@dianshuv/copilot-api 0.7.0 → 0.7.2

package/README.md

@@ -32,6 +32,16 @@ copilot-api start
 ```
 
 
+## Development
+
+```sh
+# Start the server (foreground, production mode)
+make up
+
+# Stop the server (graceful shutdown)
+make down
+```
+
 ## Command Reference
 
 | Command | Description |

package/dist/main.mjs

@@ -1213,7 +1213,7 @@ const patchClaude = defineCommand({
 
 //#endregion
 //#region package.json
-var version = "0.7.0";
+var version = "0.7.2";
 
 //#endregion
 //#region src/lib/adaptive-rate-limiter.ts
@@ -3252,13 +3252,28 @@ const getTokenCount = async (payload, model) => {
 
 //#endregion
 //#region src/services/copilot/create-chat-completions.ts
+const GPT_MODEL_PATTERN = /^gpt-/i;
 const createChatCompletions = async (payload, options) => {
 	if (!state.copilotToken) throw new Error("Copilot token not found");
-	const enableVision = payload.messages.some((x) => typeof x.content !== "string" && x.content?.some((x) => x.type === "image_url"));
-	const isAgentCall = payload.messages.some((msg) => ["assistant", "tool"].includes(msg.role));
+	const vendor = options?.resolvedModel?.vendor;
+	const isOpenAIVendor = vendor === "OpenAI" || vendor === "Azure OpenAI";
+	const isLikelyGPT = !options?.resolvedModel && GPT_MODEL_PATTERN.test(payload.model);
+	let wire = payload;
+	if (isOpenAIVendor || isLikelyGPT) {
+		const { max_tokens, max_completion_tokens, ...rest } = payload;
+		const effective = max_completion_tokens ?? max_tokens;
+		wire = {
+			...rest,
+			...effective !== null && effective !== void 0 && { max_completion_tokens: effective }
+		};
+	}
+	const enableVision = wire.messages.some((x) => typeof x.content !== "string" && x.content?.some((x) => x.type === "image_url"));
+	const isAgentCall = wire.messages.some((msg) => ["assistant", "tool"].includes(msg.role));
+	const modelSupportsVision = options?.resolvedModel?.capabilities?.supports?.vision !== false;
 	const headers = {
 		...copilotHeaders(state, {
-			vision: enableVision,
+			vision: enableVision && modelSupportsVision,
+			modelRequestHeaders: options?.resolvedModel?.request_headers,
 			intent: isAgentCall ? "conversation-agent" : "conversation-panel"
 		}),
 		"X-Initiator": options?.initiator ?? (isAgentCall ? "agent" : "user")
@@ -3266,7 +3281,7 @@ const createChatCompletions = async (payload, options) => {
 	const response = await fetch(`${copilotBaseUrl(state)}/chat/completions`, {
 		method: "POST",
 		headers,
-		body: JSON.stringify(payload)
+		body: JSON.stringify(wire)
 	});
 	if (!response.ok) {
 		consola.error("Failed to create chat completions", response);
@@ -3846,7 +3861,7 @@ async function handleCompletion$1(c) {
 async function executeRequest(opts) {
 	const { c, payload, selectedModel, ctx, trackingId } = opts;
 	try {
-		const { result: response, queueWaitMs } = await executeWithAdaptiveRateLimit(() => createChatCompletions(payload));
+		const { result: response, queueWaitMs } = await executeWithAdaptiveRateLimit(() => createChatCompletions(payload, { resolvedModel: selectedModel }));
 		ctx.queueWaitMs = queueWaitMs;
 		if (isNonStreaming(response)) return handleNonStreamingResponse$1(c, response, ctx, payload);
 		consola.debug("Streaming response");
@@ -4503,7 +4518,7 @@ async function handleGeminiGenerate(c, model, isStream) {
 			trackingId,
 			startTime
 		};
-		const { result: response, queueWaitMs } = await executeWithAdaptiveRateLimit(() => createChatCompletions(payload));
+		const { result: response, queueWaitMs } = await executeWithAdaptiveRateLimit(() => createChatCompletions(payload, { resolvedModel: selectedModel }));
 		ctx.queueWaitMs = queueWaitMs;
 		if (isNonStreaming(response)) return handleNonStreamResponse(c, response, model, ctx, payload);
 		consola.debug("Streaming Gemini response");
@@ -4587,25 +4602,27 @@ function handleNonStreamResponse(c, response, model, ctx, payload) {
 //#endregion
 //#region src/routes/gemini/model-alias.ts
 /**
-* Maps Gemini model names that aren't available on GitHub Copilot
-* to equivalent models that are.
+* Maps Gemini model names to equivalent models available on GitHub Copilot.
 *
-* The Gemini CLI's routing classifier requests gemini-2.5-flash-lite
-* and gemini-2.5-flash, which Copilot doesn't serve. We map them to
-* the closest available flash model.
