@kenkaiiii/gg-ai 4.3.204 → 4.3.206

package/dist/index.d.cts

@@ -157,7 +157,7 @@ interface StreamOptions {
     signal?: AbortSignal;
     /** Prompt cache retention preference. Providers map this to their supported values. Default: "short". */
     cacheRetention?: CacheRetention;
-    /** Stable per-session cache routing key for providers that support it (OpenAI, Moonshot). */
+    /** Stable per-session cache routing key for providers that support it (OpenAI, Moonshot, Gemini Code Assist). */
     promptCacheKey?: string;
     /** OpenAI service tier for latency-sensitive requests. Only sent to first-party OpenAI API calls. */
     serviceTier?: "auto" | "default" | "flex" | "priority";

package/dist/index.d.ts

@@ -157,7 +157,7 @@ interface StreamOptions {
     signal?: AbortSignal;
     /** Prompt cache retention preference. Providers map this to their supported values. Default: "short". */
     cacheRetention?: CacheRetention;
-    /** Stable per-session cache routing key for providers that support it (OpenAI, Moonshot). */
+    /** Stable per-session cache routing key for providers that support it (OpenAI, Moonshot, Gemini Code Assist). */
     promptCacheKey?: string;
     /** OpenAI service tier for latency-sensitive requests. Only sent to first-party OpenAI API calls. */
     serviceTier?: "auto" | "default" | "flex" | "priority";

package/dist/index.js

@@ -1932,6 +1932,8 @@ var DEFAULT_CODE_ASSIST_BASE_URL = "https://cloudcode-pa.googleapis.com";
 var CODE_ASSIST_API_VERSION = "v1internal";
 var GEMINI_CLI_USER_AGENT = "google-gemini-cli";
 var GEMINI_CLI_API_CLIENT = "gemini-cli/0.0.0";
+var CODE_ASSIST_NON_STREAMING_RETRIES = 3;
+var CODE_ASSIST_NON_STREAMING_RETRY_DELAY_MS = 1e3;
 var SYNTHETIC_THOUGHT_SIGNATURE = "skip_thought_signature_validator";
 var CODE_ASSIST_SUPPORTED_MODELS = /* @__PURE__ */ new Set([
   "gemini-3-pro-preview",
@@ -1956,7 +1958,10 @@ function getGoogleProject(options) {
   return options.projectId ?? env?.GOOGLE_CLOUD_PROJECT ?? env?.GOOGLE_CLOUD_PROJECT_ID;
 }
 function getCodeAssistEndpoint(method) {
-  return new URL(`${DEFAULT_CODE_ASSIST_BASE_URL}/${CODE_ASSIST_API_VERSION}:${method}`);
+  const env = getEnvironment();
+  const endpoint = env?.CODE_ASSIST_ENDPOINT ?? DEFAULT_CODE_ASSIST_BASE_URL;
+  const version = env?.CODE_ASSIST_API_VERSION || CODE_ASSIST_API_VERSION;
+  return new URL(`${endpoint}/${version}:${method}`);
 }
 function formatUnsupportedModelMessage(model) {
   return `Gemini OAuth is configured to use the Gemini Code Assist subscription endpoint only. That endpoint does not currently expose model "${model}".`;
@@ -2135,7 +2140,8 @@ function buildGenerateRequest(options) {
     ...systemInstruction ? { systemInstruction } : {},
     ...tools ? { tools } : {},
     ...toolConfig ? { toolConfig } : {},
-    ...Object.keys(generationConfig).length > 0 ? { generationConfig } : {}
+    ...Object.keys(generationConfig).length > 0 ? { generationConfig } : {},
+    ...options.promptCacheKey ? { session_id: options.promptCacheKey } : {}
   };
 }
 function buildCodeAssistRequest(options, request, projectId) {
@@ -2255,31 +2261,78 @@ function readFunctionCallPart(part) {
 function makeToolCallId(index, providerId) {
   return providerId ?? `gemini_call_${index}_${crypto.randomUUID().replace(/-/g, "")}`;
 }
-function streamGemini(options) {
-  return new StreamResult(runStream4(options));
+function shouldRetryCodeAssistStatus(status) {
+  return status === 429 || status === 499 || status >= 500 && status <= 599;
+}
+function isAbortError(err) {
+  return err instanceof Error && err.name === "AbortError";
+}
+async function sleep(ms, signal) {
+  if (ms <= 0) return;
+  await new Promise((resolve, reject) => {
+    const cleanup = () => signal?.removeEventListener("abort", onAbort);
+    const timer = setTimeout(() => {
+      cleanup();
+      resolve();
+    }, ms);
+    const onAbort = () => {
+      clearTimeout(timer);
+      cleanup();
+      reject(new DOMException("The operation was aborted.", "AbortError"));
+    };
+    signal?.addEventListener("abort", onAbort, { once: true });
+    if (signal?.aborted) onAbort();
+  });
 }
-async function* runStream4(options) {
-  const useStreaming = options.streaming !== false;
-  const method = useStreaming ? "streamGenerateContent" : "generateContent";
-  const plan = buildRequestPlan(options, method);
-  if (useStreaming) plan.url.searchParams.set("alt", "sse");
-  let response;
+async function fetchCodeAssist(plan, options) {
   try {
-    response = await fetch(plan.url, {
+    const response = await fetch(plan.url, {
       method: "POST",
       headers: plan.headers,
       body: JSON.stringify(plan.body),
       signal: options.signal
     });
+    if (!response.ok) {
+      const text = await response.text().catch(() => "");
+      throw new ProviderError("gemini", formatErrorMessage(response.status, text, options.model), {
+        statusCode: response.status
+      });
+    }
+    return response;
   } catch (err) {
     throw toError3(err);
   }
-  if (!response.ok) {
-    const text = await response.text().catch(() => "");
-    throw new ProviderError("gemini", formatErrorMessage(response.status, text, options.model), {
-      statusCode: response.status
-    });
+}
+async function fetchCodeAssistWithRetry(plan, options) {
+  let lastError;
+  for (let attempt = 0; attempt <= CODE_ASSIST_NON_STREAMING_RETRIES; attempt++) {
+    try {
+      return await fetchCodeAssist(plan, options);
+    } catch (err) {
+      const error = toError3(err);
+      const statusCode = error instanceof ProviderError ? error.statusCode : void 0;
+      if (options.signal?.aborted || isAbortError(error) || attempt === CODE_ASSIST_NON_STREAMING_RETRIES || statusCode != null && !shouldRetryCodeAssistStatus(statusCode)) {
+        throw error;
+      }
+      lastError = error;
+    }
+    try {
+      await sleep(CODE_ASSIST_NON_STREAMING_RETRY_DELAY_MS, options.signal);
+    } catch (err) {
+      throw toError3(err);
+    }
   }
+  throw lastError ?? new ProviderError("gemini", "Gemini Code Assist request failed.");
+}
+function streamGemini(options) {
+  return new StreamResult(runStream4(options));
+}
+async function* runStream4(options) {
+  const useStreaming = options.streaming !== false;
+  const method = useStreaming ? "streamGenerateContent" : "generateContent";
+  const plan = buildRequestPlan(options, method);
+  if (useStreaming) plan.url.searchParams.set("alt", "sse");
+  const response = useStreaming ? await fetchCodeAssist(plan, options) : await fetchCodeAssistWithRetry(plan, options);
   const contentParts = [];
   const pendingToolCalls = [];
   let textAccum = "";