@nylorun/runtime 0.28.0-beta → 0.29.0-beta

package/dist/store/postgres/drizzle/0037_model_usage_cost_split.sql

@@ -0,0 +1,4 @@
+ALTER TABLE "nylorun"."model_usage" ADD COLUMN "cost_input_usd" double precision;--> statement-breakpoint
+ALTER TABLE "nylorun"."model_usage" ADD COLUMN "cost_cache_read_usd" double precision;--> statement-breakpoint
+ALTER TABLE "nylorun"."model_usage" ADD COLUMN "cost_cache_write_usd" double precision;--> statement-breakpoint
+ALTER TABLE "nylorun"."model_usage" ADD COLUMN "cost_output_usd" double precision;

package/dist/store/postgres/drizzle/meta/_journal.json

@@ -260,6 +260,13 @@
       "when": 1792172016681,
       "tag": "0036_t5_memories",
       "breakpoints": true
+    },
+    {
+      "idx": 37,
+      "version": "7",
+      "when": 1792238141315,
+      "tag": "0037_model_usage_cost_split",
+      "breakpoints": true
     }
   ]
 }

package/dist/store/postgres/reads.js

@@ -32,6 +32,18 @@ function grantsWhere(column, grants) {
 function accessWhere(access) {
     return and(access.owner === undefined ? undefined : eq(sessions.ownerUserId, access.owner), access.agents === undefined ? undefined : grantsWhere(sessions.agentId, access.agents));
 }
+/** A row's cost by token kind, when the adapter kept it (migration 0037). */
+function costSplitOf(row) {
+    const { costInputUsd, costCacheReadUsd, costCacheWriteUsd, costOutputUsd } = row;
+    if (costInputUsd === null || costCacheReadUsd === null || costCacheWriteUsd === null || costOutputUsd === null)
+        return null;
+    return {
+        input: costInputUsd,
+        cacheRead: costCacheReadUsd,
+        cacheWrite: costCacheWriteUsd,
+        output: costOutputUsd,
+    };
+}
 function publicCall(row) {
     return {
         id: row.id,
@@ -42,6 +54,8 @@ function publicCall(row) {
         provider: row.provider,
         model: row.model,
         at: row.createdAt,
+        startedAt: row.startedAt,
+        endedAt: row.endedAt,
         outcome: "completed",
         usage: {
             inputTokens: row.inputTokens,
@@ -54,6 +68,7 @@ function publicCall(row) {
             duplicate: row.duplicate,
             tokensReported: row.tokensReported,
             costKnown: row.costKnown,
+            costSplit: costSplitOf(row),
         },
     };
 }
@@ -225,11 +240,26 @@ export function createPostgresReadStore(source, tenantId) {
                 unpricedCalls: sql `count(*) filter (where ${modelUsage.costKnown} = false)`.mapWith(Number),
                 unreportedCalls: sql `count(*) filter (where ${modelUsage.tokensReported} = false)`.mapWith(Number),
                 unknownQualityCalls: sql `count(*) filter (where ${modelUsage.tokensReported} is null or ${modelUsage.costKnown} is null)`.mapWith(Number),
+                costInput: sum(modelUsage.costInputUsd),
+                costCacheRead: sum(modelUsage.costCacheReadUsd),
+                costCacheWrite: sum(modelUsage.costCacheWriteUsd),
+                costOutput: sum(modelUsage.costOutputUsd),
+                splitCalls: sql `count(${modelUsage.costInputUsd})`.mapWith(Number),
+                // The gate's timing is ISO text; a row timed before migration 0022 has none.
+                modelTimeMs: sql `coalesce(round(sum(extract(epoch from (${modelUsage.endedAt}::timestamptz - ${modelUsage.startedAt}::timestamptz)) * 1000)), 0)`.mapWith(Number),
+                timedCalls: sql `count(*) filter (where ${modelUsage.startedAt} is not null and ${modelUsage.endedAt} is not null)`.mapWith(Number),
             })
                 .from(modelUsage)
                 .where(and(eq(modelUsage.sessionId, id), turnId === undefined ? undefined : eq(modelUsage.turnId, turnId)));
+            const { costInput, costCacheRead, costCacheWrite, costOutput, ...rest } = totals;
             return {
-                ...totals,
+                ...rest,
+                costSplit: {
+                    input: costInput,
+                    cacheRead: costCacheRead,
+                    cacheWrite: costCacheWrite,
+                    output: costOutput,
+                },
                 sessionId: id,
                 turnId: turnId ?? null,
                 asOf: new Date().toISOString(),

