@juspay/neurolink 10.8.4 → 10.8.6

package/dist/lib/server/routes/openaiProxyRoutes.js

@@ -321,7 +321,7 @@ export function createOpenAIProxyRoutes(modelRouter, basePath = "", loopbackPort
                         model: targetModel,
                     }, requestModelRouter?.getFallbackChain() ?? [], body.model, 
                     // The classifier only reads fields present on both types.
-                    adapted);
+                    adapted, requestModelRouter?.isAutoFallbackEnabled?.() ?? false);
                     const attempts = plan.attempts;
                     // --- Optional tracing ---
                     let tracer;

package/dist/lib/types/errors.d.ts

@@ -12,6 +12,11 @@ export declare class ProviderError extends BaseError {
     provider?: string | undefined;
     constructor(message: string, provider?: string | undefined);
 }
+export type OllamaHttpError = ProviderError & {
+    statusCode?: number;
+    statusText?: string;
+    responseBody?: string;
+};
 /**
  * Thrown for authentication-related errors, such as invalid or missing API keys.
  */

package/dist/lib/types/proxy.d.ts

@@ -29,6 +29,8 @@ export type ModelRouterInterface = {
     resolve(requestedModel: string): RouteResult;
     isClaudeTarget(requestedModel: string): boolean;
     getFallbackChain(): FallbackEntry[];
+    isAutoFallbackEnabled?(): boolean;
+    getMaxInflightPerAccount?(): number;
     getModelMappings?: () => ModelMapping[];
     getPassthroughModels?: () => string[];
 };
@@ -639,6 +641,27 @@ export type AnthropicSuccessResult = {
     };
 } | {
     response: Response | unknown;
+    holdsAccountAdmission?: boolean;
+};
+/** A release handle for one in-flight request admitted to an OAuth account. */
+export type AccountAdmissionLease = {
+    release(): void;
+};
+/** A cancellable queued request for per-account admission capacity. */
+export type QueuedAccountAdmission = {
+    accountKey: string;
+    promise: Promise<AccountAdmissionLease>;
+    cancel(): void;
+};
+/** One queued request waiting for per-account admission capacity. */
+export type AccountAdmissionWaiter = {
+    capacity: number;
+    resolve: (lease: AccountAdmissionLease) => void;
+};
+/** In-process request admission state for an OAuth account. */
+export type AccountAdmissionState = {
+    active: number;
+    waiters: AccountAdmissionWaiter[];
 };
 /** Result of buffering only enough upstream SSE to make a retry-safe decision. */
 export type AnthropicStreamPreflightResult = {
@@ -669,6 +692,7 @@ export type ParsedSSEBuffer = {
 };
 export type AnthropicAuthRetryResult = {
     response?: Response | unknown;
+    holdsAccountAdmission?: boolean;
     continueLoop: boolean;
     lastError: unknown;
     authFailureMessage: string | null;

package/dist/lib/types/subscription.d.ts

@@ -908,6 +908,10 @@ export type ProxyRoutingConfig = {
     strategy: "round-robin" | "fill-first";
     modelMappings: ModelMapping[];
     fallbackChain: FallbackEntry[];
+    /** Permit a last-resort provider chosen by the translation layer. Disabled by default. */
+    autoFallback?: boolean;
+    /** Maximum in-flight upstream requests per OAuth account. Defaults to two. */
+    maxInflightPerAccount?: number;
     passthroughModels?: string[];
     /** Enable quota-aware fill-first account ordering. Defaults to true. */
     quotaRouting?: boolean;

package/dist/lib/utils/providerHealth.js

@@ -7,6 +7,7 @@ import { AIProviderName, OpenAIModels, GoogleAIModels, AnthropicModels, BedrockM
 import { API_KEY_LENGTHS, PROJECT_ID_FORMAT } from "./providerConfig.js";
 import { basename } from "path";
 import { createProxyFetch } from "../proxy/proxyFetch.js";
+import { DEFAULT_OLLAMA_MODEL } from "../providers/ollama/constants.js";
 export class ProviderHealthChecker {
     static healthCache = new Map();
     static DEFAULT_TIMEOUT = 5000; // 5 seconds
@@ -690,7 +691,7 @@ export class ProviderHealthChecker {
         return new URL("/api/tags", this.getOllamaBaseUrl()).toString();
     }
     static getConfiguredOllamaModel() {
-        return process.env.OLLAMA_MODEL || "llama3.1:8b";
+        return process.env.OLLAMA_MODEL || DEFAULT_OLLAMA_MODEL;
     }
     static async fetchJsonWithTimeout(url, options = {}) {
         const controller = new AbortController();

