"VideoSplitter initialized with chunk duration: {chunk_duration_minutes} minutes")
|
||||
|
|
@ -99,7 +104,7 @@ class VideoSplitter:
|
|||
Check if a video needs to be split based on duration OR file size.
|
||||
|
||||
A video needs splitting if:
|
||||
1. Duration > 54 minutes (Gemini API time limit), OR
|
||||
1. Duration > configured chunk duration (default 53 minutes), OR
|
||||
2. File size > 500MB (conservative target to handle variable bitrate)
|
||||
With 30% variance: 500MB × 1.3 = 650MB max
|
||||
After base64 encoding: 650MB × 1.37 = 891MB (well under 1GB API limit)
|
||||
|
|
|
|||
Loading…
Add table
Reference in a new issue