@jeffreycao/copilot-api 1.9.7 → 1.9.10

package/README.md

@@ -328,7 +328,7 @@ The following command line options are available for the `start` command:
           "glm-5.1": {
             "temperature": 0.7,
             "topP": 0.95,
-            "contextCache": false,
+            "contextCache": true,
             "extraBody": {
               "preserve_thinking": true
             }

package/README.zh-CN.md

@@ -332,7 +332,7 @@ Copilot API 现在使用子命令结构，主要命令包括：
           "glm-5.1": {
             "temperature": 0.7,
             "topP": 0.95,
-            "contextCache": false,
+            "contextCache": true,
             "extraBody": {
               "preserve_thinking": true
             }

package/dist/main.js

@@ -44,7 +44,7 @@ bindElectronFetch();
 const { auth } = await import("./auth-CWEhhJYn.js");
 const { checkUsage } = await import("./check-usage-B5yr4fpk.js");
 const { debug } = await import("./debug-DcC7ZPH0.js");
-const { start } = await import("./start-BgL68ZLT.js");
+const { start } = await import("./start-CpqH2Ekm.js");
 const main = defineCommand({
 	meta: {
 		name: "copilot-api",

package/dist/{server-Bpu_LXg5.js → server-BG69Fgym.js}

@@ -1496,6 +1496,7 @@ function mapContent(content, options = {}) {
 			});
 			break;
 	}
+	if (contentParts.length === 0) return "";
 	return contentParts;
 }
 function createDocumentTextPart() {
@@ -3226,9 +3227,13 @@ function closeThinkingBlockIfOpen(state$1, events$1) {
 
 //#endregion
 //#region src/routes/messages/api-flows.ts
+const COPILOT_CONTEXT_CACHE_SYSTEM_MARKER_LIMIT = 2;
+const COPILOT_CONTEXT_CACHE_NON_SYSTEM_MARKER_LIMIT = 2;
+const COPILOT_CONTEXT_CACHE_CONTROL = { type: "ephemeral" };
 const handleWithChatCompletions = async (c, anthropicPayload, options) => {
 	const { logger: logger$7, subagentMarker, requestId, sessionId, compactType } = options;
 	const openAIPayload = translateToOpenAI(anthropicPayload);
+	prepareCopilotChatCompletionsPayload(openAIPayload);
 	const recordUsage = createCopilotUsageRecorder({
 		endpoint: "chat_completions",
 		fallbackSessionId: sessionId,
@@ -3400,6 +3405,26 @@ const handleWithMessagesApi = async (c, anthropicPayload, options) => {
 	recordUsage(normalizeAnthropicUsage(response.usage));
 	return c.json(response);
 };
+const prepareCopilotChatCompletionsPayload = (payload) => {
+	applyCopilotContextCache(payload);
+};
+const applyCopilotContextCache = (payload) => {
+	const messageIndexes = selectCopilotContextCacheMessageIndexes(payload.messages);
+	for (const messageIndex of messageIndexes) {
+		const message = payload.messages[messageIndex];
+		message.copilot_cache_control = { ...COPILOT_CONTEXT_CACHE_CONTROL };
+	}
+};
+const selectCopilotContextCacheMessageIndexes = (messages) => {
+	const systemIndexes = messages.flatMap((message, index) => message.role === "system" && isCopilotContextCacheEligible(message) ? [index] : []).slice(0, COPILOT_CONTEXT_CACHE_SYSTEM_MARKER_LIMIT);
+	const reverseNonSystemIndexes = messages.flatMap((message, index) => message.role !== "system" && isCopilotContextCacheEligible(message) ? [index] : []).reverse().slice(0, COPILOT_CONTEXT_CACHE_NON_SYSTEM_MARKER_LIMIT);
+	return uniqueIndexes$1([...systemIndexes, ...reverseNonSystemIndexes]).sort((a, b) => a - b);
+};
+const isCopilotContextCacheEligible = (message) => {
+	if (typeof message.content === "string") return message.content.length > 0;
+	return Array.isArray(message.content) && message.content.length > 0;
+};
+const uniqueIndexes$1 = (indexes) => [...new Set(indexes)];
 const isNonStreaming = (response) => Object.hasOwn(response, "choices");
 const isAsyncIterable$1 = (value) => Boolean(value) && typeof value[Symbol.asyncIterator] === "function";
 const createCopilotUsageRecorder = (options) => createCopilotTokenUsageRecorder({
@@ -3468,6 +3493,11 @@ const parseSubagentMarkerFromSystemReminder = (text) => {
 //#endregion
 //#region src/routes/messages/handler.ts
 const logger$5 = createHandlerLogger("messages-handler");
+const messagesFlowHandlers = {
+	handleWithChatCompletions,
+	handleWithMessagesApi,
+	handleWithResponsesApi
+};
 async function handleCompletion(c) {
 	await checkRateLimit(state);
 	const anthropicPayload = await c.req.json();
@@ -3490,7 +3520,7 @@ async function handleCompletion(c) {
 	if (state.manualApprove) await awaitApproval();
 	const selectedModel = findEndpointModel(anthropicPayload.model);
 	anthropicPayload.model = selectedModel?.id ?? anthropicPayload.model;
-	if (shouldUseMessagesApi(selectedModel)) return await handleWithMessagesApi(c, anthropicPayload, {
+	if (shouldUseMessagesApi(selectedModel)) return await messagesFlowHandlers.handleWithMessagesApi(c, anthropicPayload, {
 		anthropicBetaHeader: anthropicBeta,
 		subagentMarker,
 		selectedModel,
@@ -3499,7 +3529,7 @@ async function handleCompletion(c) {
 		compactType,
 		logger: logger$5
 	});
-	if (shouldUseResponsesApi(selectedModel)) return await handleWithResponsesApi(c, anthropicPayload, {
+	if (shouldUseResponsesApi(selectedModel)) return await messagesFlowHandlers.handleWithResponsesApi(c, anthropicPayload, {
 		subagentMarker,
 		selectedModel,
 		requestId,
@@ -3507,7 +3537,7 @@ async function handleCompletion(c) {
 		compactType,
 		logger: logger$5
 	});
-	return await handleWithChatCompletions(c, anthropicPayload, {
+	return await messagesFlowHandlers.handleWithChatCompletions(c, anthropicPayload, {
 		subagentMarker,
 		requestId,
 		sessionId,
@@ -4309,4 +4339,4 @@ server.route("/:provider/v1/models", providerModelRoutes);
 
 //#endregion
 export { server };
-//# sourceMappingURL=server-Bpu_LXg5.js.map
+//# sourceMappingURL=server-BG69Fgym.js.map