tappunk

Appearance

Configuration ​

Primary config file:

  • ~/.config/tnk/tnk.toml

Key settings ​

  • server_port - inference API port (default 9931)
  • workspace_root - project root for sandbox mapping (default ~/code)
  • default_provision_profile - sandbox profile default (pi)
  • default_engine_runtime - inference runtime key (llama)
  • default_model - model name injected into sandboxes as TNK_MODEL_NAME (unset by default)
toml
# tnk configuration

# API port for inference server
server_port = 9931

# Root used for project-to-sandbox mapping (must NOT be your home directory)
workspace_root = "~/code"

# Default sandbox profile
default_provision_profile = "pi"

# Inference runtime: "llama"
default_engine_runtime = "llama"

# Model name injected into sandboxes as TNK_MODEL_NAME
# default_model = "ai-fast"

default_model is required for profiled sessions (tnk run, tnk sandbox start with a provision profile, tnk sandbox shell --profile ...): tnk fails fast with a config error instead of provisioning a sandbox against a missing model.

Environment overrides ​

Set these to override file values:

  • TNK_SERVER_PORT
  • TNK_WORKSPACE_ROOT
  • TNK_PROVISION_PROFILE
  • TNK_ENGINE_RUNTIME
  • TNK_MODEL

Security-relevant guidance ​

  1. Keep workspace_root inside a dedicated subtree such as ~/code.
  2. Never set workspace_root to $HOME.
  3. Prefer explicit profile declarations in automation (--profile ...) over implicit defaults.

Inspect effective values:

bash
tnk config show

See also: Installation · Security · Troubleshooting