package/dist/store/postgres/schema.d.ts

@@ -1793,7 +1793,7 @@ export declare const agentTriggers: import("drizzle-orm/pg-core").PgTableWithCol
 }>;
 /**
  * An Agent's memories (track T5, D72): one row per Markdown file under `/memories`, its content
- * inline (the Runtime limit `memoryBytes`, 100 KB by default, bounds it). `version` counts the
+ * inline (the Runtime limit `memoryBytes`, 100 KiB by default, bounds it). `version` counts the
  * changes to this path; `description` is its frontmatter's, for the pinned index. A write locks the
  * one row it changes, so writes to different memories never wait on each other.
  */
@@ -3401,7 +3401,8 @@ export declare const signingKeys: import("drizzle-orm/pg-core").PgTableWithColum
  * flagged `duplicate`; both were billed. Budgets read their spend from it. Token counts are 0
  * when the provider sent none. `startedAt`, `endedAt` and `firstTokenMs` are the gate's own
  * timing of the call (R7.1, migration 0022); null on rows written before, and `firstTokenMs`
- * when the provider streamed nothing.
+ * when the provider streamed nothing. The `cost*Usd` columns split `costUsd` by token kind, at the
+ * prices the adapter applied (migration 0037); null on rows written before.
  */
 export declare const modelUsage: import("drizzle-orm/pg-core").PgTableWithColumns<{
     name: "model_usage";
@@ -3793,6 +3794,74 @@ export declare const modelUsage: import("drizzle-orm/pg-core").PgTableWithColumn
             identity: undefined;
             generated: undefined;
         }, {}, {}>;
+        costInputUsd: import("drizzle-orm/pg-core").PgColumn<{
+            name: "costInputUsd";
+            tableName: "model_usage";
+            dataType: "number";
+            columnType: "PgDoublePrecision";
+            data: number;
+            driverParam: string | number;
+            notNull: false;
+            hasDefault: false;
+            isPrimaryKey: false;
+            isAutoincrement: false;
+            hasRuntimeDefault: false;
+            enumValues: undefined;
+            baseColumn: never;
+            identity: undefined;
+            generated: undefined;
+        }, {}, {}>;
+        costCacheReadUsd: import("drizzle-orm/pg-core").PgColumn<{
+            name: "costCacheReadUsd";
+            tableName: "model_usage";
+            dataType: "number";
+            columnType: "PgDoublePrecision";
+            data: number;
+            driverParam: string | number;
+            notNull: false;
+            hasDefault: false;
+            isPrimaryKey: false;
+            isAutoincrement: false;
+            hasRuntimeDefault: false;
+            enumValues: undefined;
+            baseColumn: never;
+            identity: undefined;
+            generated: undefined;
+        }, {}, {}>;
+        costCacheWriteUsd: import("drizzle-orm/pg-core").PgColumn<{
+            name: "costCacheWriteUsd";
+            tableName: "model_usage";
+            dataType: "number";
+            columnType: "PgDoublePrecision";
+            data: number;
+            driverParam: string | number;
+            notNull: false;
+            hasDefault: false;
+            isPrimaryKey: false;
+            isAutoincrement: false;
+            hasRuntimeDefault: false;
+            enumValues: undefined;
+            baseColumn: never;
+            identity: undefined;
+            generated: undefined;
+        }, {}, {}>;
+        costOutputUsd: import("drizzle-orm/pg-core").PgColumn<{
+            name: "costOutputUsd";
+            tableName: "model_usage";
+            dataType: "number";
+            columnType: "PgDoublePrecision";
+            data: number;
+            driverParam: string | number;
+            notNull: false;
+            hasDefault: false;
+            isPrimaryKey: false;
+            isAutoincrement: false;
+            hasRuntimeDefault: false;
+            enumValues: undefined;
+            baseColumn: never;
+            identity: undefined;
+            generated: undefined;
+        }, {}, {}>;
     };
     dialect: 'pg';
 }>;
@@ -5656,10 +5725,14 @@ export type EnvironmentRowColumns = typeof agents.$inferSelect;
 export type ArtifactRow = typeof artifacts.$inferSelect;
 export type ArtifactVersionRow = typeof artifactVersions.$inferSelect;
 export type DefinitionFileRow = typeof definitionFiles.$inferSelect;
-export type ModelUsageWrite = Omit<ModelUsageRow, "duplicate" | "txid" | "tokensReported" | "costKnown" | "startedAt" | "endedAt" | "firstTokenMs"> & {
+export type ModelUsageWrite = Omit<ModelUsageRow, "duplicate" | "txid" | "tokensReported" | "costKnown" | "startedAt" | "endedAt" | "firstTokenMs" | "costInputUsd" | "costCacheReadUsd" | "costCacheWriteUsd" | "costOutputUsd"> & {
     tokensReported?: boolean | null;
     costKnown?: boolean | null;
     startedAt?: string | null;
     endedAt?: string | null;
     firstTokenMs?: number | null;
+    costInputUsd?: number | null;
+    costCacheReadUsd?: number | null;
+    costCacheWriteUsd?: number | null;
+    costOutputUsd?: number | null;
 };

