llm via Tokonomix

//What is llm?

llm is Simon Willison’s command-line client and Python library for large language models. It logs every prompt and response to SQLite, supports plugins, and can register any OpenAI-compatible model through a YAML file — which is exactly how Tokonomix plugs in.

• Runs prompts from the shell and pipes stdin through a model.
• Registers extra OpenAI-compatible models in extra-openai-models.yaml (model_id, model_name, api_base, api_key_name).
• Supports tool calling via the supports_tools flag on a model entry.
• Open source; keys are stored with llm keys set, not in the YAML.

//Why route llm through Tokonomix?

• One Tokonomix key in extra-openai-models.yaml reaches 130+ models without juggling per-provider keys.
• Point a model_id at tokonomix-consensus so hard queries get a multi-model verdict; use a direct slug for routine one-liners.
• Per-call spend caps bound usage, and region / origin routing keeps sensitive prompts compliant.
• Every call is metered to the cent in your Tokonomix dashboard.

//Connect Tokonomix in extra-openai-models.yaml

Add a Tokonomix model entry to extra-openai-models.yaml, then store the key with llm keys set.

# ~/.config/io.datasette.llm/extra-openai-models.yaml
- model_id: anthropic/claude-sonnet-4-6
  model_name: anthropic/claude-sonnet-4-6
  api_base: "https://tokonomix.ai/api/v1"
  api_key_name: tokonomix
  supports_tools: true
# Opt-in (planning/review only, no tool-calling): add a second entry with model_id/model_name: tokonomix-consensus

# then: llm keys set tokonomix   (paste tok-your-key)
# llm -m anthropic/claude-sonnet-4-6 "Explain quorum sensing"

//Verify it works

• Add the model entry and run llm keys set tokonomix to store your key.
• Send a small prompt: llm -m tokonomix-consensus "Explain quorum sensing".
• Confirm the call and token usage in your Tokonomix dashboard.