Apache Kafka + CaptchaAI: Streaming CAPTCHA Task Processing

When CAPTCHA solving volume reaches thousands of tasks per hour, you need more than a simple queue. Apache Kafka provides durable, ordered, high-throughput message streaming — ideal for decoupling CAPTCHA task submission from result processing at scale.

Architecture

[Scrapers] → Produce → [Kafka: captcha-tasks topic]
                              ↓
                    [CAPTCHA Worker Group]
                    (consume tasks, solve via CaptchaAI)
                              ↓
                    Produce → [Kafka: captcha-results topic]
                              ↓
                    [Result Consumer Group]
                    (process solutions, update database)

Two Kafka topics separate concerns:

captcha-tasks — CAPTCHA parameters waiting to be solved
captcha-results — Solved tokens ready for downstream use

Prerequisites

# Python
pip install kafka-python requests

# Node.js
npm install kafkajs axios

Kafka broker running on localhost:9092 (or your cluster address).

Step 1: Create Topics

kafka-topics.sh --create --topic captcha-tasks \
  --partitions 6 --replication-factor 1 \
  --bootstrap-server localhost:9092

kafka-topics.sh --create --topic captcha-results \
  --partitions 6 --replication-factor 1 \
  --bootstrap-server localhost:9092

Six partitions allow up to six parallel consumers per group.

Step 2: Task Producer (Scraper Side)

Python

import json
from kafka import KafkaProducer

producer = KafkaProducer(
    bootstrap_servers=["localhost:9092"],
    value_serializer=lambda v: json.dumps(v).encode("utf-8"),
    key_serializer=lambda k: k.encode("utf-8") if k else None,
    acks="all",  # Wait for all replicas to confirm
    retries=3
)


def enqueue_captcha(task_id, sitekey, pageurl, captcha_type="userrecaptcha"):
    """Send a CAPTCHA task to Kafka."""
    task = {
        "task_id": task_id,
        "method": captcha_type,
        "sitekey": sitekey,
        "pageurl": pageurl,
        "submitted_at": __import__("time").time()
    }

    future = producer.send(
        "captcha-tasks",
        key=task_id,  # Key ensures same task goes to same partition
        value=task
    )
    future.get(timeout=10)  # Block until confirmed
    return task_id


# Submit tasks
enqueue_captcha("task_001", "6Le-wvkSAAAAAPBMRTvw0Q4Muexq9bi0DJwx_mJ-", "https://example.com")
enqueue_captcha("task_002", "6Le-wvkSAAAAAPBMRTvw0Q4Muexq9bi0DJwx_mJ-", "https://example.com")
producer.flush()

JavaScript

const { Kafka } = require("kafkajs");

const kafka = new Kafka({
  clientId: "captcha-producer",
  brokers: ["localhost:9092"],
});

const producer = kafka.producer();

async function enqueueCaptcha(taskId, sitekey, pageurl) {
  await producer.connect();

  const task = {
    task_id: taskId,
    method: "userrecaptcha",
    sitekey: sitekey,
    pageurl: pageurl,
    submitted_at: Date.now(),
  };

  await producer.send({
    topic: "captcha-tasks",
    messages: [{ key: taskId, value: JSON.stringify(task) }],
  });
}

(async () => {
  await enqueueCaptcha(
    "task_001",
    "6Le-wvkSAAAAAPBMRTvw0Q4Muexq9bi0DJwx_mJ-",
    "https://example.com"
  );
  await producer.disconnect();
})();

Step 3: CAPTCHA Worker (Consumer + Solver)

Python

import json
import os
import time
import requests
from kafka import KafkaConsumer, KafkaProducer

API_KEY = os.environ["CAPTCHAAI_API_KEY"]

consumer = KafkaConsumer(
    "captcha-tasks",
    bootstrap_servers=["localhost:9092"],
    group_id="captcha-workers",
    value_deserializer=lambda m: json.loads(m.decode("utf-8")),
    auto_offset_reset="earliest",
    enable_auto_commit=False,  # Manual commit after processing
    max_poll_records=10
)

result_producer = KafkaProducer(
    bootstrap_servers=["localhost:9092"],
    value_serializer=lambda v: json.dumps(v).encode("utf-8")
)


def solve_captcha(task):
    """Submit to CaptchaAI and poll for result."""
    # Submit
    resp = requests.post("https://ocr.captchaai.com/in.php", data={
        "key": API_KEY,
        "method": task["method"],
        "googlekey": task["sitekey"],
        "pageurl": task["pageurl"],
        "json": 1
    })
    data = resp.json()

    if data.get("status") != 1:
        return {"error": data.get("request")}

    captcha_id = data["request"]

