@runlot/api 0.1.0-rc.190 → 0.1.0-rc.191

package/dist/schema.d.ts

@@ -12577,6 +12577,13 @@ export interface components {
              * @description The price of cached input. Absent means the model has no cache discount and cached tokens are billed at the full input price
              */
             cachedInPerMTok?: number;
+            /**
+             * Format: int64
+             * @description The price of input written to the provider's prompt cache. Absent means a write is billed at the input price
+             */
+            cacheWritePerMTok?: number;
+            /** @description Every `cache.ttl` that `env.ai.run` accepts for this model. Absent means the model takes no `cache` option */
+            cacheTTLs?: string[];
             /**
              * @description When true, this is the provider's **list price**, passed straight through with no
              *     margin added.
@@ -20171,6 +20178,8 @@ export interface operations {
                         /** @enum {string} */
                         role: "system" | "user" | "assistant";
                         content: string;
+                        /** @description End a cached prefix after this message. User messages only, on a model whose catalog row lists `cacheTTLs`. */
+                        cacheBreakpoint?: boolean;
                     }[];
                     /** @description capped at 4,096 by the lane */
                     maxTokens?: number;
@@ -20187,6 +20196,17 @@ export interface operations {
                         name?: string;
                         schema?: Record<string, never>;
                     };
+                    /** @description `env.ai.run`'s explicit prompt cache. Only on a model whose catalog row lists `cacheTTLs`. */
+                    cache?: {
+                        /** @enum {string} */
+                        mode?: "implicit" | "explicit";
+                        /** @description one of the model's `cacheTTLs` */
+                        ttl?: string;
+                        /** @description the provider's prompt cache key — calls that share it are routed together */
+                        key?: string;
+                        /** @description end a cached prefix after the system prompt */
+                        system?: boolean;
+                    };
                 };
             };
         };
@@ -20207,8 +20227,16 @@ export interface operations {
                             inputTokens?: number;
                             /** Format: int64 */
                             outputTokens?: number;
-                            /** Format: int64 */
+                            /**
+                             * Format: int64
+                             * @description the part of inputTokens read from the prompt cache
+                             */
                             cachedInputTokens?: number;
+                            /**
+                             * Format: int64
+                             * @description the part of inputTokens written to the prompt cache
+                             */
+                            cacheWriteInputTokens?: number;
                         };
                     };
                 };