@juspay/neurolink 9.23.0 → 9.25.0

package/dist/lib/adapters/tts/googleTTSHandler.js

@@ -10,6 +10,8 @@ import { TextToSpeechClient } from "@google-cloud/text-to-speech";
 import { TTSError, TTS_ERROR_CODES } from "../../utils/ttsProcessor.js";
 import { ErrorCategory, ErrorSeverity } from "../../constants/enums.js";
 import { logger } from "../../utils/logger.js";
+import { SpanSerializer, SpanType, SpanStatus, } from "../../observability/index.js";
+import { getMetricsAggregator } from "../../observability/index.js";
 export class GoogleTTSHandler {
     client = null;
     voicesCache = null;
@@ -68,18 +70,26 @@ export class GoogleTTSHandler {
                 retriable: false,
             });
         }
+        const span = SpanSerializer.createSpan(SpanType.TTS, "tts.google.listVoices", {
+            "tts.operation": "listVoices",
+            "tts.provider": "google",
+        });
         try {
             // Return cached voices if available, valid, and no language filter is specified
             if (this.voicesCache &&
                 Date.now() - this.voicesCache.timestamp <
                     GoogleTTSHandler.CACHE_TTL_MS &&
                 !languageCode) {
+                const endedSpan = SpanSerializer.endSpan(span, SpanStatus.OK);
+                getMetricsAggregator().recordSpan(endedSpan);
                 return this.voicesCache.voices;
             }
             // Call Google Cloud listVoices API
             const [response] = await this.client.listVoices(languageCode ? { languageCode } : {});
             if (!response.voices || response.voices.length === 0) {
                 logger.warn("Google Cloud TTS returned no voices");
+                const endedSpan = SpanSerializer.endSpan(span, SpanStatus.OK);
+                getMetricsAggregator().recordSpan(endedSpan);
                 return [];
             }
             const voices = [];
@@ -118,9 +128,14 @@ export class GoogleTTSHandler {
             if (!languageCode) {
                 this.voicesCache = { voices, timestamp: Date.now() };
             }
+            const endedSpan = SpanSerializer.endSpan(span, SpanStatus.OK);
+            getMetricsAggregator().recordSpan(endedSpan);
             return voices;
         }
         catch (err) {
+            // Record error span
+            const endedSpan = SpanSerializer.endSpan(span, SpanStatus.ERROR, err instanceof Error ? err.message : "Unknown error");
+            getMetricsAggregator().recordSpan(endedSpan);
             // Log error but return empty array for graceful degradation
             const message = err instanceof Error ? err.message : "Unknown error";
             logger.error(`Failed to fetch Google TTS voices: ${message}`);
@@ -144,6 +159,13 @@ export class GoogleTTSHandler {
                 retriable: false,
             });
         }
+        const voiceId = options.voice ?? "en-US-Neural2-C";
+        const span = SpanSerializer.createSpan(SpanType.TTS, "tts.google.synthesize", {
+            "tts.operation": "synthesize",
+            "tts.provider": "google",
+            "tts.voice": voiceId,
+            "tts.format": options.format ?? "mp3",
+        });
         const startTime = Date.now();
         try {
             const isSSML = text.startsWith("<speak>") && text.endsWith("</speak>");
@@ -160,7 +182,6 @@ export class GoogleTTSHandler {
                     retriable: false,
                 });
             }
-            const voiceId = options.voice ?? "en-US-Neural2-C";
             const languageCode = this.extractLanguageCode(voiceId);
             const audioEncoding = this.mapFormat(options.format ?? "mp3");
             const request = {
@@ -204,6 +225,8 @@ export class GoogleTTSHandler {
                         });
                     })();
             const latency = Date.now() - startTime;
+            const endedSpan = SpanSerializer.endSpan(span, SpanStatus.OK);
+            getMetricsAggregator().recordSpan(endedSpan);
             return {
                 buffer,
                 format: options.format ?? "mp3",
@@ -216,6 +239,8 @@ export class GoogleTTSHandler {
             };
         }
         catch (err) {
+            const endedSpan = SpanSerializer.endSpan(span, SpanStatus.ERROR, err instanceof Error ? err.message : String(err));
+            getMetricsAggregator().recordSpan(endedSpan);
             if (err instanceof TTSError) {
                 throw err;
             }

