@almadar/llm 1.0.16 → 2.0.0

package/dist/rate-limiter-9XAWfHwe.d.ts

@@ -1,98 +0,0 @@
-/**
- * Token Tracker for LLM Usage
- *
- * Tracks token usage across multiple LLM calls for:
- * - Cost estimation
- * - Usage monitoring
- * - Quota management
- *
- * @packageDocumentation
- */
-interface TokenUsage {
-    promptTokens: number;
-    completionTokens: number;
-    totalTokens: number;
-    callCount: number;
-}
-declare class TokenTracker {
-    private model;
-    private usage;
-    constructor(model?: string);
-    addUsage(promptTokens: number, completionTokens: number): void;
-    getSummary(): TokenUsage;
-    getEstimatedCost(): number;
-    getFormattedCost(): string;
-    getReport(): string;
-    reset(): void;
-    setModel(model: string): void;
-}
-declare function getGlobalTokenTracker(model?: string): TokenTracker;
-declare function resetGlobalTokenTracker(): void;
-
-/**
- * Rate Limiter for LLM API Calls
- *
- * Implements token bucket algorithm with:
- * - Configurable requests per minute/second
- * - Automatic backoff on 429 errors
- * - Queue for pending requests
- *
- * @packageDocumentation
- */
-interface RateLimiterOptions {
-    /** Maximum requests per minute (default: 60) */
-    requestsPerMinute?: number;
-    /** Maximum requests per second (default: 3) */
-    requestsPerSecond?: number;
-    /** Maximum concurrent requests (default: 5) */
-    maxConcurrent?: number;
-    /** Base delay for exponential backoff in ms (default: 1000) */
-    baseBackoffMs?: number;
-    /** Maximum backoff delay in ms (default: 60000) */
-    maxBackoffMs?: number;
-}
-/**
- * Rate limiter for LLM API calls using token bucket algorithm.
- *
- * @example
- * ```typescript
- * const limiter = new RateLimiter({ requestsPerMinute: 30 });
- * const result = await limiter.execute(() => llm.invoke(messages));
- * ```
- */
-declare class RateLimiter {
-    private requestsPerMinute;
-    private requestsPerSecond;
-    private maxConcurrent;
-    private baseBackoffMs;
-    private maxBackoffMs;
-    private minuteTokens;
-    private secondTokens;
-    private activeRequests;
-    private queue;
-    private lastMinuteReset;
-    private lastSecondReset;
-    private processing;
-    private currentBackoffMs;
-    constructor(options?: RateLimiterOptions);
-    execute<T>(fn: () => Promise<T>, _maxRetries?: number): Promise<T>;
-    getStatus(): {
-        queueLength: number;
-        activeRequests: number;
-        minuteTokens: number;
-        secondTokens: number;
-        backoffMs: number;
-    };
-    reset(): void;
-    private processQueue;
-    private refillTokens;
-    private canMakeRequest;
-    private consumeTokens;
-    private getWaitTime;
-    private isRateLimitError;
-    private sleep;
-}
-declare function getGlobalRateLimiter(options?: RateLimiterOptions): RateLimiter;
-declare function resetGlobalRateLimiter(): void;
-
-export { RateLimiter as R, TokenTracker as T, type RateLimiterOptions as a, type TokenUsage as b, getGlobalTokenTracker as c, resetGlobalTokenTracker as d, getGlobalRateLimiter as g, resetGlobalRateLimiter as r };