package/dist/lib/utils/providerUtils.js

@@ -7,6 +7,7 @@ import { logger } from "./logger.js";
 import { AIProviderName } from "../constants/enums.js";
 import { ProviderHealthChecker } from "./providerHealth.js";
 import { API_KEY_FORMATS, API_KEY_LENGTHS, PROJECT_ID_FORMAT, } from "./providerConfig.js";
+import { DEFAULT_OLLAMA_MODEL } from "../providers/ollama/constants.js";
 /**
  * Get the best available provider based on real-time availability checks
  * Enhanced version consolidated from providerUtils-fixed.ts
@@ -107,7 +108,7 @@ async function isProviderAvailable(providerName) {
         return availability.available;
     }
     if (providerName === "ollama") {
-        const availability = await ProviderHealthChecker.checkFallbackProviderAvailability(AIProviderName.OLLAMA, process.env.OLLAMA_MODEL || "llama3.1:8b");
+        const availability = await ProviderHealthChecker.checkFallbackProviderAvailability(AIProviderName.OLLAMA, process.env.OLLAMA_MODEL || DEFAULT_OLLAMA_MODEL);
         return availability.available;
     }
     try {

package/dist/providers/{amazonBedrock.d.ts → amazonBedrock/client.d.ts}

@@ -1,8 +1,8 @@
 import { BedrockRuntimeClient } from "@aws-sdk/client-bedrock-runtime";
-import type { AIProviderName } from "../constants/enums.js";
-import { BaseProvider } from "../core/baseProvider.js";
-import type { NeuroLink } from "../neurolink.js";
-import type { StreamOptions, StreamResult, EnhancedGenerateResult, TextGenerationOptions } from "../types/index.js";
+import type { AIProviderName } from "../../constants/enums.js";
+import { BaseProvider } from "../../core/baseProvider.js";
+import type { NeuroLink } from "../../neurolink.js";
+import type { StreamOptions, StreamResult, EnhancedGenerateResult, TextGenerationOptions } from "../../types/index.js";
 export declare class AmazonBedrockProvider extends BaseProvider {
     private bedrockClient;
     private conversationHistory;

package/dist/{lib/providers/amazonBedrock.js → providers/amazonBedrock/client.js}

@@ -1,24 +1,20 @@
-async function loadBedrockControl() {
-    return await import(/* @vite-ignore */ "@aws-sdk/client-bedrock");
-}
 import { BedrockRuntimeClient, ConverseCommand, ConverseStreamCommand, ImageFormat, } from "@aws-sdk/client-bedrock-runtime";
 import path from "path";
-import { createAnalytics } from "../core/analytics.js";
-import { BaseProvider } from "../core/baseProvider.js";
-import { DEFAULT_MAX_STEPS } from "../core/constants.js";
-import { AuthenticationError, ProviderError, RateLimitError, } from "../types/index.js";
-import { isAbortError, withTimeout } from "../utils/errorHandling.js";
-import { emitToolEndFromStepFinish } from "../utils/toolEndEmitter.js";
-import { logger } from "../utils/logger.js";
-import { resolveSamplingParams } from "../models/modelRegistry.js";
-import { calculateCost } from "../utils/pricing.js";
-import { buildMultimodalMessagesArray } from "../utils/messageBuilder.js";
-import { buildMultimodalOptions } from "../utils/multimodalOptionsBuilder.js";
-import { convertZodToJsonSchema } from "../utils/schemaConversion.js";
+import { createAnalytics } from "../../core/analytics.js";
+import { BaseProvider } from "../../core/baseProvider.js";
+import { DEFAULT_MAX_STEPS } from "../../core/constants.js";
+import { AuthenticationError, ProviderError, RateLimitError, } from "../../types/index.js";
+import { isAbortError, withTimeout } from "../../utils/errorHandling.js";
+import { emitToolEndFromStepFinish } from "../../utils/toolEndEmitter.js";
+import { logger } from "../../utils/logger.js";
+import { resolveSamplingParams } from "../../models/modelRegistry.js";
+import { calculateCost } from "../../utils/pricing.js";
+import { buildMultimodalMessagesArray } from "../../utils/messageBuilder.js";
+import { buildMultimodalOptions } from "../../utils/multimodalOptionsBuilder.js";
+import { convertZodToJsonSchema } from "../../utils/schemaConversion.js";
 import { SpanKind, SpanStatusCode } from "@opentelemetry/api";
-import { tracers } from "../telemetry/index.js";
-const bedrockTracer = tracers.provider;
-// Bedrock-specific types now imported from ../types/providerSpecific.js
+import { bedrockTracer } from "./constants.js";
+import { loadBedrockControl } from "./utils.js";
 export class AmazonBedrockProvider extends BaseProvider {
     bedrockClient;
     conversationHistory = [];
@@ -1826,4 +1822,3 @@ export class AmazonBedrockProvider extends BaseProvider {
         }
     }
 }
-//# sourceMappingURL=amazonBedrock.js.map

