Skip to content

LANGCHAIN_RAG_RESOURCE_REFERENCE.md

Latest commit

 

History

History
108 lines (88 loc) · 2.03 KB

LANGCHAIN_RAG_RESOURCE_REFERENCE.md

File metadata and controls

108 lines (88 loc) · 2.03 KB
Raw

LangChain RAG Resource Usage - Quick Reference

🎯 At a Glance

Metric Value Notes
Typical RAM 1.5-2GB 50K documents
Query CPU 15-30% Single query
Query Time (uncached) 250ms Average
Query Time (cached) 5ms 65-80% hit rate
Index Time 15-20 min 50K documents

💾 Memory Components

Total RAM: ~1.8GB
├─ ChromaDB:          800MB  (44%)
├─ Embedding Model:   300MB  (17%)
├─ Query Cache:       150MB  (8%)
├─ TF-IDF:           100MB  (6%)
├─ BM25 (dynamic):    150MB  (8%)
└─ Application:       300MB  (17%)

⚡ Performance Profiles

Low Memory Mode (< 2GB RAM)

dense_top_k: 30
sparse_top_k: 30
max_bm25_docs: 10
cache_max_memory_entries: 500

Result: ~1.2-1.5GB RAM, +50ms query time

Balanced Mode (2-4GB RAM) ⭐ Recommended

dense_top_k: 50
sparse_top_k: 50
max_bm25_docs: 15
cache_max_memory_entries: 1000

Result: ~1.8-2.2GB RAM, 250ms query time

High Performance Mode (4GB+ RAM)

dense_top_k: 100
sparse_top_k: 100
max_bm25_docs: 50
cache_max_memory_entries: 2000

Result: ~2.5-3GB RAM, 180ms query time

📊 Scaling Guidelines

Documents Min RAM Recommended RAM Query Time Index Time
10K 1GB 2GB 180ms 3 min
50K 1.5GB 4GB 250ms 15 min
100K 2GB 8GB 350ms 35 min
500K+ 4GB 16GB+ 600ms 3+ hours

🐳 Docker Resource Limits

Small Deployment

resources:
  limits:
    memory: "2Gi"
    cpu: "1"
  reservations:
    memory: "1Gi"
    cpu: "0.5"

Medium Deployment

resources:
  limits:
    memory: "4Gi"
    cpu: "2"
  reservations:
    memory: "2Gi"
    cpu: "1"

Large Deployment

resources:
  limits:
    memory: "8Gi"
    cpu: "4"
  reservations:
    memory: "4Gi"
    cpu: "2"

Last Updated: October 27, 2024
Quick Access: Keep this handy for production deployments!