@jeffreycao/copilot-api 1.12.7 → 1.12.8

package/dist/{server-ypsjdIs5.js → server-DQXpAOIW.js}

@@ -1,12 +1,12 @@
-import { S as PATHS, a as getMessageApiWebSearchModel, b as setModelMappings, c as getProviderConfig, d as getSmallModel, g as isResponsesApiWebSocketEnabled, h as isResponsesApiWebSearchEnabled, i as getExtraPromptForModel, l as getRawProviderConfig, m as isResponsesApiContextManagementEnabled, n as getClaudeTokenMultiplier, o as getModelMappings, p as isMessagesApiEnabled, r as getConfig, s as getModelResponsesApiCompactThreshold$1, t as getAnthropicApiKey, u as getReasoningEffortForModel, y as resolveMappedModel } from "./config-B-_X-1gj.js";
-import { C as prepareMessageProxyHeaders, E as compactMessageSections, F as createPooledWebSocketStream, I as createWebSocketUrl, L as state, M as generateTraceId, N as requestContext, O as compactSystemPromptStarts, P as resolveTraceId$1, R as HTTPError, S as prepareInteractionHeaders, T as compactAutoContinuePromptStarts, _ as getCopilotUsage, b as copilotWebSocketHeaders, d as generateRequestIdFromPayload, f as getRootSessionId, g as sleep, h as parseUserIdMetadata, j as forwardCodexResponses, m as isNullish, p as getUUID, r as setupCodexToken, s as cacheModels, v as copilotBaseUrl, x as prepareForCompact, y as copilotHeaders, z as forwardError } from "./token-CIpQTF2M.js";
+import { C as setModelMappings, S as resolveMappedModel, T as PATHS, _ as isResponsesApiWebSocketEnabled, a as getExtraPromptForModel, c as getModelResponsesApiCompactThreshold$1, d as getReasoningEffortForModel, f as getSmallModel, g as isResponsesApiWebSearchEnabled, h as isResponsesApiContextManagementEnabled, i as getConfig, l as getProviderConfig, m as isMessagesApiEnabled, n as getAnthropicApiKey, o as getMessageApiWebSearchModel, r as getClaudeTokenMultiplier, s as getModelMappings, u as getRawProviderConfig } from "./config-B71_S_hQ.js";
+import { C as prepareMessageProxyHeaders, E as compactMessageSections, F as createPooledWebSocketStream, I as createWebSocketUrl, L as state, M as generateTraceId, N as requestContext, O as compactSystemPromptStarts, P as resolveTraceId$1, R as HTTPError, S as prepareInteractionHeaders, T as compactAutoContinuePromptStarts, _ as getCopilotUsage, b as copilotWebSocketHeaders, d as generateRequestIdFromPayload, f as getRootSessionId, g as sleep, h as parseUserIdMetadata, j as forwardCodexResponses, m as isNullish, p as getUUID, r as setupCodexToken, s as cacheModels, v as copilotBaseUrl, x as prepareForCompact, y as copilotHeaders, z as forwardError } from "./token-fpw8tAQS.js";
 import { a as isDeferredToolName, c as parseMcpToolSearchSentinel, d as shouldEnableResponsesToolSearch, i as isBridgeToolSearchName, l as resolveBridgeToolSearchName, o as listDeferredToolNames, r as formatToolSearchBridgeArguments, s as normalizeToolSearchBridgeArguments, t as BRIDGE_TOOL_SEARCH_NAME, u as selectDeferredToolsByNames } from "./tool-search-D3SN0jX-.js";
 import consola from "consola";
 import { createHash } from "node:crypto";
-import { events } from "fetch-event-stream";
 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 { z } from "zod";
 import { Hono } from "hono";
 import { cors } from "hono/cors";
@@ -1309,18 +1309,18 @@ async function handleCompletion$1(c) {
 	await checkRateLimit(state);
 	debugJson(logger$7, "Request payload:", payload);
 	const selectedModel = state.models?.data.find((model) => model.id === payload.model);
-	if (selectedModel?.id === "gpt-5.4") return c.json({ error: {
