PubNub MCP Server

import { get_encoding, get_encoding_name_for_model, type Tiktoken } from "tiktoken"; import { afterAll, beforeAll, describe, expect, it } from "vitest"; import { getChatSdkDocumentation, getSdkDocumentation } from "./api"; import { chatSdkLanguageToFeatures, sdkLanguageToFeatures } from "./schemas"; const TOKEN_LIMIT = 20_000; let encoding: Tiktoken; function calculateTokenCounts(text: string) { const openAiTokens = encoding.encode(text).length; return { openAiTokens, } as const; } async function verifyDocumentationTokens<L extends string, F extends string>( label: string, mapping: Record<string, readonly string[]>, fetcher: (language: L, feature: F) => Promise<{ content: string }> ) { const failures: string[] = []; for (const [language, features] of Object.entries(mapping)) { for (const feature of features) { const docs = await fetcher(language as L, feature as F); expect(docs?.content, `${label} ${language}/${feature} returned empty content`).toBeTruthy(); const { openAiTokens } = calculateTokenCounts(docs.content); if (openAiTokens > TOKEN_LIMIT) { failures.push( `${label} ${language}/${feature} exceeds OpenAI token limit: ${openAiTokens} > ${TOKEN_LIMIT}` ); } } } const failureMessage = failures.length > 0 ? `Token limit violations detected:\n${failures.join("\n")}` : undefined; expect(failures, failureMessage).toHaveLength(0); } describe("Docs API token limits", () => { beforeAll(() => { encoding = get_encoding(get_encoding_name_for_model("gpt-5")); }); afterAll(() => { encoding.free(); }); it( "ensures all SDK documentation responses stay within token limits", { timeout: 120_000 }, async () => { await verifyDocumentationTokens("SDK", sdkLanguageToFeatures, getSdkDocumentation); } ); it( "ensures all Chat SDK documentation responses stay within token limits", { timeout: 120_000 }, async () => { await verifyDocumentationTokens( "Chat SDK", chatSdkLanguageToFeatures, getChatSdkDocumentation ); } ); });