package/dist/providers/amazonBedrock/constants.d.ts

@@ -0,0 +1 @@
+export declare const bedrockTracer: import("@opentelemetry/api").Tracer;

package/dist/providers/amazonBedrock/constants.js

@@ -0,0 +1,2 @@
+import { tracers } from "../../telemetry/index.js";
+export const bedrockTracer = tracers.provider;

package/dist/providers/amazonBedrock/index.d.ts

@@ -0,0 +1,3 @@
+export * from "./client.js";
+export * from "./constants.js";
+export * from "./utils.js";

package/dist/providers/amazonBedrock/index.js

@@ -0,0 +1,3 @@
+export * from "./client.js";
+export * from "./constants.js";
+export * from "./utils.js";

package/dist/providers/amazonBedrock/utils.d.ts

@@ -0,0 +1 @@
+export declare function loadBedrockControl(): Promise<typeof import("@aws-sdk/client-bedrock")>;

package/dist/providers/amazonBedrock/utils.js

@@ -0,0 +1,3 @@
+export async function loadBedrockControl() {
+    return await import(/* @vite-ignore */ "@aws-sdk/client-bedrock");
+}

package/dist/{lib/providers/anthropic.d.ts → providers/anthropic/client.d.ts}

@@ -1,21 +1,7 @@
-import { type AIProviderName } from "../constants/enums.js";
-import { BaseProvider } from "../core/baseProvider.js";
-import type { AnthropicProviderConfig, StreamOptions, StreamResult, ValidationSchema, EnhancedGenerateResult, TextGenerationOptions, AnthropicAuthMethod, AnthropicResponseMetadata, ClaudeSubscriptionTier, ClaudeUsageInfo } from "../types/index.js";
-import type { LanguageModel } from "../types/index.js";
-/**
- * Beta headers for Claude Code integration.
- * These enable experimental features:
- * - claude-code-20250219: Claude Code specific features
- * - fine-grained-tool-streaming-2025-05-14: Fine-grained tool streaming
- *
- * Note: interleaved-thinking-2025-05-14 was removed — it was claude-3-7-sonnet
- * specific and causes invalid_request_error (HTTP 400) on claude-4 models
- * (claude-opus-4-6, claude-sonnet-4-6) which handle thinking via the
- * `thinking` request body parameter instead.
- */
-declare const ANTHROPIC_BETA_HEADERS: {
-    "anthropic-beta": string;
-};
+import { type AIProviderName } from "../../constants/enums.js";
+import { BaseProvider } from "../../core/baseProvider.js";
+import type { AnthropicProviderConfig, StreamOptions, StreamResult, ValidationSchema, EnhancedGenerateResult, TextGenerationOptions, AnthropicAuthMethod, AnthropicResponseMetadata, ClaudeSubscriptionTier, ClaudeUsageInfo } from "../../types/index.js";
+import type { LanguageModel } from "../../types/index.js";
 /**
  * Anthropic Provider v2 - BaseProvider Implementation
  * Enhanced with OAuth support, subscription tiers, and beta headers for Claude Code integration.
@@ -91,7 +77,7 @@ export declare class AnthropicProvider extends BaseProvider {
      *
      * @returns The model capabilities or undefined if not found
      */
-    getModelCapabilities(): import("../types/subscription.js").AnthropicModelMetadata | undefined;
+    getModelCapabilities(): import("../../index.js").AnthropicModelMetadata | undefined;
     /**
      * Get the current subscription tier.
      * @returns The detected or configured subscription tier
@@ -145,5 +131,4 @@ export declare class AnthropicProvider extends BaseProvider {
     isAvailable(): Promise<boolean>;
     getModel(): LanguageModel;
 }
-export { getModelCapabilities, getRecommendedModelForTier, isModelAvailableForTier, ModelAccessError, } from "../models/anthropicModels.js";
-export { ANTHROPIC_BETA_HEADERS };
+export { getModelCapabilities, getRecommendedModelForTier, isModelAvailableForTier, ModelAccessError, } from "../../models/anthropicModels.js";