-		message: "Please use `/v1/responses` or `/v1/messages` API",
-		type: "invalid_request_error"
-	} }, 400);
 	if (state.manualApprove) await awaitApproval();
-	if (isNullish(payload.max_tokens)) {
+	if (isNullish(payload.max_tokens) && isNullish(payload.max_completion_tokens)) {
 		payload = {
 			...payload,
 			max_tokens: selectedModel?.capabilities.limits.max_output_tokens
 		};
 		debugJson(logger$7, "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$7.debug("Generated request ID:", requestId);
 	const sessionId = getUUID(requestId);
@@ -1664,6 +1664,7 @@ function translateToOpenAI(payload, options = {}) {
 	const modelId = payload.model;
 	const model = state.models?.data.find((m) => m.id === modelId);
 	const thinkingBudget = getThinkingBudget(payload, model);
+	const reasoningEffort = getReasoningEffort(payload, options);
 	return {
 		model: modelId,
 		messages: translateAnthropicMessagesToOpenAI(payload, modelId, {
@@ -1678,12 +1679,22 @@ function translateToOpenAI(payload, options = {}) {
 		user: payload.metadata?.user_id,
 		tools: translateAnthropicToolsToOpenAI(payload.tools),
 		tool_choice: translateAnthropicToolChoiceToOpenAI(payload.tool_choice),
-		thinking_budget: thinkingBudget
+		thinking_budget: thinkingBudget,
+		...reasoningEffort ? { reasoning_effort: reasoningEffort } : {}
 	};
 }
+function getReasoningEffort(payload, options) {
+	const effort = payload.output_config?.effort;
+	if (!effort) return;
+	if (!options.validateReasoningEffort) return effort;
+	const supportedEfforts = options.reasoningEffortSupport;
+	if (!supportedEfforts || supportedEfforts.length === 0) return;
+	if (supportedEfforts.includes(effort)) return effort;
+	return supportedEfforts.at(-1);
+}
 function getThinkingBudget(payload, model) {
 	const thinking = payload.thinking;
-	if (model && thinking) {
+	if (model && thinking?.budget_tokens !== void 0) {
 		const maxThinkingBudget = Math.min(model.capabilities.supports.max_thinking_budget ?? 0, (model.capabilities.limits.max_output_tokens ?? 0) - 1);
 		thinking.budget_tokens ??= maxThinkingBudget;
 		if (maxThinkingBudget > 0) {
@@ -4917,14 +4928,12 @@ async function handleProviderMessagesForProvider(c, options) {
 			c,
 			payload,
 			provider,
-			providerConfig,
 			upstreamResponse
 		});
 		return respondProviderMessagesJson(c, {
 			body: await upstreamResponse.json(),
 			payload,
-			provider,
-			providerConfig
+			provider
 		});
 	} catch (error) {
 		logger$5.error("provider.messages.error", {
@@ -5149,7 +5158,7 @@ const applyContextCacheControl = (message) => {
 const setContextCacheControl = (part) => {
 	part.cache_control = { ...OPENAI_COMPATIBLE_CONTEXT_CACHE_CONTROL };
 };
-const streamProviderMessages = ({ c, payload, provider, providerConfig, upstreamResponse }) => {
+const streamProviderMessages = ({ c, payload, provider, upstreamResponse }) => {
 	logger$5.debug("provider.messages.streaming");
 	const recordUsage = createProviderMessagesUsageRecorder(payload, provider);
 	return streamSSE(c, async (stream) => {
@@ -5167,7 +5176,7 @@ const streamProviderMessages = ({ c, payload, provider, providerConfig, upstream
 			let data = chunk.data;
 			if (!data) continue;
 			if (chunk.data === "[DONE]") break;
-			const parsed = parseProviderStreamEvent(data, providerConfig);
+			const parsed = parseProviderStreamEvent(data);
 			if (parsed) {
 				usage = mergeAnthropicUsage(usage, parsed.usage);
 				data = parsed.data;
