TypeScript SDK

The official Laghav TypeScript/Node.js SDK. Full type safety, streaming, async/await, and Edge Runtime support.

Installation

bash

npm install laghav
# types are bundled — no @types/laghav needed

LaghavClient

client.ts

import LaghavClient from 'laghav'
 
const client = new LaghavClient({
  apiKey: process.env.LAGHAV_API_KEY!,  // or auto-reads LAGHAV_API_KEY
  baseURL: 'https://api.laghav.ai',      // default
  timeout: 30000,                         // ms
  maxRetries: 2,
})
 
// Typed response
const response = await client.complete({
  messages: [{ role: 'user', content: '...' }],
  model: 'auto',
  maxTokens: 1000,
  laghavOptions: {
    compress: true,
    route: true,
    cache: true,
    score: true,
    maxAggressiveness: 0.7,
  },
})
 
// Fully typed
const meta = response.laghavMeta  // LaghavMeta type
console.log(meta.qualityScore)    // number
console.log(meta.savedUsd)        // number

Streaming

stream.ts

// Async iterator
const stream = client.complete({
  messages: [{ role: 'user', content: '...' }],
  model: 'auto',
  stream: true,
})
 
for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content ?? '')
}
 
// Helper: get final message with full laghav_meta
const final = await stream.finalMessage()
console.log(final.laghavMeta.savedUsd)

✦Next.js App Router & Edge Runtime

The SDK is compatible with Next.js Server Actions, API Routes, and Edge Runtime (Vercel, Cloudflare Workers). Use export const runtime = 'edge' in your route.

Type definitions

types.ts

import type {
  CompleteRequest,
  CompleteResponse,
  LaghavMeta,
  LaghavOptions,
  Message,
  StreamChunk,
} from 'laghav'
 
// LaghavMeta shape
interface LaghavMeta {
  originalTokens: number
  compressedTokens: number
  compressionRatio: number
  qualityScore: number
  costOriginalUsd: number
  costActualUsd: number
  savedUsd: number
  routingReason: string
  modelRequested: string
  rulesApplied: string[]
  cacheHit: boolean
  piiMasked: boolean
  latencyOverheadMs: number
  conversationId?: string
}

Python SDK LangChain