@jeffreycao/copilot-api 1.14.5 → 1.14.7

package/README.md

@@ -668,9 +668,15 @@ These endpoints mimic the OpenAI API structure.
 
 ### Codex Backend Proxy Endpoints
 
-| Endpoint             | Method | Description |
-| -------------------- | ------ | ----------- |
-| `POST /alpha/search` | `POST` | Transparently forwards the JSON body and query parameters to the Codex Alpha Search upstream. The gateway replaces client authorization and account headers with the active Codex login, forwards compatible headers such as `accept`, `content-type`, `originator`, `user-agent`, and `cookie`, and returns the upstream status, headers, and body unchanged. |
+These endpoints require an active Codex login. Each endpoint is available both without a version prefix and under `/v1`.
+
+| Endpoint | Method | Description |
+| --- | --- | --- |
+| `POST /alpha/search`<br>`POST /v1/alpha/search` | `POST` | Transparently forwards the JSON body and query parameters to the Codex Alpha Search upstream. |
+| `POST /images/generations`<br>`POST /v1/images/generations` | `POST` | Forwards a JSON image generation request to the Codex Images upstream. When the request omits `Content-Type`, the gateway defaults it to `application/json`. |
+| `POST /images/edits`<br>`POST /v1/images/edits` | `POST` | Forwards an image edit request to the Codex Images upstream. Send this request as `multipart/form-data` and let the HTTP client generate the `boundary`; the gateway preserves the incoming content type and streams the upload body. |
+
+For every endpoint above, the gateway replaces client authorization and account headers with the active Codex login, preserves query parameters and compatible request headers, and returns the upstream status, headers, and body.
 
 ### Anthropic Compatible Endpoints
 

package/README.zh-CN.md

@@ -674,9 +674,15 @@ curl http://localhost:4141/admin/config/model-mappings \
 
 ### Codex 后端代理端点
 
+这些端点要求已有可用的 Codex 登录态。每个端点同时提供无版本前缀和 `/v1` 两种路径。
+
 | 端点 | 方法 | 说明 |
 | --- | --- | --- |
-| `POST /alpha/search` | `POST` | 将 JSON 请求体和查询参数透明转发到 Codex Alpha Search 上游。网关会使用当前 Codex 登录态覆盖客户端的 authorization 和 account header，透传 `accept`、`content-type`、`originator`、`user-agent`、`cookie` 等兼容 header，并原样返回上游状态码、响应头和响应体。 |
+| `POST /alpha/search`<br>`POST /v1/alpha/search` | `POST` | 将 JSON 请求体和查询参数透明转发到 Codex Alpha Search 上游。 |
+| `POST /images/generations`<br>`POST /v1/images/generations` | `POST` | 将 JSON 图片生成请求转发到 Codex Images 上游。请求未携带 `Content-Type` 时，网关默认补充 `application/json`。 |
+| `POST /images/edits`<br>`POST /v1/images/edits` | `POST` | 将图片编辑请求转发到 Codex Images 上游。请使用 `multipart/form-data`，并让 HTTP 客户端自动生成 `boundary`；网关会保留传入的 content type，并以流式方式转发上传请求体。 |
+
+对于以上所有端点，网关都会使用当前 Codex 登录态覆盖客户端的 authorization 和 account header，保留查询参数及兼容的请求头，并返回上游状态码、响应头和响应体。
 
 ### Anthropic 兼容端点
 

package/dist/main.js

@@ -25,7 +25,7 @@ bindElectronFetch();
 const { auth } = await import("./auth-DvJ8r_tN.js");
 const { debug } = await import("./debug-Bb5hamXX.js");
 const { mcp } = await import("./mcp-BseuqgHR.js");
-const { start } = await import("./start-Beq7gtGI.js");
+const { start } = await import("./start-juT2yVlW.js");
 await runMain(defineCommand({
 	meta: {
 		name: "copilot-api",