@@ -5299,24 +5308,18 @@ const parseResponsesProviderStreamChunk = (data, providerConfig) => {
 		return null;
 	}
 };
-const parseProviderStreamEvent = (data, providerConfig) => {
+const parseProviderStreamEvent = (data) => {
 	try {
 		const parsed = JSON.parse(data);
-		if (parsed.type === "message_start") {
-			adjustInputTokens(providerConfig, parsed.message.usage);
-			return {
-				data: JSON.stringify(parsed),
-				model: parsed.message.model,
-				usage: normalizeAnthropicUsage(parsed.message.usage)
-			};
-		}
-		if (parsed.type === "message_delta") {
-			adjustInputTokens(providerConfig, parsed.usage);
-			return {
-				data: JSON.stringify(parsed),
-				usage: normalizeAnthropicUsage(parsed.usage)
-			};
-		}
+		if (parsed.type === "message_start") return {
+			data: JSON.stringify(parsed),
+			model: parsed.message.model,
+			usage: normalizeAnthropicUsage(parsed.message.usage)
+		};
+		if (parsed.type === "message_delta") return {
+			data: JSON.stringify(parsed),
+			usage: normalizeAnthropicUsage(parsed.usage)
+		};
 		return {
 			data: JSON.stringify(parsed),
 			usage: {}
@@ -5330,10 +5333,8 @@ const parseProviderStreamEvent = (data, providerConfig) => {
 	}
 };
 const respondProviderMessagesJson = (c, options) => {
-	const { body, payload, provider, providerConfig } = options;
-	const recordUsage = createProviderMessagesUsageRecorder(payload, provider);
-	adjustInputTokens(providerConfig, body.usage);
-	recordUsage(normalizeAnthropicUsage(body.usage));
+	const { body, payload, provider } = options;
+	createProviderMessagesUsageRecorder(payload, provider)(normalizeAnthropicUsage(body.usage));
 	debugJson(logger$5, "provider.messages.no_stream result:", body);
 	return c.json(body);
 };
@@ -5371,11 +5372,6 @@ const createProviderMessagesUsageRecorder = (payload, provider) => createProvide
 	providerName: provider,
 	sessionId: parseUserIdMetadata(payload.metadata?.user_id).sessionId
 });
-const adjustInputTokens = (providerConfig, usage) => {
-	if (!providerConfig.adjustInputTokens || !usage) return;
-	usage.input_tokens = Math.max(0, (usage.input_tokens ?? 0) - (usage.cache_read_input_tokens ?? 0) - (usage.cache_creation_input_tokens ?? 0));
-	debugJson(logger$5, "provider.messages.adjusted_usage:", usage);
-};
 //#endregion
 //#region src/services/copilot/create-messages.ts
 const INTERLEAVED_THINKING_BETA = "interleaved-thinking-2025-05-14";
@@ -5437,8 +5433,11 @@ const messagesApiFlowDependencies = {
 	createResponses
 };
 const handleWithChatCompletions = async (c, anthropicPayload, options) => {
-	const { logger, subagentMarker, requestId, sessionId, compactType } = options;
-	const openAIPayload = translateToOpenAI(anthropicPayload);
+	const { logger, selectedModel, subagentMarker, requestId, sessionId, compactType } = options;
+	const openAIPayload = translateToOpenAI(anthropicPayload, {
+		validateReasoningEffort: true,
+		reasoningEffortSupport: selectedModel?.capabilities.supports.reasoning_effort
+	});
 	prepareCopilotChatCompletionsPayload(openAIPayload);
 	const recordUsage = createCopilotUsageRecorder({
 		endpoint: "chat_completions",
@@ -5723,6 +5722,7 @@ async function handleCompletion(c) {
 	});
 	return await messagesFlowHandlers.handleWithChatCompletions(c, anthropicPayload, {
 		subagentMarker,
+		selectedModel,
 		requestId,
 		sessionId,
 		compactType,
@@ -6254,4 +6254,4 @@ server.route("/:provider/v1/models", providerModelRoutes);
 //#endregion
 export { server };
 
-//# sourceMappingURL=server-ypsjdIs5.js.map
+//# sourceMappingURL=server-DQXpAOIW.js.map