@jeffreycao/copilot-api 2.3.11 → 2.3.12

package/README.md

@@ -310,6 +310,8 @@ Without this configuration, Codex cannot fetch `/v1/models` while not signed in
 
 When a Codex client (`User-Agent` starts with `codex`) requests the top-level `GET /v1/models`, the gateway merges native Codex models with models available through the Messages adapter. The latter advertise `use_responses_lite: true`, except DeepSeek models, which use `use_responses_lite: false` and `tool_mode: null`. For other models, `/v1/responses` uses **Responses → Messages** for Anthropic providers, while OpenAI-compatible providers and Chat-only Copilot models reuse the existing Messages route for **Responses → Messages → Chat Completions**, then translate streaming or JSON results back to Responses.
 
+> **Note:** DeepSeek models do not use Responses Lite (`use_responses_lite: false`, `tool_mode: null`), so the tool set they advertise to Codex differs from other models, which use `tool_mode: "code_mode_only"`. Switching between a DeepSeek model and a Responses Lite model mid-session is not compatible, because tool calls and conversation history produced under one tool set do not translate to the other. Start a new Codex session when switching between them.
+
 The merged catalog is what Codex shows in its model picker, including the models exposed by your configured providers:
 
 <img src="./docs/screenshots/codex-models.png" alt="Codex model picker showing models provided by the gateway" width="900" />

package/README.zh-CN.md

@@ -326,6 +326,8 @@ args = [
 
 Codex 客户端（`User-Agent` 以 `codex` 开头）请求顶层 `GET /v1/models` 时，网关会把原生 Codex 模型与可通过 Messages 适配的模型合并返回。除 DeepSeek 模型外，后者会声明 `use_responses_lite: true`；DeepSeek 模型使用 `use_responses_lite: false` 和 `tool_mode: null`。调用 `/v1/responses` 后，Anthropic provider 走 **Responses → Messages**，OpenAI 兼容 provider 以及只支持 Chat 的 Copilot 模型则复用现有 Messages 路由继续走 **Responses → Messages → Chat Completions**，最终统一翻译回 Responses（包括流式事件）。
 
+> **注意：** DeepSeek 模型不使用 Responses Lite（`use_responses_lite: false`、`tool_mode: null`），因此向 Codex 暴露的工具集合与其他模型（`tool_mode: "code_mode_only"`）不一致。在会话中途切换 DeepSeek 模型与 Responses Lite 模型并不兼容——一套工具集合下产生的工具调用和会话历史无法直接沿用到另一套。切换模型时请新建 Codex 会话。
+
 合并后的模型列表会直接展示在 Codex 的模型选择界面中，包含各 provider 暴露的模型：
 
 <img src="./docs/screenshots/codex-models.png" alt="Codex 模型选择界面展示网关提供的模型列表" width="900" />

package/dist/main.js

@@ -30,7 +30,7 @@ if (isMcpFastPath(process.argv)) {
 	const { auth } = await import("./auth-DgKX2oyO.js");
 	const { debug } = await import("./debug-BHOitPm3.js");
 	const { mcp } = await import("./mcp-fpSlKZxK.js");
-	const { start } = await import("./start-D1lA7Gjm.js");
+	const { start } = await import("./start-BaYrtsEL.js");
 	await runMain(defineCommand({
 		meta: {
 			name: "copilot-api",