package/dist/{lib/providers/anthropic.js → providers/anthropic/client.js}

@@ -3,48 +3,32 @@ import { SpanKind, trace } from "@opentelemetry/api";
 import { existsSync, mkdirSync, readFileSync, renameSync, writeFileSync, } from "fs";
 import { homedir } from "os";
 import { join } from "path";
-import { ANTHROPIC_TOKEN_URL, CLAUDE_CLI_USER_AGENT, CLAUDE_CODE_CLIENT_ID, CLAUDE_CODE_OAUTH_BETAS, } from "../auth/anthropicOAuth.js";
-import { AnthropicModels, TOKEN_EXPIRY_BUFFER_MS, } from "../constants/enums.js";
-import { BaseProvider } from "../core/baseProvider.js";
-import { DEFAULT_MAX_STEPS } from "../core/constants.js";
-import { streamAnalyticsCollector } from "../core/streamAnalytics.js";
-import { getModelCapabilities, getRecommendedModelForTier, isModelAvailableForTier, } from "../models/anthropicModels.js";
-import { createOAuthFetch } from "../proxy/oauthFetch.js";
-import { createProxyFetch } from "../proxy/proxyFetch.js";
-import { AuthenticationError, NetworkError, ProviderError, RateLimitError, } from "../types/index.js";
-import { logger } from "../utils/logger.js";
-import { redactUrlCredentials } from "../utils/logSanitize.js";
-import { ANTHROPIC_MAX_CACHE_BREAKPOINTS, applyAnthropicHistoryCacheBreakpoints, countAnthropicCacheMarkers, } from "../utils/anthropicCacheBreakpoints.js";
-import { calculateCost } from "../utils/pricing.js";
-import { resolveDeferredTool } from "../tools/toolDiscovery.js";
-import { createAnthropicConfig, getProviderModel, validateApiKey, } from "../utils/providerConfig.js";
-import { composeAbortSignals, createTimeoutController, mergeAbortSignals, TimeoutError, } from "../utils/timeout.js";
-import { resolveToolChoice } from "../utils/toolChoice.js";
-import { emitToolEndFromStepFinish } from "../utils/toolEndEmitter.js";
-import { NoOutputGeneratedError } from "../utils/generationErrors.js";
-import { buildNoOutputSentinel, stampNoOutputSpan, } from "../utils/noOutputSentinel.js";
-import { convertZodToJsonSchema } from "../utils/schemaConversion.js";
-import { resolveClaudeMaxTokens } from "../utils/tokenLimits.js";
-import { toAnthropicImageBlock, fileToAnthropicBlock, } from "./anthropicImageBlocks.js";
-import { resolveSamplingParams } from "../models/modelRegistry.js";
-import { createChunkQueue, createDeferredAnalytics, stringifyToolInput, } from "./openaiChatCompletionsClient.js";
-/**
- * Beta headers for Claude Code integration.
- * These enable experimental features:
- * - claude-code-20250219: Claude Code specific features
- * - fine-grained-tool-streaming-2025-05-14: Fine-grained tool streaming
- *
- * Note: interleaved-thinking-2025-05-14 was removed — it was claude-3-7-sonnet
- * specific and causes invalid_request_error (HTTP 400) on claude-4 models
- * (claude-opus-4-6, claude-sonnet-4-6) which handle thinking via the
- * `thinking` request body parameter instead.
- */
-const ANTHROPIC_BETA_HEADERS = {
-    "anthropic-beta": [
-        "claude-code-20250219",
-        "fine-grained-tool-streaming-2025-05-14",
-    ].join(","),
-};
+import { ANTHROPIC_TOKEN_URL, CLAUDE_CLI_USER_AGENT, CLAUDE_CODE_CLIENT_ID, CLAUDE_CODE_OAUTH_BETAS, } from "../../auth/anthropicOAuth.js";
+import { AnthropicModels, TOKEN_EXPIRY_BUFFER_MS, } from "../../constants/enums.js";
+import { BaseProvider } from "../../core/baseProvider.js";
+import { DEFAULT_MAX_STEPS } from "../../core/constants.js";
+import { streamAnalyticsCollector } from "../../core/streamAnalytics.js";
+import { getModelCapabilities, getRecommendedModelForTier, isModelAvailableForTier, } from "../../models/anthropicModels.js";
+import { createOAuthFetch } from "../../proxy/oauthFetch.js";
+import { createProxyFetch } from "../../proxy/proxyFetch.js";
+import { AuthenticationError, NetworkError, ProviderError, RateLimitError, } from "../../types/index.js";
+import { logger } from "../../utils/logger.js";
+import { redactUrlCredentials } from "../../utils/logSanitize.js";
+import { ANTHROPIC_MAX_CACHE_BREAKPOINTS, applyAnthropicHistoryCacheBreakpoints, countAnthropicCacheMarkers, } from "../../utils/anthropicCacheBreakpoints.js";
+import { calculateCost } from "../../utils/pricing.js";
+import { resolveDeferredTool } from "../../tools/toolDiscovery.js";
+import { createAnthropicConfig, getProviderModel, validateApiKey, } from "../../utils/providerConfig.js";
+import { composeAbortSignals, createTimeoutController, mergeAbortSignals, TimeoutError, } from "../../utils/timeout.js";
+import { resolveToolChoice } from "../../utils/toolChoice.js";
+import { emitToolEndFromStepFinish } from "../../utils/toolEndEmitter.js";
+import { NoOutputGeneratedError } from "../../utils/generationErrors.js";
+import { buildNoOutputSentinel, stampNoOutputSpan, } from "../../utils/noOutputSentinel.js";
+import { convertZodToJsonSchema } from "../../utils/schemaConversion.js";
+import { resolveClaudeMaxTokens } from "../../utils/tokenLimits.js";
+import { toAnthropicImageBlock, fileToAnthropicBlock, } from "../anthropicImageBlocks.js";
+import { resolveSamplingParams } from "../../models/modelRegistry.js";
+import { createChunkQueue, createDeferredAnalytics, stringifyToolInput, } from "../openaiChatCompletionsClient.js";
+import { ANTHROPIC_BETA_HEADERS } from "./constants.js";
 // AnthropicProviderConfig is imported from types/providers.ts
 // Re-export for backward compatibility
 // Configuration helpers - now using consolidated utility
@@ -1805,7 +1789,4 @@ export class AnthropicProvider extends BaseProvider {
     }
 }
 // Re-export types and utilities for convenience
-export { getModelCapabilities, getRecommendedModelForTier, isModelAvailableForTier, ModelAccessError, } from "../models/anthropicModels.js";
-// Export beta headers constant for external use
-export { ANTHROPIC_BETA_HEADERS };
-//# sourceMappingURL=anthropic.js.map
+export { getModelCapabilities, getRecommendedModelForTier, isModelAvailableForTier, ModelAccessError, } from "../../models/anthropicModels.js";

