Skip to main content

Configuration reference

AI Rulebook compiler

The creation page and POST /api/v1/compile fail closed when the selected provider is incomplete. Groq uses one configured model at a time. The configured model IDs are openai/gpt-oss-120b and qwen/qwen3.6-27b; selecting one never silently falls back to the other. GPT-OSS uses Groq JSON Schema constrained decoding. Qwen uses Groq JSON Object mode and then passes through the same local Zod schema and canonical Rulebook validation. The SDK also accepts any AI SDK LanguageModel without changing Slip’s canonical validation.

Solana

Deployment gate

This flag does not upgrade or detect a binary by itself. It is an operator assertion made after deployment and a live lifecycle check. With the flag false, market reads remain available but writes are absent from the UI. The server and browser settlement-mint values must identify the same mint before writes are enabled. Markets using another six-decimal mint are not projected into the product, preventing an arbitrary mint from being labelled as the configured settlement asset. Market creation additionally requires the AI variables above. A deployed program alone does not cause the UI to expose a nonfunctional natural-language compiler.

TxLINE

Keeper operations

Invalid or missing required keeper variables fail during startup. There is no local or sample fallback.