fix(server): OpenCode ChatGPT usage limits keep their reset time

apps/server/src/orchestration-v2/Adapters/OpenCode2AdapterV2.test.ts

@@ -561,6 +561,21 @@ it.layer(McpProviderSessions.layer)("OpenCode2 adapter", (it) => {
         status: 429,
       },
     },
+    {
+      name: "a ChatGPT provider.quota usage window (429)",
+      failureClass: "usage_limit",
+      error: {
+        type: "provider.quota",
+        message: "The usage limit has been reached",
+        status: 429,
+        response: {
+          body: JSON.stringify({
+            error: { type: "usage_limit_reached", resets_at: 1793114741, resets_in_seconds: 222 },
+          }),
+        },
+      },
+      resetAt: "2026-10-27T15:25:41.000Z",
+    },
     {
       name: "a provider.rate-limit failure (429)",
       failureClass: "usage_limit",
@@ -586,7 +601,7 @@ it.layer(McpProviderSessions.layer)("OpenCode2 adapter", (it) => {
         message: "Request failed. Resets at 2026-11-02T20:00:00Z.",
       },
     },
-  ] as const)("classifies $name as $failureClass", ({ error, failureClass }) =>
+  ] as const)("classifies $name as $failureClass", ({ error, failureClass, ...row }) =>
     Effect.gen(function* () {
       const { runtime, thread } = yield* resumed([
         out("session.prompt", { sessionID: SESSION, text: "<any>" }),
@@ -602,6 +617,7 @@ it.layer(McpProviderSessions.layer)("OpenCode2 adapter", (it) => {
         code: error.type,
         message: error.message,
         retryable: null,
+        ...("resetAt" in row ? { resetAt: row.resetAt } : {}),
       });
     }).pipe(Effect.scoped),
   );

packages/provider-opencode/src/server/v2/adapter.ts

@@ -45,6 +45,7 @@ import * as Effect from "effect/Effect";
 import * as Hex from "effect/encoding/Hex";
 import * as Exit from "effect/Exit";
 import * as FileSystem from "effect/FileSystem";
+import * as Option from "effect/Option";
 import * as Queue from "effect/Queue";
 import * as Schedule from "effect/Schedule";
 import * as Scope from "effect/Scope";
@@ -531,6 +532,17 @@ const claudeUsageLimit = (message: string | undefined) =>
     ? { resetAt: /\bResets at (\d{4}-\d\d-\d\dT[\d:.]+Z)\.?\s*$/.exec(message)?.[1] ?? null }
     : undefined;
 
+// ChatGPT-backed OpenAI models put the usage-limit reset in the 429 body as Unix seconds.
+const decodeUsageLimitBody = Schema.decodeUnknownOption(
+  Schema.fromJsonString(Schema.Struct({ error: Schema.Struct({ resets_at: Schema.Finite }) })),
+);
+const usageLimitResetAt = (body: string | undefined) =>
+  decodeUsageLimitBody(body).pipe(
+    Option.flatMap(({ error }) => DateTime.make(error.resets_at * 1000)),
+    Option.map(DateTime.formatIso),
+    Option.getOrNull,
+  );
+
 const INTERRUPT_TIMEOUT = "10 seconds";
 const INSTRUCTIONS_KEY = "t3-code";
 const RECONNECT_ATTEMPTS = 5;
@@ -2168,7 +2180,7 @@ export const make = Effect.fn("OpenCode2Adapter.make")(function* (instanceId: Pr
               message: readableProviderErrorMessage(message, status, response?.body),
               code: type,
               class: usageLimit ? "usage_limit" : "provider_error",
-              resetAt: claudeLimit?.resetAt ?? null,
+              resetAt: claudeLimit?.resetAt ?? usageLimitResetAt(response?.body),
             }),
           });
         }