VK Chat Bot Development: Turnkey Solutions

Imagine your business losing up to 30% of leads daily because managers can't respond in time on messengers. A VK chat bot solves this—processing orders, answering questions, and forwarding complex requests to operators. For example, for a coffee chain we implemented a bot that takes orders, integrates with CRM, and reduced processing time by 60%. This helped increase conversion by 25%. Over the years, we have developed more than 50 such bots for e-commerce, services, and media, each handling up to 10,000 messages per minute. Savings on operational costs amount to tens of thousands of rubles monthly.

We use a modern stack: Python FastAPI, vk_api, JavaScript for Mini Apps. We'll evaluate your project in one day—just contact us.

VK Chat Bot Development: Features and Tech Stack

How to Choose Between Callback API and Long Poll?

Callback API—VK sends a POST to your HTTPS server for each event. It's a webhook analog, suitable for production.

Long Poll—your server makes a GET request that "hangs" until a new event appears. Works without HTTPS, convenient for development.

Comparison:

For production bots, we always use Callback API. It ensures guaranteed delivery and automatic retries on failures.

Callback API is configured in the "Management" → "Working with API" → "Callback API" section of the VK group. When adding, VK sends a confirmation event—you must return the string from settings:

from fastapi import FastAPI, Request

app = FastAPI()
CONFIRMATION_TOKEN = "abc123xyz"  # From group settings
SECRET_KEY = "your_secret"        # For signature verification

@app.post("/vk/webhook")
asy nc def vk_webhook(request: Request):
    data = await request.json()

    # Signature check
    if data.get("secret") != SECRET_KEY:
        return "forbidden"

    if data["type"] == "confirmation":
        return CONFIRMATION_TOKEN

    if data["type"] == "message_new":
        message = data["object"]["message"]
        await handle_message(message)

    return "ok"  # VK requires exactly the string "ok"

If you return anything other than "ok", VK will resend the event up to 3 times, then mark delivery as failed.

According to VK API documentation, Callback API guarantees delivery with retries.

How to Send Messages with a Keyboard?

import vk_api
from vk_api.bot_longpoll import VkBotLongPoll, VkBotEventType

# Through vk_api library
vk_session = vk_api.VkApi(token=GROUP_TOKEN)
vk = vk_session.get_api()

def send_message(peer_id: int, text: str, keyboard=None):
    params = {
        "peer_id": peer_id,
        "message": text,
        "random_id": 0  # 0 = auto random_id for duplicate protection
    }
    if keyboard:
        params["keyboard"] = json.dumps(keyboard)
    vk.messages.send(**params)

random_id is important: VK deduplicates messages by (peer_id, random_id). With random_id=0 it's automatic. With a fixed value, a repeat call won't create a new message—protection against double sending.

Keyboards come in two types: regular (replaces system keyboard) and inline (attached to a message). Example inline keyboard:

keyboard = {
    "inline": True,
    "buttons": [
        [
            {
                "action": {
                    "type": "text",
                    "label": "Catalog",
                    "payload": json.dumps({"command": "catalog"})
                },
                "color": "primary"
            }
        ]
    ]
}

Button colors: primary, secondary, positive, negative. Maximum 4 buttons per row, 10 rows.

What are VK Mini Apps and How to Integrate Them?

VK Mini Apps are web applications inside VK, similar to Telegram Mini Apps. SDK: @vkontakte/vk-bridge.

import bridge from '@vkontakte/vk-bridge';

bridge.subscribe((e) => {
    if (e.detail.type === 'VKWebAppUpdateConfig') {
        // Theme (light/dark), color scheme
        const scheme = e.detail.data.scheme;
        document.body.setAttribute('scheme', scheme);
    }
});

// Get user data
const userInfo = await bridge.send('VKWebAppGetUserInfo');
// { id, first_name, last_name, photo_200, ... }

vk-bridge allows: get user data, open payment dialog (VKWebAppOpenPayForm), request geolocation, copy to clipboard, open QR scanner.

Mini App authorization: launch_params in URL contain signed data (sign field). Verification on server via HMAC-SHA256 with API Secret from app settings—mandatory.

Common Development Errors

1. Incorrect request verification. If you don't check secret or don't return "ok", VK will repeat requests and eventually disable the callback.
2. Incorrect random_id. Using a fixed value can cause message loss. Best to always pass 0 for automatic generation.
3. Ignoring keyboard limits. Max 4 buttons per row and 10 rows. Exceeding returns an error.
4. Lack of API error handling. When sending a message, VK may return an error, e.g., if the user has disabled messages from the group.

On every project, we conduct code audits and test these cases. Want to avoid these errors? Contact us—we'll help.

Work Process and Timelines

Typical timeline: bot with keyboard—1–2 weeks, with VK Mini App—4–8 weeks.

What's Included in Development?

• VK group setup and Callback API configuration
• Message handling logic (commands, scenarios)
• Keyboards and carousels implementation
• Integration with external services (CRM, databases)
• VK Mini App development (if required)
• Operation documentation
• Team training
• 1 month free technical support

Why Choose Us?

• Years of experience developing bots for VK and other platforms
• Numerous successful projects for small and medium businesses
• Certified VK API developers
• Guarantee of stable operation and adherence to deadlines

Development cost is discussed individually, but on average it pays off within six months through automation. Request a consultation—we'll evaluate your project in one day.

