AxContextCacheOptions

type AxContextCacheOptions = object;

Defined in: https://github.com/ax-llm/ax/blob/5b28f9093bb70863b59459bb6df5062d005bce41/src/ax/ai/types.ts#L755

Options for explicit context caching (e.g., Gemini/Vertex context caching). Allows caching large prompt prefixes for cost savings and lower latency.

When this option is present, caching is enabled. The system will:

• Automatically cache the system prompt and any content marked with cache: true
• Reuse existing caches when content hash matches
• Create new caches when content changes
• Auto-refresh TTL when cache is near expiration

Properties

cacheBreakpoint?

optional cacheBreakpoint: "system" | "after-functions" | "after-examples";

Defined in: https://github.com/ax-llm/ax/blob/5b28f9093bb70863b59459bb6df5062d005bce41/src/ax/ai/types.ts#L797

Controls where the cache breakpoint is set in the prompt prefix. Prefix order: System → Functions → Examples → User Input

• ‘after-examples’ (default): Cache includes system + functions + examples
• ‘after-functions’: Cache includes system + functions only (use when examples are dynamic)
• ‘system’: Cache includes only system prompt (use when functions are dynamic)

minTokens?

optional minTokens: number;

Defined in: https://github.com/ax-llm/ax/blob/5b28f9093bb70863b59459bb6df5062d005bce41/src/ax/ai/types.ts#L774

Minimum token threshold for creating explicit caches. Content below this threshold won’t create explicit caches (implicit caching still applies). Default: 2048 (Gemini minimum requirement)

name?

optional name: string;

Defined in: https://github.com/ax-llm/ax/blob/5b28f9093bb70863b59459bb6df5062d005bce41/src/ax/ai/types.ts#L761

Explicit cache resource name/ID. If provided, this cache will be used directly (bypasses auto-creation). If omitted, a cache will be created/looked up automatically.

refreshWindowSeconds?

optional refreshWindowSeconds: number;

Defined in: https://github.com/ax-llm/ax/blob/5b28f9093bb70863b59459bb6df5062d005bce41/src/ax/ai/types.ts#L780

Window in seconds before expiration to trigger automatic TTL refresh. Default: 300 (5 minutes)

registry?

optional registry: AxContextCacheRegistry;

Defined in: https://github.com/ax-llm/ax/blob/5b28f9093bb70863b59459bb6df5062d005bce41/src/ax/ai/types.ts#L787

External registry for persisting cache metadata. If provided, cache lookups and storage will use this registry instead of in-memory storage. Useful for serverless/short-lived processes.

ttlSeconds?

optional ttlSeconds: number;

Defined in: https://github.com/ax-llm/ax/blob/5b28f9093bb70863b59459bb6df5062d005bce41/src/ax/ai/types.ts#L767

TTL (Time To Live) in seconds for the cache. Default: 3600 (1 hour). Maximum varies by provider.