package/dist/providers/anthropic/constants.d.ts

@@ -0,0 +1,4 @@
+export declare const ANTHROPIC_BETA_HEADERS: {
+    "anthropic-beta": string;
+};
+export declare const streamTracer: import("@opentelemetry/api").Tracer;

package/dist/providers/anthropic/constants.js

@@ -0,0 +1,8 @@
+import { trace } from "@opentelemetry/api";
+export const ANTHROPIC_BETA_HEADERS = {
+    "anthropic-beta": [
+        "claude-code-20250219",
+        "fine-grained-tool-streaming-2025-05-14",
+    ].join(","),
+};
+export const streamTracer = trace.getTracer("neurolink.provider.anthropic");

package/dist/providers/anthropic/index.d.ts

@@ -0,0 +1,3 @@
+export * from "./client.js";
+export * from "./constants.js";
+export * from "./utils.js";

package/dist/providers/anthropic/index.js

@@ -0,0 +1,3 @@
+export * from "./client.js";
+export * from "./constants.js";
+export * from "./utils.js";

package/dist/providers/anthropic/utils.d.ts

@@ -0,0 +1,7 @@
+import type { AnthropicAuthMethod, AnthropicRateLimitInfo, ClaudeSubscriptionTier, OAuthToken } from "../../types/index.js";
+export declare const getAnthropicApiKey: () => string;
+export declare const getDefaultAnthropicModel: () => string;
+export declare const getOAuthToken: () => OAuthToken | null;
+export declare const detectSubscriptionTier: (oauthToken: OAuthToken | null) => ClaudeSubscriptionTier;
+export declare const detectAuthMethod: (oauthToken: OAuthToken | null) => AnthropicAuthMethod;
+export declare const parseRateLimitHeaders: (headers: Headers | Record<string, string>) => AnthropicRateLimitInfo;