+* Two types of aliases:
 *
-* Aliases are only applied when the requested model is absent from
-* the Copilot model list, so if Copilot adds support for these models
-* natively, requests will go through unchanged.
+* - **Forced**: Always applied regardless of Copilot model availability.
+*   Use when the old model name should never reach the backend.
+*
+* - **Conditional**: Only applied when the requested model is absent from
+*   the Copilot model list, so if Copilot adds native support the request
+*   goes through unchanged.
 */
-const GEMINI_MODEL_ALIASES = {
+const GEMINI_FORCED_ALIASES = { "gemini-2.5-pro": "gemini-3.1-pro-preview" };
+const GEMINI_CONDITIONAL_ALIASES = {
 	"gemini-2.5-flash-lite": "gemini-3-flash-preview",
 	"gemini-2.5-flash": "gemini-3-flash-preview"
 };
 function resolveGeminiModelAlias(model) {
-	if (!(model in GEMINI_MODEL_ALIASES)) return model;
-	if (state.models?.data.some((m) => m.id === model)) return model;
-	return GEMINI_MODEL_ALIASES[model];
+	if (model in GEMINI_FORCED_ALIASES) return GEMINI_FORCED_ALIASES[model];
+	if (!(model in GEMINI_CONDITIONAL_ALIASES)) return model;
+	if (findModelById(model)) return model;
+	return GEMINI_CONDITIONAL_ALIASES[model];
 }
 
 //#endregion
@@ -7702,7 +7719,10 @@ async function handleTranslatedCompletion(c, anthropicPayload, ctx, initiatorOve
 	if (truncateResult) ctx.truncateResult = truncateResult;
 	if (state.manualApprove) await awaitApproval();
 	try {
-		const { result: response, queueWaitMs } = await executeWithAdaptiveRateLimit(() => createChatCompletions(openAIPayload, { initiator: initiatorOverride }));
+		const { result: response, queueWaitMs } = await executeWithAdaptiveRateLimit(() => createChatCompletions(openAIPayload, {
+			initiator: initiatorOverride,
+			resolvedModel: selectedModel
+		}));
 		ctx.queueWaitMs = queueWaitMs;
 		if (isNonStreaming(response)) return handleNonStreamingResponse({
 			c,
@@ -8056,10 +8076,14 @@ modelRoutes.get("/", async (c) => {
 
 //#endregion
 //#region src/services/copilot/create-responses.ts
-const createResponses = async (payload, { vision, initiator }) => {
+const createResponses = async (payload, { vision, initiator, resolvedModel }) => {
 	if (!state.copilotToken) throw new Error("Copilot token not found");
+	const modelSupportsVision = resolvedModel?.capabilities?.supports?.vision !== false;
 	const headers = {
-		...copilotHeaders(state, { vision }),
+		...copilotHeaders(state, {
+			vision: vision && modelSupportsVision,
+			modelRequestHeaders: resolvedModel?.request_headers
+		}),
 		"X-Initiator": initiator
 	};
 	payload.service_tier = null;
@@ -8322,7 +8346,8 @@ const handleResponses = async (c) => {
 		trackingId,
 		startTime
 	};
-	if (!(findModelById(payload.model)?.supported_endpoints?.includes(RESPONSES_ENDPOINT) ?? false)) {
+	const selectedModel = findModelById(payload.model);
+	if (!(selectedModel?.supported_endpoints?.includes(RESPONSES_ENDPOINT) ?? false)) {
 		recordErrorResponse(ctx, model, /* @__PURE__ */ new Error("This model does not support the responses endpoint."));
 		return c.json({ error: {
 			message: "This model does not support the responses endpoint. Please choose a different model.",
@@ -8334,7 +8359,8 @@ const handleResponses = async (c) => {
 	try {
 		const { result: response, queueWaitMs } = await executeWithAdaptiveRateLimit(() => createResponses(payload, {
 			vision,
-			initiator
+			initiator,
+			resolvedModel: selectedModel
 		}));
 		ctx.queueWaitMs = queueWaitMs;
 		if (isStreamingRequested(payload) && isAsyncIterable(response)) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@dianshuv/copilot-api",
-  "version": "0.7.0",
+  "version": "0.7.2",
   "description": "Turn GitHub Copilot into OpenAI/Anthropic API compatible server. Usable with Claude Code!",
   "author": "dianshuv",
   "type": "module",
@@ -20,7 +20,7 @@
     "prepare": "npm run build && (command -v bun >/dev/null 2>&1 && simple-git-hooks || true)",
     "prepublishOnly": "npm run typecheck && npm run lint:all && npm run test",
     "release": "npm publish --access public --//registry.npmjs.org/:_authToken=$NPM_TOKEN",
-    "start": "NODE_ENV=production bun run ./src/main.ts",
+    "start": "NODE_ENV=production bun run ./src/main.ts start",
     "test": "bun test tests/*.test.ts",
     "test:all": "bun test tests/*.test.ts && bun test tests/integration/",
     "test:integration": "bun test tests/integration/",