fix DeepSeek Responses UUID item ids for Codex

docs-site/src/content/docs/ja/reference/configuration/providers.md

@@ -84,7 +84,7 @@ namespace 付き combo alias はその namespace prefix に selector を再利
 | `noTopPModels?` | `string[]` |発信者指定の`top_p`を拒否するモデル。 |
 | `noPenaltyModels?` | `string[]` |存在/周波数ペナルティを拒否するモデル。 |
 | `parallelToolCalls?` | `boolean` |並列ツール呼び出しを切り替えます。 OpenAI Chat はデフォルトでオンになっています。非チャット アダプターは明示的な `true` でのみアドバタイズします。 |
-| `responsesItemIdRepair?` | `{ message?: string[]; reasoning?: string[]; repairMissingTerminalIds?: boolean }` |正確なプレースホルダー ID および欠落している端末 ID に対するダウンストリーム SSE 修復はデフォルトで無効になっています。関数呼び出し ID は決して書き換えられません。 |
+| `responsesItemIdRepair?` | `{ message?: string[]; reasoning?: string[]; repairMissingTerminalIds?: boolean; rewriteNonCanonicalIds?: boolean }` |正確なプレースホルダー ID、欠落している端末 ID、および非標準の UUID 形式 response/item ID（`rewriteNonCanonicalIds`）に対するダウンストリーム SSE 修復はデフォルトで無効です。関数呼び出し ID は決して書き換えられません。 |
 | `autoToolChoiceOnlyModels?` | `string[]` | `tool_choice` が `auto` または `none` のみを受け入れるモデル。強制的な選択は格下げされます。 |
 | `preserveReasoningContentModels?` | `string[]` |チャット履歴に以前のアシスタント `reasoning_content` が必要なモデル。 |
 | `thinkingToggleModels?` | `string[]` |エフォート ラダーではなく `thinking.enabled` を使用してモデルをチャットします。 |
@@ -197,14 +197,15 @@ Anthropic アカウント ポリシーのリスクを理解していない限り
       "responsesItemIdRepair": {
         "reasoning": ["rs_0"],
         "message": ["msg_0"],
-        "repairMissingTerminalIds": true
+        "repairMissingTerminalIds": true,
+        "rewriteNonCanonicalIds": true
       }
     }
   }
 }
 ```
 
-プレースホルダー リストは完全に一致します。通常/ステートフル応答プロバイダーのフィールドを未設定のままにして、パススルーがバイトごとに同一になるようにします。
+プレースホルダー リストは完全に一致します。DeepSeek Responses のように UUID 形式の response/item ID と生の `reasoning_text` ストリームを返すゲートウェイでは `rewriteNonCanonicalIds` を有効にします。OpenCodex はそれらの ID を Codex 向けプレフィックスへ書き換え、reasoning を `encrypted_content` に正規化しつつ replay 用の平文 `content` を保持し、終端の `[DONE]` を補完します。通常/ステートフルな Responses プロバイダーではこのフィールドを未設定のままにして、パススルーがバイト単位で同一になるようにします。
 
 ## Cursor プロバイダー (`adapter: "cursor"`)
 

docs-site/src/content/docs/ko/reference/configuration/providers.md

@@ -84,7 +84,7 @@ target도 selector로 재사용할 수 없습니다. raw account id와 email은
 | `noTopPModels?` | `string[]` | 호출자가 지정한 `top_p`를 거부하는 모델입니다. |
 | `noPenaltyModels?` | `string[]` | presence/frequency penalty를 허용하지 않는 모델입니다. |
 | `parallelToolCalls?` | `boolean` | 병렬 도구 호출을 켜거나 끕니다. OpenAI Chat은 기본으로 켜져 있고, 비-chat 어댑터는 명시적으로 `true`일 때만 이를 노출합니다. |
-| `responsesItemIdRepair?` | `{ message?: string[]; reasoning?: string[]; repairMissingTerminalIds?: boolean }` | 기본값이 꺼진 downstream SSE 복구입니다. 정확한 자리표시자 id와 누락된 종료 id를 복구합니다. function-call id는 다시 쓰지 않습니다. |
+| `responsesItemIdRepair?` | `{ message?: string[]; reasoning?: string[]; repairMissingTerminalIds?: boolean; rewriteNonCanonicalIds?: boolean }` | 기본값이 꺼진 downstream SSE 복구입니다. 정확한 자리표시자 id, 누락된 종료 id, 비표준 UUID 스타일 response/item id(`rewriteNonCanonicalIds`)를 복구합니다. function-call id는 다시 쓰지 않습니다. |
 | `autoToolChoiceOnlyModels?` | `string[]` | `tool_choice`가 `auto` 또는 `none`만 받는 모델입니다. 강제 선택은 낮은 수준으로 바뀝니다. |
 | `preserveReasoningContentModels?` | `string[]` | chat 기록에서 이전 assistant `reasoning_content`가 필요한 모델입니다. |
 | `thinkingToggleModels?` | `string[]` | effort 계층 대신 `thinking.enabled`를 쓰는 chat 모델입니다. |
@@ -200,14 +200,15 @@ Anthropic 계정 정책 위험을 이해하지 못한다면 이 기능은 꺼두
       "responsesItemIdRepair": {
         "reasoning": ["rs_0"],
         "message": ["msg_0"],
-        "repairMissingTerminalIds": true
+        "repairMissingTerminalIds": true,
+        "rewriteNonCanonicalIds": true
       }
     }
   }
 }
 ```
 