package/dist/providers/anthropic/utils.js

@@ -0,0 +1,149 @@
+import { existsSync, readFileSync } from "fs";
+import { homedir } from "os";
+import { join } from "path";
+import { AnthropicModels } from "../../constants/enums.js";
+import { logger } from "../../utils/logger.js";
+import { createAnthropicConfig, getProviderModel, validateApiKey, } from "../../utils/providerConfig.js";
+export const getAnthropicApiKey = () => {
+    return validateApiKey(createAnthropicConfig());
+};
+export const getDefaultAnthropicModel = () => {
+    return getProviderModel("ANTHROPIC_MODEL", AnthropicModels.CLAUDE_SONNET_4_6);
+};
+export const getOAuthToken = () => {
+    // First, check stored credentials file (highest priority)
+    try {
+        const credentialsPath = join(homedir(), ".neurolink", "anthropic-credentials.json");
+        if (existsSync(credentialsPath)) {
+            const credentialsContent = readFileSync(credentialsPath, "utf-8");
+            const credentials = JSON.parse(credentialsContent);
+            if (credentials.type === "oauth" && credentials.oauth?.accessToken) {
+                logger.debug("[AnthropicProvider] Using OAuth token from stored credentials file");
+                return credentials.oauth;
+            }
+        }
+    }
+    catch (error) {
+        logger.debug("[AnthropicProvider] Failed to read stored credentials:", error);
+    }
+    // Fallback to environment variables
+    const tokenString = process.env.ANTHROPIC_OAUTH_TOKEN || process.env.CLAUDE_OAUTH_TOKEN;
+    if (!tokenString) {
+        return null;
+    }
+    // Try to parse as JSON (for full token object with refresh token and expiry)
+    try {
+        const parsed = JSON.parse(tokenString);
+        if (typeof parsed === "object" && parsed.accessToken) {
+            return parsed;
+        }
+        // If it's a simple string in JSON, use it as access token
+        if (typeof parsed === "string") {
+            return { accessToken: parsed };
+        }
+    }
+    catch {
+        // Not JSON, treat as plain access token string
+    }
+    // Treat as plain access token string
+    return { accessToken: tokenString };
+};
+export const detectSubscriptionTier = (oauthToken) => {
+    // Check explicit environment variable first
+    const envTier = process.env.ANTHROPIC_SUBSCRIPTION_TIER?.toLowerCase();
+    if (envTier) {
+        const validTiers = [
+            "free",
+            "pro",
+            "max",
+            "max_5",
+            "max_20",
+            "api",
+        ];
+        if (validTiers.includes(envTier)) {
+            logger.debug("[detectSubscriptionTier] Using environment override", {
+                tier: envTier,
+            });
+            return envTier;
+        }
+        logger.warn("[detectSubscriptionTier] Invalid ANTHROPIC_SUBSCRIPTION_TIER", {
+            value: envTier,
+            validTiers,
+        });
+    }
+    // If using OAuth, default to 'pro' (most common subscription tier)
+    if (oauthToken) {
+        // Check if token scopes indicate tier (future-proofing)
+        const scopes = oauthToken.scopes ?? [];
+        let detectedTier = "pro";
+        if (scopes.includes("max_20")) {
+            detectedTier = "max_20";
+        }
+        else if (scopes.includes("max_5")) {
+            detectedTier = "max_5";
+        }
+        else if (scopes.includes("max")) {
+            detectedTier = "max";
+        }
+        logger.debug("[detectSubscriptionTier] Detected from OAuth token", {
+            tier: detectedTier,
+            scopes,
+        });
+        return detectedTier;
+    }
+    // Default to 'api' for API key authentication
+    logger.debug("[detectSubscriptionTier] No OAuth token, defaulting to API tier");
+    return "api";
+};
+export const detectAuthMethod = (oauthToken) => {
+    // Explicit env var takes highest precedence — allows forcing api_key mode
+    // even when OAuth credentials exist (e.g., when using a proxy that handles auth)
+    const explicit = process.env.ANTHROPIC_AUTH_METHOD?.toLowerCase();
+    if (explicit === "api_key" || explicit === "apikey") {
+        logger.debug("[detectAuthMethod] Forced to api_key by ANTHROPIC_AUTH_METHOD env var");
+        return "api_key";
+    }
+    if (explicit === "oauth") {
+        if (oauthToken) {
+            logger.debug("[detectAuthMethod] Forced to oauth by ANTHROPIC_AUTH_METHOD env var");
+            return "oauth";
+        }
+        logger.warn("[detectAuthMethod] ANTHROPIC_AUTH_METHOD=oauth but no OAuth token found; falling through to auto-detection");
+    }
+    else if (explicit) {
+        logger.warn("[detectAuthMethod] Unrecognized ANTHROPIC_AUTH_METHOD value; falling through to auto-detection", {
+            value: explicit,
+        });
+    }
+    // Auto-detect: OAuth takes precedence if available
+    const method = oauthToken ? "oauth" : "api_key";
+    logger.debug("[detectAuthMethod] Auth method resolved", {
+        method,
+        hasOAuthToken: !!oauthToken,
+    });
+    return method;
+};
+export const parseRateLimitHeaders = (headers) => {
+    const getHeader = (name) => {
+        if (headers instanceof Headers) {
+            return headers.get(name);
+        }
+        return headers[name] || headers[name.toLowerCase()] || null;
+    };
+    const parseNumber = (value) => {
+        if (!value) {
+            return undefined;
+        }
+        const num = parseInt(value, 10);
+        return isNaN(num) ? undefined : num;
+    };
+    return {
+        requestsLimit: parseNumber(getHeader("anthropic-ratelimit-requests-limit")),
+        requestsRemaining: parseNumber(getHeader("anthropic-ratelimit-requests-remaining")),
+        requestsReset: getHeader("anthropic-ratelimit-requests-reset") || undefined,
+        tokensLimit: parseNumber(getHeader("anthropic-ratelimit-tokens-limit")),
+        tokensRemaining: parseNumber(getHeader("anthropic-ratelimit-tokens-remaining")),
+        tokensReset: getHeader("anthropic-ratelimit-tokens-reset") || undefined,
+        retryAfter: parseNumber(getHeader("retry-after")),
+    };
+};