package/dist/structured-output.d.ts

@@ -1,113 +0,0 @@
-import { a as RateLimiterOptions, b as TokenUsage } from './rate-limiter-9XAWfHwe.js';
-import { z } from 'zod';
-
-/**
- * JSON Schema type used for OpenAI structured outputs.
- */
-interface JsonSchema {
-    type?: string | string[];
-    properties?: Record<string, JsonSchema>;
-    required?: string[];
-    items?: JsonSchema;
-    enum?: unknown[];
-    const?: unknown;
-    anyOf?: JsonSchema[];
-    oneOf?: JsonSchema[];
-    allOf?: JsonSchema[];
-    $ref?: string;
-    $defs?: Record<string, JsonSchema>;
-    definitions?: Record<string, JsonSchema>;
-    additionalProperties?: boolean | JsonSchema;
-    description?: string;
-    default?: unknown;
-    minItems?: number;
-    maxItems?: number;
-    minLength?: number;
-}
-interface StructuredOutputOptions {
-    model?: string;
-    temperature?: number;
-    maxTokens?: number;
-    rateLimiter?: RateLimiterOptions;
-    useGlobalRateLimiter?: boolean;
-    trackTokens?: boolean;
-}
-interface StructuredGenerationOptions {
-    /** User's natural language request */
-    userRequest: string;
-    /** Model to use (overrides client default) */
-    model?: string;
-    /** Temperature (overrides client default) */
-    temperature?: number;
-    /** Maximum tokens (overrides client default) */
-    maxTokens?: number;
-    /** JSON Schema for structured output */
-    jsonSchema?: JsonSchema;
-    /** Schema name for the json_schema response format */
-    schemaName?: string;
-    /** System prompt override */
-    systemPrompt?: string;
-    /** System prompt builder function (called dynamically) */
-    buildSystemPrompt?: () => string;
-    /** Additional system prompt instructions */
-    additionalInstructions?: string;
-    /** Existing context for updates (e.g., existing schema JSON) */
-    existingContext?: string;
-    /** Skip post-generation validation (default: false) */
-    skipValidation?: boolean;
-}
-interface StructuredGenerationResult<T = unknown> {
-    /** Generated data (guaranteed to match JSON Schema structure) */
-    data: T;
-    /** Raw JSON string from API */
-    raw: string;
-    /** Token usage statistics */
-    usage: {
-        promptTokens: number;
-        completionTokens: number;
-        totalTokens: number;
-    };
-    /** Generation latency in milliseconds */
-    latencyMs: number;
-    /** Model used for generation */
-    model: string;
-    /** Zod validation result (if not skipped) */
-    zodValidation?: {
-        success: boolean;
-        errors?: z.ZodError['errors'];
-    };
-}
-declare const STRUCTURED_OUTPUT_MODELS: {
-    readonly GPT5_MINI: "gpt-5-mini";
-    readonly GPT4O_MINI: "gpt-4o-mini";
-    readonly GPT4O: "gpt-4o";
-    readonly GPT4O_2024_08_06: "gpt-4o-2024-08-06";
-};
-declare class StructuredOutputClient {
-    private openai;
-    private rateLimiter;
-    private tokenTracker;
-    private defaultModel;
-    private defaultTemperature;
-    private defaultMaxTokens;
-    constructor(options?: StructuredOutputOptions);
-    private usesMaxCompletionTokens;
-    /**
-     * Generate structured output with guaranteed JSON Schema compliance.
-     */
-    generate<T = unknown>(options: StructuredGenerationOptions): Promise<StructuredGenerationResult<T>>;
-    getModel(): string;
-    getRateLimiterStatus(): {
-        queueLength: number;
-        activeRequests: number;
-        minuteTokens: number;
-        secondTokens: number;
-        backoffMs: number;
-    };
-    getTokenUsage(): TokenUsage | null;
-}
-declare function getStructuredOutputClient(options?: StructuredOutputOptions): StructuredOutputClient;
-declare function resetStructuredOutputClient(): void;
-declare function isStructuredOutputAvailable(): boolean;
-
-export { type JsonSchema, STRUCTURED_OUTPUT_MODELS, type StructuredGenerationOptions, type StructuredGenerationResult, StructuredOutputClient, type StructuredOutputOptions, getStructuredOutputClient, isStructuredOutputAvailable, resetStructuredOutputClient };