-자리표시자 목록은 정확히 일치해야 합니다. 일반적인 상태 유지형 Responses 공급자에서는 필드를 설정하지 마십시오. 그래야 passthrough가 바이트 단위로 동일하게 유지됩니다.
+자리표시자 목록은 정확히 일치해야 합니다. DeepSeek Responses처럼 UUID 형식 response/item id와 원본 `reasoning_text` 스트림을 반환하는 게이트웨이에서는 `rewriteNonCanonicalIds`를 켜십시오. OpenCodex는 해당 id를 Codex 친화적 prefix로 다시 쓰고, reasoning을 `encrypted_content`로 정규화하면서 replay용 평문 `content`는 유지하며, 종료용 `[DONE]`을 보장합니다. 일반/상태 유지형 Responses 공급자에서는 필드를 설정하지 마십시오. 그래야 passthrough가 바이트 단위로 동일하게 유지됩니다.
 
 ## Cursor 공급자 (`adapter: "cursor"`)
 

docs-site/src/content/docs/reference/configuration/providers.md

@@ -91,7 +91,7 @@ differing backup and rewrites known legacy namespaced selected ids to bare ids.
 | `noTopPModels?` | `string[]` | Models that reject caller-specified `top_p`. |
 | `noPenaltyModels?` | `string[]` | Models that reject presence/frequency penalties. |
 | `parallelToolCalls?` | `boolean` | Toggle parallel tool calls. OpenAI Chat defaults on; non-chat adapters advertise only on explicit `true`. |
-| `responsesItemIdRepair?` | `{ message?: string[]; reasoning?: string[]; repairMissingTerminalIds?: boolean }` | Disabled-by-default downstream SSE repair for exact placeholder ids and missing terminal ids. Function-call ids are never rewritten. |
+| `responsesItemIdRepair?` | `{ message?: string[]; reasoning?: string[]; repairMissingTerminalIds?: boolean; rewriteNonCanonicalIds?: boolean }` | Disabled-by-default downstream SSE repair for exact placeholder ids, missing terminal ids, and non-canonical UUID-style response/item ids (`rewriteNonCanonicalIds`). Function-call ids are never rewritten. |
 | `autoToolChoiceOnlyModels?` | `string[]` | Models whose `tool_choice` accepts only `auto` or `none`; forced choices are downgraded. |
 | `preserveReasoningContentModels?` | `string[]` | Models requiring prior assistant `reasoning_content` in chat history. |
 | `thinkingToggleModels?` | `string[]` | Chat models using `thinking.enabled` rather than an effort ladder. |
@@ -237,15 +237,15 @@ For a broken `openai-responses` gateway, repair belongs on the provider object:
       "responsesItemIdRepair": {
         "reasoning": ["rs_0"],
         "message": ["msg_0"],
-        "repairMissingTerminalIds": true
+        "repairMissingTerminalIds": true,
+        "rewriteNonCanonicalIds": true
       }
     }
   }
 }
 ```
 
-Placeholder lists are exact matches. Leave the field unset for normal/stateful Responses providers
-so passthrough stays byte-for-byte identical.
+Placeholder lists are exact matches. Set `rewriteNonCanonicalIds` for gateways such as DeepSeek Responses that emit UUID item/response ids and raw `reasoning_text` streams; OpenCodex rewrites those ids to Codex-friendly prefixes, folds reasoning into `encrypted_content` (while keeping plaintext `content` for replay), and ensures a terminal `[DONE]`. Leave the field unset for normal/stateful Responses providers so passthrough stays byte-for-byte identical.
 