package/dist/providers/{googleAiStudio.d.ts → googleAiStudio/client.d.ts}

@@ -1,7 +1,7 @@
-import { type AIProviderName } from "../constants/enums.js";
-import { BaseProvider } from "../core/baseProvider.js";
-import type { ZodUnknownSchema, EnhancedGenerateResult, TextGenerationOptions, StreamOptions, StreamResult } from "../types/index.js";
-import type { LanguageModel, Schema } from "../types/index.js";
+import { type AIProviderName } from "../../constants/enums.js";
+import { BaseProvider } from "../../core/baseProvider.js";
+import type { ZodUnknownSchema, EnhancedGenerateResult, TextGenerationOptions, StreamOptions, StreamResult } from "../../types/index.js";
+import type { LanguageModel, Schema } from "../../types/index.js";
 /**
  * Google AI Studio provider implementation using BaseProvider
  * Migrated from original GoogleAIStudio class to new factory pattern

package/dist/{lib/providers/googleAiStudio.js → providers/googleAiStudio/client.js}

@@ -1,18 +1,18 @@
-import { ErrorCategory, ErrorSeverity, GoogleAIModels, } from "../constants/enums.js";
-import { BaseProvider } from "../core/baseProvider.js";
-import { IMAGE_GENERATION_MODELS, TOOL_STORAGE_TIMEOUT_MS, } from "../core/constants.js";
-import { processUnifiedFilesArray } from "../utils/messageBuilder.js";
-import { ATTR, tracers, withClientSpan, withClientStreamSpan, withSpan, } from "../telemetry/index.js";
-import { AuthenticationError, InvalidModelError, NetworkError, ProviderError, RateLimitError, } from "../types/index.js";
-import { ERROR_CODES, NeuroLinkError } from "../utils/errorHandling.js";
-import { logger } from "../utils/logger.js";
-import { composeAbortSignals, createTimeoutController, TimeoutError, } from "../utils/timeout.js";
-import { withTimeout } from "../utils/async/index.js";
-import { estimateTokens } from "../utils/tokenEstimation.js";
-import { transformToolExecutions } from "../utils/transformationUtils.js";
-import { resolveToolExecutionRecords } from "../core/toolExecutionRecorder.js";
-import { buildGeminiResponseSchema, buildNativeConfig, buildNativeToolDeclarations, collectStreamChunks, collectStreamChunksIncremental, computeMaxSteps, createTextChannel, buildUserPartsWithMultimodal, executeNativeToolCalls, extractTextFromParts, extractThoughtSignature, handleMaxStepsTermination, prependConversationMessages, pushModelResponseToHistory, refreshNativeToolDeclarations, DedupExecuteMap, } from "./googleNativeGemini3.js";
-import { createProxyFetch } from "../proxy/proxyFetch.js";
+import { ErrorCategory, ErrorSeverity, GoogleAIModels, } from "../../constants/enums.js";
+import { BaseProvider } from "../../core/baseProvider.js";
+import { IMAGE_GENERATION_MODELS, TOOL_STORAGE_TIMEOUT_MS, } from "../../core/constants.js";
+import { processUnifiedFilesArray } from "../../utils/messageBuilder.js";
+import { ATTR, tracers, withClientSpan, withClientStreamSpan, withSpan, } from "../../telemetry/index.js";
+import { AuthenticationError, InvalidModelError, NetworkError, ProviderError, RateLimitError, } from "../../types/index.js";
+import { ERROR_CODES, NeuroLinkError } from "../../utils/errorHandling.js";
+import { logger } from "../../utils/logger.js";
+import { composeAbortSignals, createTimeoutController, TimeoutError, } from "../../utils/timeout.js";
+import { withTimeout } from "../../utils/async/index.js";
+import { estimateTokens } from "../../utils/tokenEstimation.js";
+import { transformToolExecutions } from "../../utils/transformationUtils.js";
+import { resolveToolExecutionRecords } from "../../core/toolExecutionRecorder.js";
+import { buildGeminiResponseSchema, buildNativeConfig, buildNativeToolDeclarations, collectStreamChunks, collectStreamChunksIncremental, computeMaxSteps, createTextChannel, buildUserPartsWithMultimodal, executeNativeToolCalls, extractTextFromParts, extractThoughtSignature, handleMaxStepsTermination, prependConversationMessages, pushModelResponseToHistory, refreshNativeToolDeclarations, DedupExecuteMap, } from "../googleNativeGemini3/index.js";
+import { createProxyFetch } from "../../proxy/proxyFetch.js";
 // Google AI Live API types now imported from ../types/providerSpecific.js
 // Import proper types for multimodal message handling
 // Create Google GenAI client
@@ -1442,4 +1442,3 @@ export class GoogleAIStudioProvider extends BaseProvider {
     }
 }
 export default GoogleAIStudioProvider;
-//# sourceMappingURL=googleAiStudio.js.map