package/dist/{server-CmynIOAW.js → server-CF0okzl9.js}

@@ -7,6 +7,7 @@ import fs, { readFileSync } from "node:fs";
 import fs$1 from "node:fs/promises";
 import path from "node:path";
 import { events } from "fetch-event-stream";
+import { fetch as fetch$1, getGlobalDispatcher } from "undici";
 import { z } from "zod";
 import { Hono } from "hono";
 import { cors } from "hono/cors";
@@ -299,6 +300,10 @@ const debugLazy = (logger, factory) => {
 const debugJson = (logger, label, value) => {
 	debugLazy(logger, () => [label, JSON.stringify(value)]);
 };
+const debugJsonAsync = async (logger, label, factory) => {
+	if (!state.verbose) return;
+	logger.debug(label, JSON.stringify(await factory()));
+};
 const debugJsonTail = (logger, label, { value, tailLength = 400 }) => {
 	debugLazy(logger, () => [label, JSON.stringify(value).slice(-tailLength)]);
 };
@@ -441,7 +446,7 @@ async function forwardProviderModels(providerConfig, requestHeaders) {
 }
 //#endregion
 //#region src/routes/alpha-search/route.ts
-const logger$10 = createHandlerLogger("alpha-search-handler");
+const logger$11 = createHandlerLogger("alpha-search-handler");
 const alphaSearchRoutes = new Hono();
 alphaSearchRoutes.post("/", async (c) => {
 	try {
@@ -449,15 +454,15 @@ alphaSearchRoutes.post("/", async (c) => {
 			message: "Provider 'codex' not found or disabled",
 			type: "invalid_request_error"
 		} }, 404);
-		debugJson(logger$10, "alpha_search.codex.request", { body: await c.req.raw.clone().text() });
+		await debugJsonAsync(logger$11, "alpha_search.codex.request", async () => ({ body: await c.req.raw.clone().text() }));
 		const upstreamResponse = await forwardCodexAlphaSearch(c.req.raw);
-		debugJson(logger$10, "alpha_search.codex.response", {
+		await debugJsonAsync(logger$11, "alpha_search.codex.response", async () => ({
 			body: await upstreamResponse.clone().text(),
 			statusCode: upstreamResponse.status
-		});
+		}));
 		return createProviderProxyResponse(upstreamResponse);
 	} catch (error) {
-		logger$10.error("alpha_search.codex.error", { error });
+		logger$11.error("alpha_search.codex.error", { error });
 		return await forwardError(c, error);
 	}
 });
@@ -1589,7 +1594,7 @@ const setContextCacheControl = (part) => {
 };
 //#endregion
 //#region src/routes/provider/chat-completions/handler.ts
-const logger$9 = createHandlerLogger("provider-chat-completions-handler");
+const logger$10 = createHandlerLogger("provider-chat-completions-handler");
 async function handleProviderChatCompletionsForProvider(c, options) {
 	const { payload, provider } = options;
 	const providerConfig = await resolveProviderConfig(provider);
@@ -1603,13 +1608,13 @@ async function handleProviderChatCompletionsForProvider(c, options) {
 	applyProviderStreamOptions(payload);
 	applyDashScopePreserveThinkingDefault(payload, providerConfig);
 	applyProviderContextCache(payload, modelConfig, providerConfig);
-	debugJson(logger$9, "provider.chat_completions.request", {
+	debugJson(logger$10, "provider.chat_completions.request", {
 		payload,
 		provider
 	});
 	const upstreamResponse = await forwardProviderChatCompletions(providerConfig, payload, c.req.raw.headers);
 	if (!upstreamResponse.ok) {
-		logger$9.error("Failed to create provider chat completions", {
+		logger$10.error("Failed to create provider chat completions", {
 			provider,
 			statusCode: upstreamResponse.status
 		});