    # Poll for result
    for _ in range(60):
        time.sleep(5)
        result = requests.get("https://ocr.captchaai.com/res.php", params={
            "key": API_KEY,
            "action": "get",
            "id": captcha_id,
            "json": 1
        }).json()

        if result.get("status") == 1:
            return {"solution": result["request"]}
        if result.get("request") != "CAPCHA_NOT_READY":
            return {"error": result.get("request")}

    return {"error": "TIMEOUT"}


# Main consumer loop
print("CAPTCHA worker started. Waiting for tasks...")
for message in consumer:
    task = message.value
    print(f"Processing {task['task_id']}...")

    result = solve_captcha(task)
    result["task_id"] = task["task_id"]
    result["solved_at"] = time.time()

    # Publish result
    result_producer.send("captcha-results", value=result)
    result_producer.flush()

    # Commit offset after successful processing
    consumer.commit()
    print(f"  → {task['task_id']}: {'solved' if 'solution' in result else result.get('error')}")

JavaScript

const { Kafka } = require("kafkajs");
const axios = require("axios");

const API_KEY = process.env.CAPTCHAAI_API_KEY;

const kafka = new Kafka({
  clientId: "captcha-worker",
  brokers: ["localhost:9092"],
});

const consumer = kafka.consumer({ groupId: "captcha-workers" });
const producer = kafka.producer();

function sleep(ms) {
  return new Promise((resolve) => setTimeout(resolve, ms));
}

async function solveCaptcha(task) {
  const submitResp = await axios.post(
    "https://ocr.captchaai.com/in.php",
    null,
    {
      params: {
        key: API_KEY,
        method: task.method,
        googlekey: task.sitekey,
        pageurl: task.pageurl,
        json: 1,
      },
    }
  );

  if (submitResp.data.status !== 1) {
    return { error: submitResp.data.request };
  }

  const captchaId = submitResp.data.request;

  for (let i = 0; i < 60; i++) {
    await sleep(5000);
    const result = await axios.get("https://ocr.captchaai.com/res.php", {
      params: { key: API_KEY, action: "get", id: captchaId, json: 1 },
    });

    if (result.data.status === 1) return { solution: result.data.request };
    if (result.data.request !== "CAPCHA_NOT_READY")
      return { error: result.data.request };
  }

  return { error: "TIMEOUT" };
}

async function run() {
  await consumer.connect();
  await producer.connect();
  await consumer.subscribe({ topic: "captcha-tasks", fromBeginning: false });

  await consumer.run({
    eachMessage: async ({ message }) => {
      const task = JSON.parse(message.value.toString());
      console.log(`Processing ${task.task_id}...`);

      const result = await solveCaptcha(task);
      result.task_id = task.task_id;
      result.solved_at = Date.now();

      await producer.send({
        topic: "captcha-results",
        messages: [{ value: JSON.stringify(result) }],
      });

      console.log(
        `  → ${task.task_id}: ${result.solution ? "solved" : result.error}`
      );
    },
  });
}

run();

Scaling Workers

Kafka consumer groups automatically distribute partitions across workers:

# 6 partitions, 3 workers → each worker gets 2 partitions
Worker-1: partitions 0, 1
Worker-2: partitions 2, 3
Worker-3: partitions 4, 5

# Add Worker-4 → rebalance
Worker-1: partitions 0, 1
Worker-2: partitions 2
Worker-3: partitions 3, 4
Worker-4: partition 5

Scale up to the number of partitions. Beyond that, add more partitions.

Monitoring

Track key metrics via Kafka consumer lag:

kafka-consumer-groups.sh --describe --group captcha-workers \
  --bootstrap-server localhost:9092

Metric	Healthy	Warning
Consumer lag	< 100	> 1000 (add workers)
Messages/sec in	Matches scraper rate	Spikes indicate burst
Messages/sec out	Matches in rate	Falling behind = bottleneck

Troubleshooting

Issue	Cause	Fix
Consumer lag growing	Workers can't keep up with task rate	Add more worker instances (up to partition count)
Duplicate results	Worker crashes before committing offset	Add idempotency check on task_id in result consumer
Rebalancing too frequently	Workers crashing/restarting	Increase `session.timeout.ms`; check for OOM
Tasks not distributed evenly	Poor key distribution	Use randomized keys or more partitions

FAQ

Why Kafka instead of Redis or RabbitMQ?

Kafka is ideal when you need message durability (replay capability), high throughput (100K+ messages/sec), and consumer group scaling. For simpler setups under 1,000 tasks/hour, Redis or RabbitMQ is sufficient.

Should I use one topic or two?

Two topics (tasks + results) cleanly separate producers and consumers. The task producer doesn't need to know about result consumers, and vice versa.

How do I handle poison messages (unsolvable CAPTCHAs)?

Set a retry limit in the worker. After max retries, publish to a captcha-dead-letter topic for manual inspection. Don't block the partition with infinite retries.