Skip to main content

Overview

Olis uses sophisticated intent detection to understand user queries and provide relevant responses. The system supports multiple backends for flexibility and performance.

Detection Modes

Automatic

Intelligently selects the best available option based on system capabilities

SLM (Local)

Local small language model for privacy-focused, offline operation

LLM (Remote)

Cloud-hosted language model for faster, more powerful processing

NLP (Fallback)

Fast heuristic-based detection using pattern matching

Configuration

Environment Variables

# Model Selection
OLIS_INTENT_MODEL=phi3

# Remote URL (for LLM mode)
OLIS_INTENT_REMOTE_URL=http://localhost:11434

# Model Download (for SLM mode)
OLIS_INTENT_MODEL_DOWNLOAD_URL=https://example.com/model.gguf

# Timeout
OLIS_INTENT_TIMEOUT_MS=30000

How It Works

Best Practices

  • Use SLM for privacy-sensitive data
  • Use LLM for complex queries
  • Configure timeout appropriately
  • Monitor performance metrics

Next Steps