@@ -1623,7 +1628,7 @@ async function handleProviderChatCompletionsForProvider(c, options) {
 	});
 	const responseBody = await upstreamResponse.clone().json();
 	recordUsage(normalizeOpenAIUsage(responseBody.usage));
-	debugJson(logger$9, "provider.chat_completions.response", responseBody);
+	debugJson(logger$10, "provider.chat_completions.response", responseBody);
 	return createProviderProxyResponse(upstreamResponse);
 }
 const applyProviderModelDefaults = (payload, modelConfig) => {
@@ -1653,12 +1658,12 @@ const createProviderChatCompletionsUsageRecorder = (payload, provider, modelConf
 	providerName: provider
 });
 const streamProviderChatCompletions = (c, upstreamResponse, options) => {
-	logger$9.debug("provider.chat_completions.streaming", { provider: options.provider });
+	logger$10.debug("provider.chat_completions.streaming", { provider: options.provider });
 	return streamSSE(c, async (stream) => {
 		let usage = {};
 		try {
 			for await (const chunk of events(upstreamResponse)) {
-				debugJson(logger$9, "provider.chat_completions.stream_chunk", chunk);
+				debugJson(logger$10, "provider.chat_completions.stream_chunk", chunk);
 				if (chunk.data && chunk.data !== "[DONE]") {
 					const parsedChunk = parseChatCompletionChunkData(chunk.data);
 					if (parsedChunk?.usage) usage = normalizeOpenAIUsage(parsedChunk.usage);
@@ -1785,7 +1790,7 @@ const createChatCompletions = async (payload, options) => {
 };
 //#endregion
 //#region src/routes/chat-completions/handler.ts
-const logger$8 = createHandlerLogger("chat-completions-handler");
+const logger$9 = createHandlerLogger("chat-completions-handler");
 async function handleCompletion$1(c) {
 	let payload = await c.req.json();
 	const requestedModel = payload.model;
@@ -1799,23 +1804,23 @@ async function handleCompletion$1(c) {
 			provider: providerModelAlias.provider
 		});
 	}
-	debugJson(logger$8, "Request payload:", payload);
+	debugJson(logger$9, "Request payload:", payload);
 	const selectedModel = state.models?.data.find((model) => model.id === payload.model);
 	if (isNullish(payload.max_tokens) && isNullish(payload.max_completion_tokens)) {
 		payload = {
 			...payload,
 			max_tokens: selectedModel?.capabilities.limits.max_output_tokens
 		};
-		debugJson(logger$8, "Set max_tokens to:", payload.max_tokens);
+		debugJson(logger$9, "Set max_tokens to:", payload.max_tokens);
 	}
 	if (payload.model.includes("gpt")) {
 		if (isNullish(payload.max_completion_tokens)) payload.max_completion_tokens = payload.max_tokens;
 		delete payload.max_tokens;
 	}
 	const requestId = generateRequestIdFromPayload(payload);
-	logger$8.debug("Generated request ID:", requestId);
+	logger$9.debug("Generated request ID:", requestId);
 	const sessionId = getUUID(requestId);
-	logger$8.debug("Extracted session ID:", sessionId);
+	logger$9.debug("Extracted session ID:", sessionId);
 	const recordUsage = createCopilotTokenUsageRecorder({
 		endpoint: "chat_completions",
 		fallbackSessionId: sessionId,
@@ -1826,18 +1831,18 @@ async function handleCompletion$1(c) {
 		sessionId
 	});
 	if (isNonStreaming$1(response)) {
-		debugJson(logger$8, "Non-streaming response:", response);
+		debugJson(logger$9, "Non-streaming response:", response);
 		recordUsage({
 			...normalizeOpenAIUsage(response.usage),
 			total_nano_aiu: normalizeOptionalToken(response.copilot_usage?.total_nano_aiu)
 		});
 		return c.json(response);
 	}
-	logger$8.debug("Streaming response");
+	logger$9.debug("Streaming response");