package/dist/{server-m80zRYma.js → server-fiF0jU-7.js}

@@ -8697,6 +8697,9 @@ providerModelRoutes.get("/", async (c) => {
 //#endregion
 //#region src/routes/responses/messages-translation.ts
 const MESSAGES_COMPACTION_PREFIX = "copilot-api:messages-compaction:v1:";
+const MESSAGES_REASONING_ID_SUFFIX = "__a1";
+const markMessagesReasoningId = (id) => `${id}${MESSAGES_REASONING_ID_SUFFIX}`;
+const isMessagesReasoningId = (id) => typeof id === "string" && id.endsWith(MESSAGES_REASONING_ID_SUFFIX);
 const MESSAGES_COMPACTION_PROMPT = [
 	"You are performing a CONTEXT CHECKPOINT COMPACTION. Create a handoff summary for another LLM that will resume the task.",
 	"Do NOT continue the task, make changes, or call any tools. Your only output must be the handoff summary.",
@@ -9206,7 +9209,7 @@ function translateAssistantOutput(response, registry) {
 	for (const [index, block] of response.content.entries()) {
 		if (block.type === "thinking") {
 			output.push({
-				id: `rs_${createStableHash(`${response.id}:${index}:reasoning`)}`,
+				id: markMessagesReasoningId(`rs_${createStableHash(`${response.id}:${index}:reasoning`)}`),
 				type: "reasoning",
 				status: "completed",
 				...block.thinking && block.thinking !== "Thinking..." ? { summary: [{
@@ -9238,7 +9241,6 @@ function translateAssistantOutput(response, registry) {
 function translateToolUseOutput(block, registry, idSeed) {
 	const descriptor = resolveToolDescriptor(registry, block.name);
 	const common = {
-		id: `fc_${createStableHash(idSeed)}`,
 		call_id: block.id,
 		name: descriptor.name,
 		status: "completed",
@@ -9246,11 +9248,13 @@ function translateToolUseOutput(block, registry, idSeed) {
 	};
 	if (descriptor.kind === "custom") return {
 		...common,
+		id: `ctc_${createStableHash(idSeed)}`,
 		type: "custom_tool_call",
 		input: decodeCustomToolInput(block.input)
 	};
 	return {
 		...common,
+		id: `fc_${createStableHash(idSeed)}`,
 		type: "function_call",
 		arguments: JSON.stringify(block.input)
 	};
@@ -9845,7 +9849,7 @@ function* startContentBlock(state, event) {
 	}
 	if (block.type === "thinking") {
 		const item = {
-			id: `rs_${state.responseId.slice(-18)}_${event.index}`,
+			id: markMessagesReasoningId(`rs_${state.responseId.slice(-18)}_${event.index}`),
 			type: "reasoning",
 			status: "in_progress",
 			summary: [],
@@ -9872,7 +9876,6 @@ function* startContentBlock(state, event) {
 	if (block.type === "tool_use") {
 		const descriptor = resolveToolDescriptor(state.context.registry, block.name);
 		const common = {
-			id: `fc_${state.responseId.slice(-18)}_${event.index}`,
 			call_id: block.id,
 			name: descriptor.name,
 			status: "in_progress",
@@ -9881,6 +9884,7 @@ function* startContentBlock(state, event) {
 		if (descriptor.kind === "custom") {
 			const item = {
 				...common,
+				id: `ctc_${state.responseId.slice(-18)}_${event.index}`,
 				type: "custom_tool_call",
 				input: ""
 			};
@@ -9910,6 +9914,7 @@ function* startContentBlock(state, event) {
 		}
 		const item = {
 			...common,
+			id: `fc_${state.responseId.slice(-18)}_${event.index}`,
 			type: "function_call",
 			arguments: ""
 		};
@@ -10523,18 +10528,22 @@ const handleResponses = async (c) => {
 	const normalizedReasoningEffort = normalizeResponsesReasoningEffort(payload, selectedModel?.capabilities?.supports?.reasoning_effort);
 	if (normalizedReasoningEffort) logger$1.debug(`Normalized reasoning effort from ${normalizedReasoningEffort.from} to ${normalizedReasoningEffort.to} based on the selected model capabilities`);
 	const responsesTransport = getResponsesTransportForModel(selectedModel);
-	if (shouldFallbackToMessages(c, payload.model, selectedModel, responsesTransport)) return await handleResponsesViaMessages(c, {
-		payload,
-		publicModel: requestedModel,
-		targetModel: payload.model,
-		subagentMarker,
-		requestId,
-		sessionId: fallbackSessionId
-	});
+	if (shouldFallbackToMessages(c, payload.model, selectedModel, responsesTransport)) {
+		filterReasoningForTransport(payload, true);
+		return await handleResponsesViaMessages(c, {
+			payload,
+			publicModel: requestedModel,
+			targetModel: payload.model,
+			subagentMarker,
+			requestId,
+			sessionId: fallbackSessionId
+		});
+	}
 	if (!responsesTransport) return c.json({ error: {
 		message: "This model does not support the responses endpoint. Please choose a different model.",
 		type: "invalid_request_error"
 	} }, 400);
+	filterReasoningForTransport(payload, false);
 	const recordUsage = createCopilotTokenUsageRecorder({
 		endpoint: "responses",
 		fallbackSessionId,
@@ -10605,6 +10614,13 @@ const handleResponses = async (c) => {
 	return c.json(result);
 };
 const isStreamingRequested = (payload) => Boolean(payload.stream);
+const filterReasoningForTransport = (payload, useMessagesFallback) => {
+	if (!Array.isArray(payload.input)) return;
+	payload.input = payload.input.filter((item) => {
+		if (item.type !== "reasoning") return true;
+		return isMessagesReasoningId(item.id) === useMessagesFallback;
+	});
+};
 const shouldFallbackToMessages = (c, modelId, selectedModel, responsesTransport) => {
 	if (isCodexUserAgent(c.req.header("user-agent"))) return !(modelId.startsWith("gpt") || modelId.startsWith("codex"));
 	if (responsesTransport) return false;
@@ -10792,4 +10808,4 @@ server.route("/:provider/images", providerImageRoutes);
 //#endregion
 export { server };
 
-//# sourceMappingURL=server-m80zRYma.js.map
+//# sourceMappingURL=server-fiF0jU-7.js.map