package/dist/lib/adapters/video/vertexVideoHandler.js

@@ -70,7 +70,14 @@ const DEFAULT_LOCATION = "us-central1";
  * ```
  */
 export function isVertexVideoConfigured() {
+    // Same credential detection as googleVertex.ts hasGoogleCredentials().
+    // GoogleAuth (used by getAccessToken) also supports ADC from
+    // `gcloud auth application-default login` automatically, so we only
+    // gate on the explicit env vars here — if none are set, we still
+    // allow the call through and let GoogleAuth resolve ADC at runtime.
+    // This avoids duplicating GoogleAuth's discovery logic.
     return !!(process.env.GOOGLE_APPLICATION_CREDENTIALS ||
+        process.env.GOOGLE_APPLICATION_CREDENTIALS_NEUROLINK ||
         process.env.GOOGLE_SERVICE_ACCOUNT_KEY ||
         (process.env.GOOGLE_AUTH_CLIENT_EMAIL &&
             process.env.GOOGLE_AUTH_PRIVATE_KEY));
@@ -270,21 +277,10 @@ function calculateDimensions(resolution, aspectRatio) {
  * ```
  */
 export async function generateVideoWithVertex(image, prompt, options = {}, region) {
-    // Validate configuration
-    if (!isVertexVideoConfigured()) {
-        throw new VideoError({
-            code: VIDEO_ERROR_CODES.PROVIDER_NOT_CONFIGURED,
-            message: "Vertex AI credentials not configured. Set GOOGLE_APPLICATION_CREDENTIALS environment variable",
-            category: ErrorCategory.CONFIGURATION,
-            severity: ErrorSeverity.HIGH,
-            retriable: false,
-            context: {
-                provider: "vertex",
-                feature: "video-generation",
-                suggestion: "Set GOOGLE_APPLICATION_CREDENTIALS to the path of your service account JSON file",
-            },
-        });
-    }
+    // Credential validation is deferred to getAccessToken() which uses
+    // GoogleAuth — it handles env vars, service accounts, AND ADC from
+    // `gcloud auth application-default login` automatically.
+    // Same pattern as googleVertex.ts — no synchronous pre-check needed.
     const config = await getVertexConfig();
     const project = config.project;
     const location = region || config.location;
@@ -312,7 +308,12 @@ export async function generateVideoWithVertex(image, prompt, options = {}, regio
         // Get auth token
         const accessToken = await getAccessToken();
         // Construct API request - predictLongRunning endpoint
-        const endpoint = `https://${location}-aiplatform.googleapis.com/v1/projects/${project}/locations/${location}/publishers/google/models/${VEO_MODEL}:predictLongRunning`;
+        // Global endpoint uses aiplatform.googleapis.com (no region prefix),
+        // same pattern as googleVertex.ts createVertexSettings
+        const apiHost = location === "global"
+            ? "aiplatform.googleapis.com"
+            : `${location}-aiplatform.googleapis.com`;
+        const endpoint = `https://${apiHost}/v1/projects/${project}/locations/${location}/publishers/google/models/${VEO_MODEL}:predictLongRunning`;
         // Request body structure (verified working from video.js reference)
         const requestBody = {
             instances: [
@@ -727,7 +728,12 @@ export async function generateTransitionWithVertex(firstFrame, lastFrame, prompt
  */
 async function pollOperation(modelOrEndpoint, operationName, accessToken, project, location, timeoutMs) {
     const startTime = Date.now();
-    const pollEndpoint = `https://${location}-aiplatform.googleapis.com/v1/projects/${project}/locations/${location}/publishers/google/models/${modelOrEndpoint}:fetchPredictOperation`;
+    // Global endpoint uses aiplatform.googleapis.com (no region prefix),
+    // same pattern as the predictLongRunning endpoint at line 374
+    const pollHost = location === "global"
+        ? "aiplatform.googleapis.com"
+        : `${location}-aiplatform.googleapis.com`;
+    const pollEndpoint = `https://${pollHost}/v1/projects/${project}/locations/${location}/publishers/google/models/${modelOrEndpoint}:fetchPredictOperation`;
     while (Date.now() - startTime < timeoutMs) {
         const result = await makePollRequest(pollEndpoint, operationName, accessToken);
         if (result.done) {

