/**
* Performance optimization constants for the MCP server
* These values are tuned for fast indexing and efficient resource usage
*/
export const PERFORMANCE_CONFIG = {
// Embedding processing
EMBEDDING: {
// Concurrent requests for different providers
IBTHINK_CONCURRENT_REQUESTS: 3,
OPENAI_CONCURRENT_REQUESTS: 5,
VOYAGEAI_CONCURRENT_REQUESTS: 4,
GEMINI_CONCURRENT_REQUESTS: 2,
OLLAMA_CONCURRENT_REQUESTS: 1, // Local processing, avoid overwhelming
// Batch sizes for different providers
IBTHINK_CHUNK_SIZE: 50,
OPENAI_CHUNK_SIZE: 100,
VOYAGEAI_CHUNK_SIZE: 75,
GEMINI_CHUNK_SIZE: 25,
OLLAMA_CHUNK_SIZE: 10,
// Rate limiting delays (milliseconds)
IBTHINK_BATCH_DELAY: 100,
OPENAI_BATCH_DELAY: 200,
VOYAGEAI_BATCH_DELAY: 150,
GEMINI_BATCH_DELAY: 300,
OLLAMA_BATCH_DELAY: 0, // No delay for local
},
// Snapshot management
SNAPSHOT: {
SAVE_INTERVAL_MS: 15000, // 15 seconds between saves
THROTTLE_INTERVAL_MS: 1000, // 1 second min between saves
BATCH_UPDATE_SIZE: 100, // Batch multiple updates
},
// File operations
FILE_OPS: {
MAX_CONCURRENT_READS: 50,
READ_CHUNK_SIZE: 8192, // 8KB chunks
DIRECTORY_SCAN_BATCH_SIZE: 1000,
FILE_STAT_BATCH_SIZE: 100,
},
// Milvus database
MILVUS: {
// Optimized for fast indexing
SEGMENT_SIZE_MB: 1024,
SEGMENT_SEAL_PROPORTION: 0.25,
INSERT_BUFFER_SIZE_MB: 512,
INDEX_BUILD_WORKERS: 4,
MAX_SEGMENT_NUM: 16,
// Cache sizes
QUERY_NODE_CACHE_SIZE_MB: 4096,
INDEX_ENGINE_CACHE_SIZE_MB: 2048,
// Performance tuning
SEARCH_ENGINE_NUM_THREADS: 8,
BUILD_INDEX_THREADS: 4,
// Connection pooling
MAX_CONNECTIONS: 10,
CONNECTION_TIMEOUT_MS: 30000,
},
// Indexing process
INDEXING: {
MAX_CONCURRENT_CODEBASES: 2,
CHUNK_PROCESSING_BATCH_SIZE: 500,
PROGRESS_UPDATE_INTERVAL_MS: 5000, // 5 seconds
// Memory management
MAX_MEMORY_USAGE_MB: 4096, // 4GB max
GC_INTERVAL_MS: 60000, // Force GC every minute
// Error handling
MAX_RETRIES: 3,
RETRY_DELAY_MS: 1000,
EXPONENTIAL_BACKOFF: true,
},
// API rate limits
RATE_LIMITS: {
REQUESTS_PER_MINUTE: {
IBTHINK: 100,
OPENAI: 3000,
VOYAGEAI: 300,
GEMINI: 60,
OLLAMA: 1000, // Local, higher limit
},
TOKENS_PER_MINUTE: {
IBTHINK: 100000,
OPENAI: 1000000,
VOYAGEAI: 300000,
GEMINI: 50000,
OLLAMA: 500000,
}
},
// Logging
LOGGING: {
// Reduce logging in performance-critical paths
ENABLE_PROGRESS_LOGS: true,
ENABLE_DEBUG_LOGS: false,
LOG_BATCH_SIZE: 50,
LOG_FLUSH_INTERVAL_MS: 10000,
}
} as const;
/**
* Environment-specific overrides
*/
export const getPerformanceConfig = (env: 'development' | 'production' | 'test' = 'production') => {
// Create a deep copy and make it mutable
const config = JSON.parse(JSON.stringify(PERFORMANCE_CONFIG));
switch (env) {
case 'development':
// More conservative settings for development
config.EMBEDDING.IBTHINK_CONCURRENT_REQUESTS = 2;
config.EMBEDDING.OPENAI_CONCURRENT_REQUESTS = 3;
config.MILVUS.MAX_CONNECTIONS = 5;
config.INDEXING.MAX_CONCURRENT_CODEBASES = 1;
config.LOGGING.ENABLE_DEBUG_LOGS = true;
break;
case 'test':
// Minimal settings for testing
config.EMBEDDING.IBTHINK_CONCURRENT_REQUESTS = 1;
config.EMBEDDING.OPENAI_CONCURRENT_REQUESTS = 1;
config.SNAPSHOT.SAVE_INTERVAL_MS = 5000;
config.MILVUS.MAX_CONNECTIONS = 2;
config.LOGGING.ENABLE_PROGRESS_LOGS = false;
break;
case 'production':
default:
// Use default optimized settings
break;
}
return config;
};
/**
* Helper function to get provider-specific embedding config
*/
export const getEmbeddingConfig = (provider: string) => {
const config = PERFORMANCE_CONFIG.EMBEDDING;
const providerUpper = provider.toUpperCase();
return {
concurrentRequests: config[`${providerUpper}_CONCURRENT_REQUESTS` as keyof typeof config] || 2,
chunkSize: config[`${providerUpper}_CHUNK_SIZE` as keyof typeof config] || 50,
batchDelay: config[`${providerUpper}_BATCH_DELAY` as keyof typeof config] || 200,
};
};