 	return streamSSE(c, async (stream) => {
 		let usage = {};
 		for await (const chunk of response) {
-			debugJson(logger$8, "Streaming chunk:", chunk);
+			debugJson(logger$9, "Streaming chunk:", chunk);
 			const parsedChunk = parseChatCompletionChunk(chunk);
 			if (parsedChunk?.usage || parsedChunk?.copilot_usage) usage = {
 				...normalizeOpenAIUsage(parsedChunk.usage),
@@ -1926,6 +1931,71 @@ embeddingRoutes.post("/", async (c) => {
 	}
 });
 //#endregion
+//#region src/services/codex/images.ts
+const CODEX_IMAGES_TIMEOUT_MS = 900 * 1e3;
+const codexImagesDispatcher = { dispatch(options, handler) {
+	return getGlobalDispatcher().dispatch({
+		...options,
+		bodyTimeout: CODEX_IMAGES_TIMEOUT_MS,
+		headersTimeout: CODEX_IMAGES_TIMEOUT_MS
+	}, handler);
+} };
+function resolveCodexImagesUrl(requestUrl, operation) {
+	const upstreamUrl = new URL(`${CODEX_API_BASE_URL}/codex/images/${operation}`);
+	upstreamUrl.search = new URL(requestUrl, "http://localhost").search;
+	return upstreamUrl.toString();
+}
+async function forwardCodexImages(request, operation) {
+	const headers = buildCodexRequestHeaders(request.headers);
+	if (!headers.has("accept")) headers.set("accept", "application/json");
+	if (operation === "generations" && !headers.has("content-type")) headers.set("content-type", "application/json");
+	const init = {
+		method: "POST",
+		headers,
+		body: request.body,
+		duplex: "half",
+		signal: AbortSignal.timeout(CODEX_IMAGES_TIMEOUT_MS)
+	};
+	const upstreamUrl = resolveCodexImagesUrl(request.url, operation);
+	if (typeof Bun !== "undefined") return await fetch(upstreamUrl, init);
+	return await fetch$1(upstreamUrl, {
+		...init,
+		dispatcher: codexImagesDispatcher
+	});
+}
+//#endregion
+//#region src/routes/images/route.ts
+const logger$8 = createHandlerLogger("images-handler");
+const imageRoutes = new Hono();
+const imageRouteDependencies = { debugJsonAsync };
+function getContentMetadata(headers) {
+	return {
+		contentType: headers.get("content-type"),
+		contentLength: headers.get("content-length")
+	};
+}
+async function handleCodexImages(c, operation) {
+	try {
+		if (!await resolveProviderConfig("codex")) return c.json({ error: {
+			message: "Provider 'codex' not found or disabled",
+			type: "invalid_request_error"
+		} }, 404);
+		if (operation === "generations") await imageRouteDependencies.debugJsonAsync(logger$8, "images.generations.codex.request", async () => ({ body: await c.req.raw.clone().text() }));
+		else debugJson(logger$8, "images.edits.codex.request", getContentMetadata(c.req.raw.headers));
+		const upstreamResponse = await forwardCodexImages(c.req.raw, operation);
+		debugJson(logger$8, `images.${operation}.codex.response`, {
+			...getContentMetadata(upstreamResponse.headers),
+			statusCode: upstreamResponse.status
+		});
+		return createProviderProxyResponse(upstreamResponse);
+	} catch (error) {
+		logger$8.error(`images.${operation}.codex.error`, { error });
+		return await forwardError(c, error);
+	}
+}
+imageRoutes.post("/generations", (c) => handleCodexImages(c, "generations"));
+imageRoutes.post("/edits", (c) => handleCodexImages(c, "edits"));
+//#endregion
 //#region src/lib/tokenizer.ts
 const ENCODING_MAP = {
 	o200k_base: () => import("gpt-tokenizer/encoding/o200k_base"),
@@ -3568,6 +3638,7 @@ const MESSAGE_TYPE = "message";
 const COMPACTION_SIGNATURE_PREFIX = "cm1#";
 const COMPACTION_SIGNATURE_SEPARATOR = "@";