package/dist/lib/constants/contextWindows.js

@@ -88,7 +88,10 @@ export const MODEL_CONTEXT_WINDOWS = {
     },
     "google-ai": {
         _default: 1_048_576,
+        "gemini-3.1-pro": 1_048_576,
         "gemini-3.1-pro-preview": 1_048_576,
+        "gemini-3.1-flash": 1_048_576,
+        "gemini-3.1-flash-lite": 1_048_576,
         "gemini-3-pro-preview": 1_048_576,
         "gemini-3-pro-image-preview": 65_536,
         "gemini-3-flash-preview": 1_048_576,
@@ -114,7 +117,10 @@ export const MODEL_CONTEXT_WINDOWS = {
         "claude-opus-4-20250514": 200_000,
         "claude-opus-4": 200_000,
         // Gemini on Vertex
+        "gemini-3.1-pro": 1_048_576,
         "gemini-3.1-pro-preview": 1_048_576,
+        "gemini-3.1-flash": 1_048_576,
+        "gemini-3.1-flash-lite": 1_048_576,
         "gemini-3-pro-preview": 1_048_576,
         "gemini-3-pro-latest": 1_048_576,
         "gemini-3-flash-preview": 1_048_576,

package/dist/lib/constants/enums.d.ts

@@ -234,7 +234,10 @@ export declare enum VertexModels {
     CLAUDE_3_SONNET = "claude-3-sonnet-20240229",
     CLAUDE_3_OPUS = "claude-3-opus-20240229",
     CLAUDE_3_HAIKU = "claude-3-haiku-20240307",
+    GEMINI_3_1_PRO = "gemini-3.1-pro",
     GEMINI_3_1_PRO_PREVIEW = "gemini-3.1-pro-preview",
+    GEMINI_3_1_FLASH = "gemini-3.1-flash",
+    GEMINI_3_1_FLASH_LITE = "gemini-3.1-flash-lite",
     GEMINI_3_PRO = "gemini-3-pro",
     GEMINI_3_PRO_PREVIEW_11_2025 = "gemini-3-pro-preview-11-2025",
     GEMINI_3_PRO_LATEST = "gemini-3-pro-latest",
@@ -258,7 +261,10 @@ export declare enum VertexModels {
  * Supported Models for Google AI Studio
  */
 export declare enum GoogleAIModels {
+    GEMINI_3_1_PRO = "gemini-3.1-pro",
     GEMINI_3_1_PRO_PREVIEW = "gemini-3.1-pro-preview",
+    GEMINI_3_1_FLASH = "gemini-3.1-flash",
+    GEMINI_3_1_FLASH_LITE = "gemini-3.1-flash-lite",
     GEMINI_3_PRO_PREVIEW = "gemini-3-pro-preview",
     GEMINI_3_PRO_IMAGE_PREVIEW = "gemini-3-pro-image-preview",
     GEMINI_3_FLASH = "gemini-3-flash",