Machine Learning in Mobile Apps: CoreML, TFLite, and On-Device Models

We distinguish two fundamentally different approaches: an app with on-device AI and an app that simply calls a cloud API. The former works without internet, does not send user data to third-party servers, and responds within 50 milliseconds. The latter depends on network latency and pricing plans. Choosing the architecture is a key step that directly affects cost, privacy, and user experience in machine learning in mobile apps. Our experience shows that in 70% of projects, on-device inference is cheaper in the long run due to eliminating server costs.

How to Choose Between CoreML and TFLite for On-Device Inference?

CoreML — Apple's native framework for running ML models on device. Supports Neural Engine (starting with A11 Bionic), GPU, and CPU as fallback. Models are converted to .mlmodel format via coremltools from PyTorch, ONNX, or TensorFlow. Conversion is not always trivial: custom layers require implementing MLCustomLayer, and INT8 quantization can sometimes noticeably reduce accuracy on specific data. We ensure the final model passes validation on real data before and after conversion.

TensorFlow Lite — cross-platform alternative for Android and Flutter. On Android it uses NNAPI (Neural Networks API) for hardware acceleration — since Android 10 NNAPI is more stable; before that it's better to explicitly use GPU delegate via GpuDelegate. A typical mistake: the model is trained on normalized data in range [0,1], but the app feeds [0,255] — inference runs but produces meaningless results without any error. We include an automatic input data validation module in the SDK.

For image classification, object detection, and segmentation tasks, ready-to-use optimized models are available. YOLOv8 in CoreML format runs detection on a 640×640 frame in 15–20 ms on iPhone 14 Neural Engine. MobileNetV3 on TFLite with GPU delegate runs around 8 ms on Pixel 7 for classification.

What If You Need Text Generation On-Device?

Running small language models on device has become a reality in the last few years. Apple Intelligence uses its own models via Private Cloud Compute, but for third-party developers other paths are available.

llama.cpp with Metal backend on iOS is a working approach for phi-3-mini (3.8B parameters, 4-bit quantization, ~2.3 GB). Inference: 15–25 tokens/second on iPhone 15 Pro. For integration in Swift, use the Swift Package llama.swift or a wrapper via C interface llama.h. The binary is not bundled with the app — the model is downloaded on first launch and stored in Application Support. Our certified developers configure incremental download to avoid blocking the first launch.

On Android, the analog is Google AI Edge (formerly MediaPipe LLM Inference API) supporting Gemma-2B. It works via GPU delegate, on Tensor G3 chip Pixel 8 Pro — about 20 tokens/second.

Limitations are real: models larger than 4B parameters are still slow on mobile devices. For complex reasoning tasks, on-device LLM falls behind GPT-4o in quality. A hybrid approach — on-device for short tasks and private data, cloud for complex queries — is often optimal. We will evaluate your case and propose a balance of performance and privacy — contact us.

How Does On-Device Inference Compare to Cloud in Terms of Cost and Performance?

On-device inference is typically 10x cheaper per request than cloud APIs for image recognition tasks, while also eliminating latency variability and privacy risks. The table below summarizes the trade-offs.

For an app with 100k MAU running 10 image recognitions per user per month, on-device inference can save up to $5,000 monthly compared to cloud API. Get a free consultation on your ML architecture today.

Integrating OpenAI API and Other Cloud Models

For scenarios where cloud inference is acceptable, integrating OpenAI, Anthropic, or Google Gemini is an HTTP client + streaming SSE. In Swift, AsyncThrowingStream is convenient for streaming responses. In Kotlin, use Flow.

Critically: API keys must never be stored in the app bundle. Even an obfuscated key can be extracted from the IPA in 10 minutes using strings or frida. Correct architecture: mobile app → your own backend → OpenAI API. The backend controls rate limiting, logs requests, and protects the key.

What Is Included in the Work (Deliverables)

• Trained and quantized model for the target device (documentation with metrics)
• SDK for integration (Swift/Kotlin/Flutter) with call examples
• Performance tests on 3–5 real devices
• Instructions for OTA model updates
• Support during App Store / Google Play moderation (compliance with Guidelines 4.2, 5.1)
• 2 weeks of technical support after release

Typical Project Pipeline

1. Task analysis — measure latency, privacy, size, supported devices.
2. Model prototyping — in Python, evaluate accuracy on target data.
3. Conversion and quantization — for CoreML/TFLite with validation.
4. Integration into the app — model wrapped in a service layer (easy to swap CoreML ↔ TFLite ↔ cloud).
5. Testing — on real devices, measure FPS, RAM, battery.
6. Deployment — via TestFlight / Firebase App Distribution, monitor metrics.

Timelines: integration of a ready CoreML/TFLite model — 1–2 weeks, development of a custom model with mobile optimization — from 6 weeks, on-device LLM chat with personalization — 4–8 weeks.

Why We Take on Complex Cases?

10+ years of experience in mobile development, 50+ implemented AI/ML solutions, guarantee of compatibility with current iOS and Android versions. All projects undergo code review and load testing. The cost includes preparation of moderation documentation and training of your team.

Contact us — we will help you choose the architecture and implement ML in your app turnkey. Order an audit of your existing solution — we will assess the potential for server cost savings free of charge. In some projects, savings can reach significant amounts per month.