 const THINKING_TEXT = "Thinking...";
+const REASONING_SUMMARY_SEPARATOR = "⁣\n\n";
 const resolveReasoningEffort = (payload) => payload.output_config?.effort ?? getReasoningEffortForModel(payload.model);
 const buildPromptCacheKey = (basePromptCacheKey, subagentAgentId) => {
 	if (!basePromptCacheKey) return null;
@@ -3609,7 +3680,7 @@ const translateAnthropicMessagesToResponsesPayload = (payload, subagentAgentId)
 		parallel_tool_calls: true,
 		reasoning: {
 			effort: resolveReasoningEffort(payload),
-			summary: "detailed",
+			summary: "auto",
 			context: isSupportAllTurns(payload) ? "all_turns" : "auto"
 		},
 		include: ["reasoning.encrypted_content"]
@@ -3755,17 +3826,24 @@ const createFileContent = (block) => ({
 });
 const createReasoningContent = (block) => {
 	const { encryptedContent, id } = parseReasoningSignature(block.signature);
-	const thinking = block.thinking === "Thinking..." ? "" : block.thinking;
 	return {
 		id,
 		type: "reasoning",
-		summary: thinking ? [{
-			type: "summary_text",
-			text: thinking
-		}] : [],
+		summary: createReasoningSummary(block.thinking === "Thinking..." ? "" : block.thinking),
 		encrypted_content: encryptedContent
 	};
 };
+const createReasoningSummary = (thinking) => {
+	if (thinking.length === 0) return [];
+	if (!thinking.includes("⁣\n\n")) return [{
+		type: "summary_text",
+		text: thinking
+	}];
+	return thinking.split(REASONING_SUMMARY_SEPARATOR).map((text) => ({
+		type: "summary_text",
+		text
+	}));
+};
 const createCompactionContent = (block) => {
 	const compaction = decodeCompactionCarrierSignature(block.signature);
 	if (!compaction) return;
@@ -4042,7 +4120,7 @@ const extractReasoningText = (item) => {
 	};
 	if (!item.summary || item.summary.length === 0) return THINKING_TEXT;
 	collectFromBlocks(item.summary);
-	return segments.join("").trim();
+	return segments.join(REASONING_SUMMARY_SEPARATOR).trim();
 };
 const createToolUseContentBlock = (call) => {
 	const toolId = call.call_id;
@@ -4193,6 +4271,7 @@ const createResponsesStreamState = (options) => ({
 	blockIndexByKey: /* @__PURE__ */ new Map(),
 	openBlocks: /* @__PURE__ */ new Set(),
 	blockHasDelta: /* @__PURE__ */ new Set(),
+	reasoningSummaryHasDelta: /* @__PURE__ */ new Set(),
 	functionCallStateByOutputIndex: /* @__PURE__ */ new Map(),
 	toolSearchName: options?.toolSearchName ?? "mcp__tool_search__search",
 	hasToolCall: false
@@ -4202,6 +4281,7 @@ const translateResponsesStreamEvent = (rawEvent, state) => {
 		case "response.created": return handleResponseCreated(rawEvent, state);
 		case "response.output_item.added": return handleOutputItemAdded$1(rawEvent, state);
 		case "response.reasoning_summary_text.delta": return handleReasoningSummaryTextDelta(rawEvent, state);
+		case "response.reasoning_summary_part.added": return handleReasoningSummaryPartAdded(rawEvent, state);
 		case "response.output_text.delta": return handleOutputTextDelta(rawEvent, state);
 		case "response.reasoning_summary_text.done": return handleReasoningSummaryTextDone(rawEvent, state);
 		case "response.output_text.done": return handleOutputTextDone(rawEvent, state);
@@ -4298,7 +4378,7 @@ const handleOutputItemDone$1 = (rawEvent, state) => {
 	const blockIndex = openThinkingBlockIfNeeded(state, outputIndex, events);
 	const signature = (item.encrypted_content ?? "") + "@" + item.id;
 	if (signature) {
-		if (!item.summary || item.summary.length === 0) events.push({
