config.example.yaml

# Example configuration for AI CLI
# This shows different configuration options you can use

# NOTE: Models are now managed dynamically via the ModelRegistry (stored in Redis)
# Use CLI commands to manage models:
#   /model general add <url> <model_name>  - Add a general chat model
#   /model coder add <url> <model_name>    - Add a code-specialized model
#   /model list                            - List all registered models
#   /model status                          - Check model availability
#   /model remove <model_id>               - Remove a model
#
# Or use the Web UI at /models to manage models visually

# Local tinyollama configuration (lightweight fallback model)
# This is the only model that remains in config - it's used as a fallback
# when the primary models are unavailable
tinyollama:
  url: "http://localhost:11434"
  model: "tinyllama"
  timeout: 60
  # Features disabled for tinyollama (lightweight model limitations)
  disabled_features:
    - code_mode        # /code command - requires larger models for code generation
    - coder_model      # Code editing - requires specialized coder models
    - repomap_create   # Repository mapping - requires large context understanding
    - repomap_update   # Repository map update - requires large context understanding
    - datamap_create   # Data mapping - requires large context understanding
    - datamap_update   # Data map update - requires large context understanding

# Chat configuration
chat:
  system_prompt: "You are a helpful AI assistant."
  max_context_length: 10
  temperature: 0.7
  stream: true

# Tabular GMD (Gaussian Multinomial Diffusion) configuration
# Service for generating synthetic tabular data using diffusion models
# The generate_fake_data_with_ddpm tool will try this endpoint first,
# then fall back to local numpy implementation if unavailable
tabular_gmd:
  url: "http://192.168.31.23:15432"  # Change to your tabular-gmd service URL
  timeout: 300  # 5 minutes - synthetic data generation can take time

---

# Migration note:
# If you're upgrading from an older version with ollama.model and ollama.coder_model
# in config.yaml, those models will be automatically migrated to the ModelRegistry
# on first startup. After migration, you can remove those fields from your config.

---

# Example: Adding models via CLI

# Add a general chat model:
# /model general add http://192.168.1.100:11434 llama3.1:8b

# Add a coder model:
# /model coder add http://192.168.1.100:11434 qwen2.5-coder:7b

# Check model status:
# /model status

# List all models:
# /model list

---

# Available Ollama models (examples):
# - tinyllama: Lightweight model (~1GB, CPU-friendly) - Fallback model
# - llama2/llama3: General-purpose models
# - mistral: Fast and efficient model
# - codellama: Specialized for code
# - qwen2.5-coder: Advanced code generation model
# - phi: Smaller, faster model
# - neural-chat: Conversational model
#
# Pull models:
# Docker: docker compose exec ollama ollama pull <model-name>
# Local: ollama pull <model-name>