package/dist/lib/constants/enums.js

@@ -331,8 +331,11 @@ export var VertexModels;
     VertexModels["CLAUDE_3_SONNET"] = "claude-3-sonnet-20240229";
     VertexModels["CLAUDE_3_OPUS"] = "claude-3-opus-20240229";
     VertexModels["CLAUDE_3_HAIKU"] = "claude-3-haiku-20240307";
-    // Gemini 3.1 Series (Released February 2026)
+    // Gemini 3.1 Series (Released March 2026)
+    VertexModels["GEMINI_3_1_PRO"] = "gemini-3.1-pro";
     VertexModels["GEMINI_3_1_PRO_PREVIEW"] = "gemini-3.1-pro-preview";
+    VertexModels["GEMINI_3_1_FLASH"] = "gemini-3.1-flash";
+    VertexModels["GEMINI_3_1_FLASH_LITE"] = "gemini-3.1-flash-lite";
     // Gemini 3 Series (Preview)
     VertexModels["GEMINI_3_PRO"] = "gemini-3-pro";
     VertexModels["GEMINI_3_PRO_PREVIEW_11_2025"] = "gemini-3-pro-preview-11-2025";
@@ -361,8 +364,11 @@ export var VertexModels;
  */
 export var GoogleAIModels;
 (function (GoogleAIModels) {
-    // Gemini 3.1 Series (Released February 2026)
+    // Gemini 3.1 Series (Released March 2026)
+    GoogleAIModels["GEMINI_3_1_PRO"] = "gemini-3.1-pro";
     GoogleAIModels["GEMINI_3_1_PRO_PREVIEW"] = "gemini-3.1-pro-preview";
+    GoogleAIModels["GEMINI_3_1_FLASH"] = "gemini-3.1-flash";
+    GoogleAIModels["GEMINI_3_1_FLASH_LITE"] = "gemini-3.1-flash-lite";
     // Gemini 3 Series (Preview)
     GoogleAIModels["GEMINI_3_PRO_PREVIEW"] = "gemini-3-pro-preview";
     GoogleAIModels["GEMINI_3_PRO_IMAGE_PREVIEW"] = "gemini-3-pro-image-preview";