 ## Cursor provider (`adapter: "cursor"`)
 

docs-site/src/content/docs/ru/reference/configuration/providers.md

@@ -94,7 +94,7 @@ cross-route credential fallback не существует. Строки API GPT-
 | `noTopPModels?` | `string[]` | Модели, отвергающие переданный вызывающей стороной `top_p`. |
 | `noPenaltyModels?` | `string[]` | Модели, отвергающие penalty presence/frequency. |
 | `parallelToolCalls?` | `boolean` | Переключатель parallel tool call'ов. Для OpenAI Chat по умолчанию включено; не-chat adapter'ы рекламируют это только при явном `true`. |
-| `responsesItemIdRepair?` | `{ message?: string[]; reasoning?: string[]; repairMissingTerminalIds?: boolean }` | По умолчанию выключенная downstream SSE-repair для exact placeholder-id и отсутствующих terminal-id. Function-call id никогда не переписываются. |
+| `responsesItemIdRepair?` | `{ message?: string[]; reasoning?: string[]; repairMissingTerminalIds?: boolean; rewriteNonCanonicalIds?: boolean }` | По умолчанию выключенная downstream SSE-repair для exact placeholder-id, отсутствующих terminal-id и неканонических UUID-style response/item id (`rewriteNonCanonicalIds`). Function-call id никогда не переписываются. |
 | `autoToolChoiceOnlyModels?` | `string[]` | Модели, у которых `tool_choice` принимает только `auto` или `none`; forced choice понижается. |
 | `preserveReasoningContentModels?` | `string[]` | Модели, которым нужен предыдущий assistant `reasoning_content` в chat history. |
 | `thinkingToggleModels?` | `string[]` | Chat-модели, использующие `thinking.enabled` вместо effort-ladder. |
@@ -245,14 +245,15 @@ Beijing, а `alibaba-token-plan-intl` обслуживает междунаро
       "responsesItemIdRepair": {
         "reasoning": ["rs_0"],
         "message": ["msg_0"],
-        "repairMissingTerminalIds": true
+        "repairMissingTerminalIds": true,
+        "rewriteNonCanonicalIds": true
       }
     }
   }
 }
 ```
 
-Списки placeholder'ов сравниваются по exact-match. Для обычных/stateful Responses-провайдеров это
+Списки placeholder'ов сравниваются по exact-match. Для шлюзов вроде DeepSeek Responses, которые возвращают UUID-style response/item id и сырые `reasoning_text` потоки, включайте `rewriteNonCanonicalIds`: OpenCodex переписывает id в Codex-friendly prefix, нормализует reasoning в `encrypted_content` (сохраняя plaintext `content` для replay) и гарантирует terminal `[DONE]`. Для обычных/stateful Responses-провайдеров это
 поле оставляйте unset, чтобы passthrough оставался byte-for-byte идентичным.
 