package/dist/store/postgres/schema.js

@@ -251,7 +251,7 @@ export const agentTriggers = nylorun.table("agent_triggers", {
 ]);
 /**
  * An Agent's memories (track T5, D72): one row per Markdown file under `/memories`, its content
- * inline (the Runtime limit `memoryBytes`, 100 KB by default, bounds it). `version` counts the
+ * inline (the Runtime limit `memoryBytes`, 100 KiB by default, bounds it). `version` counts the
  * changes to this path; `description` is its frontmatter's, for the pinned index. A write locks the
  * one row it changes, so writes to different memories never wait on each other.
  */
@@ -425,7 +425,8 @@ export const signingKeys = nylorun.table("signing_keys", {
  * flagged `duplicate`; both were billed. Budgets read their spend from it. Token counts are 0
  * when the provider sent none. `startedAt`, `endedAt` and `firstTokenMs` are the gate's own
  * timing of the call (R7.1, migration 0022); null on rows written before, and `firstTokenMs`
- * when the provider streamed nothing.
+ * when the provider streamed nothing. The `cost*Usd` columns split `costUsd` by token kind, at the
+ * prices the adapter applied (migration 0037); null on rows written before.
  */
 export const modelUsage = nylorun.table("model_usage", {
     id: textC().primaryKey(),
@@ -459,6 +460,14 @@ export const modelUsage = nylorun.table("model_usage", {
     endedAt: text(),
     /** From `startedAt` to the provider's first stream event. */
     firstTokenMs: integer(),
+    /** What the uncached input cost. */
+    costInputUsd: doublePrecision(),
+    /** What the input read from the provider's cache cost. */
+    costCacheReadUsd: doublePrecision(),
+    /** What the input written to the provider's cache cost. */
+    costCacheWriteUsd: doublePrecision(),
+    /** What the output cost. */
+    costOutputUsd: doublePrecision(),
 }, (t) => [
     index("model_usage_session").on(t.sessionId, t.createdAt, t.id),
     index("model_usage_export").on(t.txid, t.id),

package/dist/tenant/types.d.ts

@@ -74,6 +74,11 @@ export interface TenantConfig {
      * the identity file (`NYLORUN_IDENTITY_FILE`). Absent: none.
      */
     issuers?: import("./issuers.js").TrustedIssuers;
+    /**
+     * The URL clients reach the Runtime API at (`NYLORUN_PUBLIC_URL`, no trailing slash): the base
+     * of webhook URLs and the OAuth resource. Absent: the origin each request reached.
+     */
+    publicUrl?: string;
     logger: Logger;
 }
 export interface Logger {

package/dist/version.d.ts

@@ -1 +1 @@
-export declare const RUNTIME_VERSION = "0.28.0-beta";
+export declare const RUNTIME_VERSION = "0.29.0-beta";

package/dist/version.js

@@ -1,3 +1,3 @@
 // Kept in step with package.json by runtime/server/scripts/check-package.mjs, which runs in npm run check.
 // release:prepare also rewrites this when it bumps @nylorun/runtime.
-export const RUNTIME_VERSION = "0.28.0-beta";
+export const RUNTIME_VERSION = "0.29.0-beta";

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@nylorun/runtime",
-  "version": "0.28.0-beta",
+  "version": "0.29.0-beta",
   "description": "Nylorun Runtime: the agent execution Host on Postgres, Restate and S2, and model providers.",
   "type": "module",
   "license": "Apache-2.0",
@@ -65,8 +65,8 @@
     "@hono/node-server": "^2.1.5",
     "@hono/zod-openapi": "^1.6.3",
     "@modelcontextprotocol/sdk": "^1.32.1",
-    "@nylorun/core": "0.24.0-beta",
-    "@nylorun/harness": "0.29.0-beta",
+    "@nylorun/core": "0.25.0-beta",
+    "@nylorun/harness": "0.29.1-beta",
     "@restatedev/restate-sdk": "1.17.3",
     "@restatedev/restate-sdk-clients": "1.17.3",
     "@s2-dev/streamstore": "0.28.0",