package/dist/lib/context/budgetChecker.js

@@ -8,6 +8,8 @@
  */
 import { getAvailableInputTokens } from "../constants/contextWindows.js";
 import { estimateMessagesTokens, estimateTokens, TOKENS_PER_MESSAGE, } from "../utils/tokenEstimation.js";
+import { SpanSerializer, SpanType, SpanStatus, } from "../observability/index.js";
+import { getMetricsAggregator } from "../observability/index.js";
 /** Default compaction threshold (80% of available input) */
 const DEFAULT_COMPACTION_THRESHOLD = 0.8;
 /** Estimated tokens per tool definition */
@@ -20,53 +22,78 @@ const TOKENS_PER_TOOL_DEFINITION = 200;
  * against available input space.
  */
 export function checkContextBudget(params) {
-    const { provider, model, maxTokens, systemPrompt, conversationMessages, currentPrompt, toolDefinitions, fileAttachments, compactionThreshold = DEFAULT_COMPACTION_THRESHOLD, } = params;
-    const availableInputTokens = getAvailableInputTokens(provider, model, maxTokens);
-    // Estimate each category
-    const systemPromptTokens = systemPrompt
-        ? estimateTokens(systemPrompt, provider) + TOKENS_PER_MESSAGE
-        : 0;
-    const conversationHistoryTokens = conversationMessages?.length
-        ? estimateMessagesTokens(conversationMessages, provider)
-        : 0;
-    const currentPromptTokens = currentPrompt
-        ? estimateTokens(currentPrompt, provider) + TOKENS_PER_MESSAGE
-        : 0;
-    const toolDefinitionTokens = toolDefinitions?.length
-        ? toolDefinitions.reduce((sum, tool) => {
-            try {
-                const serialized = JSON.stringify(tool);
-                return sum + estimateTokens(serialized, provider);
-            }
-            catch {
-                return sum + TOKENS_PER_TOOL_DEFINITION;
-            }
-        }, 0)
-        : 0;
-    const fileAttachmentTokens = fileAttachments?.length
-        ? fileAttachments.reduce((sum, file) => sum + estimateTokens(file.content, provider), 0)
-        : 0;
-    const estimatedInputTokens = systemPromptTokens +
-        conversationHistoryTokens +
-        currentPromptTokens +
-        toolDefinitionTokens +
-        fileAttachmentTokens;
-    const usageRatio = availableInputTokens > 0 ? estimatedInputTokens / availableInputTokens : 1;
-    const withinBudget = estimatedInputTokens <= availableInputTokens;
-    const shouldCompact = usageRatio >= compactionThreshold;
-    return {
-        withinBudget,
-        estimatedInputTokens,
-        availableInputTokens,
-        usageRatio,
-        shouldCompact,
-        breakdown: {
-            systemPrompt: systemPromptTokens,
-            conversationHistory: conversationHistoryTokens,
-            currentPrompt: currentPromptTokens,
-            toolDefinitions: toolDefinitionTokens,
-            fileAttachments: fileAttachmentTokens,
-        },
-    };
+    const span = SpanSerializer.createSpan(SpanType.CONTEXT_COMPACTION, "context.budgetCheck", {
+        "context.operation": "budgetCheck",
+    });
+    const startTime = Date.now();
+    try {
+        const { provider, model, maxTokens, systemPrompt, conversationMessages, currentPrompt, toolDefinitions, fileAttachments, compactionThreshold = DEFAULT_COMPACTION_THRESHOLD, } = params;
+        const availableInputTokens = getAvailableInputTokens(provider, model, maxTokens);
+        // Estimate each category
+        const systemPromptTokens = systemPrompt
+            ? estimateTokens(systemPrompt, provider) + TOKENS_PER_MESSAGE
+            : 0;
+        const conversationHistoryTokens = conversationMessages?.length
+            ? estimateMessagesTokens(conversationMessages, provider)
+            : 0;
+        const currentPromptTokens = currentPrompt
+            ? estimateTokens(currentPrompt, provider) + TOKENS_PER_MESSAGE
+            : 0;
+        const toolDefinitionTokens = toolDefinitions?.length
+            ? toolDefinitions.reduce((sum, tool) => {
+                try {
+                    const serialized = JSON.stringify(tool);
+                    return sum + estimateTokens(serialized, provider);
+                }
+                catch {
+                    return sum + TOKENS_PER_TOOL_DEFINITION;
+                }
+            }, 0)
+            : 0;
+        const fileAttachmentTokens = fileAttachments?.length
+            ? fileAttachments.reduce((sum, file) => sum + estimateTokens(file.content, provider), 0)
+            : 0;
+        const estimatedInputTokens = systemPromptTokens +
+            conversationHistoryTokens +
+            currentPromptTokens +
+            toolDefinitionTokens +
+            fileAttachmentTokens;
+        const usageRatio = availableInputTokens > 0
+            ? estimatedInputTokens / availableInputTokens
+            : 1;
+        const withinBudget = estimatedInputTokens <= availableInputTokens;
+        const shouldCompact = usageRatio >= compactionThreshold;
+        const result = {
+            withinBudget,
+            estimatedInputTokens,
+            availableInputTokens,
+            usageRatio,
+            shouldCompact,
+            breakdown: {
+                systemPrompt: systemPromptTokens,
+                conversationHistory: conversationHistoryTokens,
+                currentPrompt: currentPromptTokens,
+                toolDefinitions: toolDefinitionTokens,
+                fileAttachments: fileAttachmentTokens,
+            },
+        };
+        span.durationMs = Date.now() - startTime;
+        const endedSpan = SpanSerializer.endSpan(SpanSerializer.updateAttributes(span, {
+            "context.budgetUsage": usageRatio,
+            "context.triggered": shouldCompact,
+            "context.estimatedTokens": estimatedInputTokens,
+            "context.availableTokens": availableInputTokens,
+        }), SpanStatus.OK);
+        getMetricsAggregator().recordSpan(endedSpan);
+        return result;
+    }
+    catch (error) {
+        span.durationMs = Date.now() - startTime;
+        const endedSpan = SpanSerializer.endSpan(span, SpanStatus.ERROR);
+        endedSpan.statusMessage =
+            error instanceof Error ? error.message : String(error);
+        getMetricsAggregator().recordSpan(endedSpan);
+        throw error;
+    }
 }
 //# sourceMappingURL=budgetChecker.js.map