 ## Провайдер Cursor (`adapter: "cursor"`)

docs-site/src/content/docs/zh-cn/reference/configuration/providers.md

@@ -83,7 +83,7 @@ pool account id（不能是内部 `__main__`），或用 `"@main"` 表示 Codex
 | `noTopPModels?` | `string[]` | 会拒绝调用方指定 `top_p` 的模型。 |
 | `noPenaltyModels?` | `string[]` | 会拒绝 presence/frequency penalty 的模型。 |
 | `parallelToolCalls?` | `boolean` | 切换并行工具调用。OpenAI Chat 默认开启；非 chat 适配器只有显式 `true` 时才会声明支持。 |
-| `responsesItemIdRepair?` | `{ message?: string[]; reasoning?: string[]; repairMissingTerminalIds?: boolean }` | 默认关闭的下游 SSE 修复，用于精确占位 id 和缺失的终止 id。function-call id 永远不会被重写。 |
+| `responsesItemIdRepair?` | `{ message?: string[]; reasoning?: string[]; repairMissingTerminalIds?: boolean; rewriteNonCanonicalIds?: boolean }` | 默认关闭的下游 SSE 修复，用于精确占位 id、缺失的终止 id，以及非标准 UUID 风格的 response/item id（`rewriteNonCanonicalIds`）。function-call id 永远不会被重写。 |
 | `autoToolChoiceOnlyModels?` | `string[]` | `tool_choice` 只接受 `auto` 或 `none` 的模型；强制选择会被降级。 |
 | `preserveReasoningContentModels?` | `string[]` | 需要在聊天历史中保留先前 assistant `reasoning_content` 的模型。 |
 | `thinkingToggleModels?` | `string[]` | 使用 `thinking.enabled` 而不是 effort 阶梯的 chat 模型。 |
@@ -193,14 +193,15 @@ affinity。这些策略不能规避 provider enforcement。
       "responsesItemIdRepair": {
         "reasoning": ["rs_0"],
         "message": ["msg_0"],
-        "repairMissingTerminalIds": true
+        "repairMissingTerminalIds": true,
+        "rewriteNonCanonicalIds": true
       }
     }
   }
 }
 ```
 
-占位列表必须精确匹配。对于正常/有状态的 Responses 提供者，请保持该字段未设置，以便转发能保持逐字节一致。
+占位列表必须精确匹配。对 DeepSeek Responses 这类会返回 UUID item/response id 与原始 `reasoning_text` 流的网关，请设置 `rewriteNonCanonicalIds`：OpenCodex 会把 id 改写成 Codex 友好前缀，把 reasoning 折叠进 `encrypted_content`（同时保留明文 `content` 便于回放），并补齐终端 `[DONE]`。对于正常/有状态的 Responses 提供者，请保持该字段未设置，以便转发能保持逐字节一致。
 
 ## Cursor 提供者（`adapter: "cursor"`）
 

src/config.ts

@@ -567,6 +567,7 @@ const providerConfigSchema = z.object({
     message: z.array(z.string().min(1)).optional(),
     reasoning: z.array(z.string().min(1)).optional(),
     repairMissingTerminalIds: z.boolean().optional(),
+    rewriteNonCanonicalIds: z.boolean().optional(),
   }).strict().optional(),
 }).passthrough();
 

src/providers/derive.ts

@@ -269,6 +269,14 @@ export function enrichProviderFromRegistry(name: string, prov: OcxProviderConfig
   // the entry so an explicit user value stays distinguishable from the default.
   if (prov.supportsServiceTier === undefined && entry.supportsServiceTier !== undefined) prov.supportsServiceTier = entry.supportsServiceTier;
   if (prov.preserveResponsesReasoningContent === undefined && entry.preserveResponsesReasoningContent !== undefined) prov.preserveResponsesReasoningContent = entry.preserveResponsesReasoningContent;
+  if (prov.responsesItemIdRepair === undefined && entry.responsesItemIdRepair !== undefined) {
+    const repair = entry.responsesItemIdRepair;
+    prov.responsesItemIdRepair = {
+      ...repair,
+      ...(repair.message ? { message: [...repair.message] } : {}),
+      ...(repair.reasoning ? { reasoning: [...repair.reasoning] } : {}),
+    };
+  }
   if (!prov.autoToolChoiceOnlyModels && seed.autoToolChoiceOnlyModels) prov.autoToolChoiceOnlyModels = [...seed.autoToolChoiceOnlyModels];
   if (!prov.preserveReasoningContentModels && seed.preserveReasoningContentModels) prov.preserveReasoningContentModels = [...seed.preserveReasoningContentModels];
   if (!prov.reasoningSplitModels && seed.reasoningSplitModels) prov.reasoningSplitModels = [...seed.reasoningSplitModels];

src/providers/registry.ts

@@ -184,6 +184,8 @@ export interface ProviderRegistryEntry {
   supportsServiceTier?: boolean;
   /** Registry default for plaintext reasoning replay; see `OcxProviderConfig.preserveResponsesReasoningContent`. Registry-only like `supportsServiceTier`. */
   preserveResponsesReasoningContent?: boolean;
+  /** Registry default for client-facing Responses item-id repair (fill-only into runtime provider config). */