package/dist/providers/googleAiStudio/index.d.ts

@@ -0,0 +1,2 @@
+export * from "./client.js";
+export * from "./utils.js";

package/dist/providers/googleAiStudio/index.js

@@ -0,0 +1,2 @@
+export * from "./client.js";
+export * from "./utils.js";

package/dist/providers/googleAiStudio/utils.d.ts

@@ -0,0 +1,2 @@
+import type { GenAIClient } from "../../types/index.js";
+export declare function createGoogleGenAIClient(apiKey: string): Promise<GenAIClient>;

package/dist/providers/googleAiStudio/utils.js

@@ -0,0 +1,18 @@
+import { ErrorCategory, ErrorSeverity } from "../../constants/enums.js";
+import { ERROR_CODES, NeuroLinkError } from "../../utils/errorHandling.js";
+export async function createGoogleGenAIClient(apiKey) {
+    const mod = await import("@google/genai");
+    const ctor = mod.GoogleGenAI;
+    if (!ctor) {
+        throw new NeuroLinkError({
+            code: ERROR_CODES.INVALID_CONFIGURATION,
+            message: "@google/genai does not export GoogleGenAI",
+            category: ErrorCategory.CONFIGURATION,
+            severity: ErrorSeverity.CRITICAL,
+            retriable: false,
+            context: { module: "@google/genai", expectedExport: "GoogleGenAI" },
+        });
+    }
+    const Ctor = ctor;
+    return new Ctor({ apiKey });
+}

package/dist/providers/googleNativeGemini3/constants.d.ts

@@ -0,0 +1 @@
+export declare const GEMINI3_NATIVE_MAX_STEPS = 100;

package/dist/providers/googleNativeGemini3/constants.js

@@ -0,0 +1 @@
+export const GEMINI3_NATIVE_MAX_STEPS = 100;

package/dist/providers/googleNativeGemini3/index.d.ts

@@ -0,0 +1,2 @@
+export * from "./constants.js";
+export * from "./utils.js";

package/dist/providers/googleNativeGemini3/index.js

@@ -0,0 +1,2 @@
+export * from "./constants.js";
+export * from "./utils.js";

package/dist/providers/{googleNativeGemini3.d.ts → googleNativeGemini3/utils.d.ts}

@@ -8,8 +8,8 @@
  * This module extracts the functions that are duplicated between the two
  * providers so they can share a single implementation.
  */
-import type { GenerateStopReason, ThinkingConfig, ChatMessage, CollectedChunkResult, MinimalChatMessage, NativeFunctionCall, NativeFunctionResponse, NativeToolDeclarationsResult, NativeToolsConfig, TextChannel, VertexNativePart, GeminiMultimodalInput } from "../types/index.js";
-import type { Tool } from "../types/index.js";
+import type { GenerateStopReason, ThinkingConfig, ChatMessage, CollectedChunkResult, MinimalChatMessage, NativeFunctionCall, NativeFunctionResponse, NativeToolDeclarationsResult, NativeToolsConfig, TextChannel, VertexNativePart, GeminiMultimodalInput } from "../../types/index.js";
+import type { Tool } from "../../types/index.js";
 /**
  * A per-turn tool execute map that deduplicates identical tool calls.
  *