Skip to main content

API Reference (TypeScript)

The complete TypeScript API for LLM Context Forge. The API surface is intentionally identical in behavior to the Python SDK.

TokenCounter​

Provides exact token counting for specific model encodings.

import { TokenCounter } from 'llm-context-forge';

const counter = new TokenCounter("gpt-4o");

Methods​

count(text: string): number​

Calculates the exact token count.

fitsInWindow(text: string, reserveOutput: number = 0): boolean​

Returns true if the tokens for text + reserveOutput fit within the model's limit.

DocumentChunker​

Splits text into token-safe portions using specified strategies.

import { DocumentChunker, ChunkStrategy } from 'llm-context-forge';

const chunker = new DocumentChunker("claude-3-5-sonnet");

Options Interface​

interface ChunkOptions {
  maxTokens: number;
  overlapTokens?: number;
}

Methods​

chunk(text: string, strategy: ChunkStrategy, options: ChunkOptions): string[]​

Splits the string according to the strategy, ensuring no chunk exceeds maxTokens.

ContextWindow​

Provides a priority-based packing mechanism for assembling RAG prompts.

import { ContextWindow, Priority } from 'llm-context-forge';

const window = new ContextWindow("gpt-4o");

Methods​

addBlock(content: string, priority: Priority, blockId?: string): void​

Enqueues content. Passing a priority of 0 (Priority.CRITICAL) guarantees the content is included, or it throws a ContextOverflowError.

assemble(options?: { maxTokens?: number }): string​

Packs the prompt, honoring exact limits. If maxTokens is omitted, it defaults to the model's context window.

usage(): UsageStats​

Returns { tokensUsed: number, included: string[], excluded: string[] }.

CostCalculator​

Calculates pricing estimates for prompts and completions.

import { CostCalculator } from 'llm-context-forge';

const calc = new CostCalculator("gpt-4o");

Methods​

estimatePrompt(text: string): CostEstimate​

Returns { usd: number } for the given prompt based on exact token counting.