+  responsesItemIdRepair?: import("../types").ResponsesItemIdRepairConfig;
   modelDiscovery?: ProviderModelDiscoverySpec;
   contextWindow?: number;
   modelContextWindows?: Record<string, number>;
@@ -1093,6 +1095,14 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
     // route REQUIRES replay on tool-call continuations is an inference from the
     // Chat Thinking-Mode docs, not a confirmed Responses contract.)
     preserveResponsesReasoningContent: true,
+    // DeepSeek Responses emits UUID item/response ids and raw reasoning_text streams that
+    // leave Codex App/CLI stuck on Thinking/Working even after HTTP 200 (#938 family).
+    // Enable the client-facing SSE repair by default for the built-in provider so Codex
+    // Desktop/CLI do not need a hand-edited provider config for the native route.
+    responsesItemIdRepair: {
+      rewriteNonCanonicalIds: true,
+      repairMissingTerminalIds: true,
+    },
     // "The API is stateless: responses and conversations are not stored on the
     // server." https://api-docs.deepseek.com/api/create-response/
     statelessResponses: true,

src/server/relay-eager.ts

@@ -58,6 +58,8 @@ export type EagerRelayHooks = {
 };
 
 export type EagerRelayOptions = {
+  /** Optional trailer emitted after rewrite flush (e.g. data: [DONE]). */
+  trailer?: string | (() => string | undefined);
   /** Bounded client queue in bytes; producer pauses above it. Default 8 MiB. */
   maxQueueBytes?: number;
   /** Transient-budget owner for the inline-rewrite frame buffer. */
@@ -123,14 +125,16 @@ export function relaySseEagerBounded(
       if (!next) break;
       const payload = sseDataPayload(next.block);
       const rewrittenPayload = payload === null ? null : rewrite!(payload);
+      frameBuffer = next.rest;
+      // Returning null from rewrite drops the entire SSE event.
+      if (payload !== null && rewrittenPayload === null) continue;
       // Replace only on an actual change: replaceSseDataPayload collapses
       // multi-data-line events and normalizes newline style even when the
       // payload is identical, which corrupts valid streams.
-      const block = payload !== null && rewrittenPayload !== payload
-        ? replaceSseDataPayload(next.block, rewrittenPayload!)
+      const block = payload !== null && rewrittenPayload !== null && rewrittenPayload !== payload
+        ? replaceSseDataPayload(next.block, rewrittenPayload)
         : next.block;
       out += block + next.delimiter;
-      frameBuffer = next.rest;
     }
     if (rewriteBudget) {
       const remaining = rewriteEncoder!.encode(frameBuffer).byteLength;
@@ -149,6 +153,14 @@ export function relaySseEagerBounded(
     const payload = sseDataPayload(tail);
     if (payload !== null) {
       const rewrittenPayload = rewrite(payload);
+      if (rewrittenPayload === null) {
+        frameBuffer = "";
+        if (rewriteBudget && frameBufferBytes > 0) {
+          rewriteBudget.releaseRetained(frameBufferBytes, { kind: "live_transient" });
+        }
+        frameBufferBytes = 0;
+        return rewriteEncoder!.encode("");
+      }
       if (rewrittenPayload !== payload) tail = replaceSseDataPayload(tail, rewrittenPayload);
     }
     frameBuffer = "";
@@ -222,6 +234,12 @@ export function relaySseEagerBounded(
               queuedBytes += tail.byteLength;
               try { controllerRef?.enqueue(tail); } catch { /* client already gone */ }
             }
+            const trailer = typeof opts?.trailer === "function" ? opts.trailer() : opts?.trailer;
+            if (trailer && !cancelled) {
+              const trailerBytes = new TextEncoder().encode(trailer);
+              queuedBytes += trailerBytes.byteLength;
+              try { controllerRef?.enqueue(trailerBytes); } catch { /* client already gone */ }
+            }
           }
           if (!hooks.sawTerminal() && !cancelled && !upstream.signal.aborted) {
             syntheticKind = "incomplete";