+		if ((!item.summary || item.summary.length === 0) && !state.blockHasDelta.has(blockIndex)) events.push({
 			type: "content_block_delta",
 			index: blockIndex,
 			delta: {
@@ -4389,8 +4469,10 @@ const handleOutputTextDelta = (rawEvent, state) => {
 };
 const handleReasoningSummaryTextDelta = (rawEvent, state) => {
 	const outputIndex = rawEvent.output_index;
+	const summaryIndex = rawEvent.summary_index;
 	const deltaText = rawEvent.delta;
 	const events = new Array();
+	if (!deltaText) return events;
 	const blockIndex = openThinkingBlockIfNeeded(state, outputIndex, events);
 	events.push({
 		type: "content_block_delta",
@@ -4401,21 +4483,42 @@ const handleReasoningSummaryTextDelta = (rawEvent, state) => {
 		}
 	});
 	state.blockHasDelta.add(blockIndex);
+	state.reasoningSummaryHasDelta.add(getBlockKey(outputIndex, summaryIndex));
 	return events;
 };
-const handleReasoningSummaryTextDone = (rawEvent, state) => {
-	const outputIndex = rawEvent.output_index;
-	const text = rawEvent.text;
+const handleReasoningSummaryPartAdded = (rawEvent, state) => {
 	const events = new Array();
-	const blockIndex = openThinkingBlockIfNeeded(state, outputIndex, events);
-	if (text && !state.blockHasDelta.has(blockIndex)) events.push({
+	const blockIndex = openThinkingBlockIfNeeded(state, rawEvent.output_index, events);
+	if (rawEvent.summary_index === 0) return events;
+	events.push({
 		type: "content_block_delta",
 		index: blockIndex,
 		delta: {
 			type: "thinking_delta",
-			thinking: text
+			thinking: REASONING_SUMMARY_SEPARATOR
 		}
 	});
+	state.blockHasDelta.add(blockIndex);
+	return events;
+};
+const handleReasoningSummaryTextDone = (rawEvent, state) => {
+	const outputIndex = rawEvent.output_index;
+	const summaryIndex = rawEvent.summary_index;
+	const text = rawEvent.text;
+	const events = new Array();
+	const blockIndex = openThinkingBlockIfNeeded(state, outputIndex, events);
+	const summaryKey = getBlockKey(outputIndex, summaryIndex);
+	if (text && !state.reasoningSummaryHasDelta.has(summaryKey)) {
+		events.push({
+			type: "content_block_delta",
+			index: blockIndex,
+			delta: {
+				type: "thinking_delta",
+				thinking: text
+			}
+		});
+		state.blockHasDelta.add(blockIndex);
+	}
 	return events;
 };
 const handleOutputTextDone = (rawEvent, state) => {
@@ -4562,6 +4665,7 @@ const closeOpenBlocks = (state, events) => {
 const closeAllOpenBlocks = (state, events) => {
 	closeOpenBlocks(state, events);
 	state.functionCallStateByOutputIndex.clear();
+	state.reasoningSummaryHasDelta.clear();
 };
 const buildErrorEvent = (message) => ({
 	type: "error",
@@ -6876,14 +6980,17 @@ server.route("/token-usage", tokenUsageRoute);
 server.route("/token", tokenRoute);
 server.route("/responses", responsesRoutes);
 server.route("/alpha/search", alphaSearchRoutes);
+server.route("/images", imageRoutes);
 server.route("/v1/chat/completions", completionRoutes);
 server.route("/v1/models", modelRoutes);
 server.route("/v1/embeddings", embeddingRoutes);
 server.route("/v1/responses", responsesRoutes);
+server.route("/v1/alpha/search", alphaSearchRoutes);
+server.route("/v1/images", imageRoutes);
 server.route("/v1/messages", messageRoutes);
 server.route("/:provider/v1/messages", providerMessageRoutes);
 server.route("/:provider/v1/models", providerModelRoutes);
 //#endregion
 export { server };
 
-//# sourceMappingURL=server-CmynIOAW.js.map
+//# sourceMappingURL=server-CF0okzl9.js.map