Developer API & Webhooks Documentation

All requests to the Pak-LLM API require a secret Bearer API key generated from the Developer Panel (/developer) or Admin Dashboard. Keys are prefixed with pak_live_....

Generates conversational completions with native Nastaliq subword compression. Supports real-time Server-Sent Events (SSE) streaming and deterministic tool calling.

{
  "model": "pak-llm-awam-v2",
  "messages": [
    {
      "role": "user",
      "content": "پاکستان میں آئی ٹی کمپنیوں پر ٹیکس استثنیٰ کی تفصیل بتائیں۔"
    }
  ],
  "temperature": 0.3
}

{
  "id": "chatcmpl_9f821a7c",
  "object": "chat.completion",
  "created": 1789456200,
  "model": "pak-llm-awam-v2",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "پاکستان میں PSEB کے ساتھ رجسٹرڈ 100% آئی ٹی برآمدی ترسیلات پر..."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 18,
    "completion_tokens": 74,
    "total_tokens": 92
  }
}

List available foundation models and generate semantic multilingual vector embeddings:

Webhooks push real-time JSON events to your specified HTTPS destination endpoint as soon as operations complete in Pak-LLM.

Every webhook payload is dispatched with cryptographic signature headers to ensure message authenticity and prevent tampering:

• x-pakllm-signature-256: sha256=HEX_DIGEST
• x-pakllm-timestamp: Unix epoch timestamp in milliseconds
• x-pakllm-delivery-id: Unique delivery trace identifier

import crypto from "crypto";

export function verifyPakLlmWebhook(rawBody: string, signatureHeader: string, secret: string): boolean {
  if (!signatureHeader || !signatureHeader.startsWith("sha256=")) return false;
  
  const expectedSignature = "sha256=" + crypto
    .createHmac("sha256", secret)
    .update(rawBody)
    .digest("hex");

  return crypto.timingSafeEqual(
    Buffer.from(signatureHeader),
    Buffer.from(expectedSignature)
  );
}

Because Pak-LLM follows standard OpenAI REST specifications, you can use official client libraries by overriding the baseURL parameter:

from openai import OpenAI

client = OpenAI(
    api_key="pak_live_your_key_here",
    base_url="https://pak-llm.com/api/v1",
)

response = client.chat.completions.create(
    model="pak-llm-awam-v2",
    messages=[{"role": "user", "content": "How are you?"}]
)
print(response.choices[0].message.content)

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "pak_live_your_key_here",
  baseURL: "https://pak-llm.com/api/v1",
});

const completion = await client.chat.completions.create({
  model: "pak-llm-awam-v2",
  messages: [{ role: "user", content: "Hello Pak-LLM!" }],
});
console.log(completion.choices[0].message.content);

Under our sovereign security policy, API request prompt bodies execute in volatile memory under strict Zero Data Retention (ZDR) agreements. No customer queries are used to train foundational models. All API keys are sealed using SHA-256 one-way hashes and secret vaults protected by NIST FIPS 203 Post-Quantum Cryptography (ML-KEM-768).