llamavault.yaml

# ============================================================================
# LlamaVault Configuration
# ============================================================================
# Single config file for the entire system. All scripts and services read
# from this file. Edit this, then run: make dev / make deploy / make up
# ============================================================================

# --- Model ------------------------------------------------------------------
model:
  name: llama-3.2-3b-instruct           # Model name from models/models.yaml
  file: llama-3.2-3b-instruct-q4_k_m.gguf
  context_size: 4096
  max_tokens: 4096
  temperature: 0.7
  threads: 4                             # CPU threads for inference
  parallel: 2                            # Concurrent request slots

# --- Agent ------------------------------------------------------------------
agent:
  memory_window: 100                     # Messages kept in context
  restrict_to_workspace: true            # Sandbox file access to workspace
  port: 18790                            # Agent gateway port
  # api_key: ""                          # Set to protect the gateway API

# --- MCP Servers ------------------------------------------------------------
# External capabilities via Model Context Protocol.
# Each entry is a named server with either stdio or HTTP transport.
mcp_servers:
  # Example — filesystem access:
  # filesystem:
  #   command: npx
  #   args: ["-y", "@anthropic/mcp-filesystem-server", "/data"]
  #
  # Example — GitHub:
  # github:
  #   command: npx
  #   args: ["-y", "@anthropic/mcp-github-server"]
  #   env:
  #     GITHUB_TOKEN: ghp_xxx
  #
  # Example — HTTP transport:
  # custom:
  #   url: http://localhost:3000/mcp
  #   headers:
  #     Authorization: "Bearer xxx"

mcp_allowed_commands: [npx, uvx]

# --- Security (AI SAFE² Gateway) -------------------------------------------
security:
  level: strict
  block_prompt_injection: true
  block_pii: true
  max_requests_per_minute: 60
  kill_switch: false

# --- Paths ------------------------------------------------------------------
paths:
  models_dir: ./models/data
  data_dir: ./data/nanobot

# --- Deploy (AWS Lightsail) -------------------------------------------------
deploy:
  region: us-east-1
  availability_zone: us-east-1a
  bundle: 2xlarge_3_0                    # xlarge_3_0 (8GB), 2xlarge_3_0 (16GB), 4xlarge_3_0 (32GB)
  disk_size_gb: 60
  instance_name: llamavault
  # ssh_cidrs: ["YOUR_IP/32"]           # REQUIRED for deploy — restrict SSH access