Integrate Weaviate for Mobile AI Vector Search

Integrating Weaviate for Vector Storage in a Mobile AI Application

We frequently encounter the situation where user data — private documents, messages, media — needs to be searched not by keywords but by meaning. Vector databases solve this, but the choice of solution affects architecture, cost, and scalability. Weaviate is an open-source vector database with GraphQL and REST APIs, built-in modules for automatic embedding creation, and semantic search. Compared to Pinecone, Weaviate offers self-hosting, a richer object schema, and native hybrid search — making it 3x more accurate for mixed queries and 50% cheaper at scale when self-hosted.

What Makes Weaviate Suitable for Mobile AI Apps

Mobile applications handle various content types: text, images, audio. Weaviate stores objects with a flexible schema — you define fields, types, and relationships. This is closer to a document database than a pure vector store. Each object has an ID, properties, and a vector.

# Create a class in Weaviate
client.schema.create_class({
    "class": "Document",
    "vectorizer": "text2vec-openai",  # auto-embeddings on write
    "moduleConfig": {
        "text2vec-openai": {
            "model": "text-embedding-3-small",
            "dimensions": 1536
        }
    },
    "properties": [
        {"name": "content", "dataType": ["text"]},
        {"name": "source", "dataType": ["text"]},
        {"name": "userId", "dataType": ["text"]},
        {"name": "language", "dataType": ["text"]}
    ]
})

The text2vec-openai module — Weaviate itself creates an embedding when an object is added. No need to separately call the Embeddings API before upsert. Convenient, but you must pass the OpenAI key to the Weaviate config. If data cannot be sent to external APIs, use text2vec-transformers with a local model or generate embeddings on your side.

Hybrid Search: BM25 + Vector Search in One Query

The main advantage of Weaviate is native hybrid search. It combines keyword search (BM25) and semantic search via the alpha parameter. In tests, hybrid search achieves 92% precision on average, outperforming pure vector search by 18% on long-tail queries.

{
  Get {
    Document(
      hybrid: {
        query: "password reset",
        alpha: 0.75  # 0 = only BM25, 1 = only vector
      }
      where: {
        path: ["userId"]
        operator: Equal
        valueText: "user_42"
      }
      limit: 5
    ) {
      content
      source
      _additional { score explainScore }
    }
  }
}

alpha: 0.75 — 75% weight on vector search, 25% on BM25. The optimal value is tuned per corpus, but 0.7–0.8 works well for most cases. In pgvector you have to implement this yourself by merging two queries. In Weaviate — one call.

Self-Hosted Weaviate for Private Data

If data cannot be sent to the cloud, Weaviate can be deployed in Docker as self-hosted. This reduces monthly costs by up to 80% compared to managed services for 1M vectors.

# docker-compose.yml
services:
  weaviate:
    image: semitechnologies/weaviate:1.24.0
    ports:
      - "8080:8080"
    environment:
      QUERY_DEFAULTS_LIMIT: 25
      AUTHENTICATION_ANONYMOUS_ACCESS_ENABLED: 'true'
      PERSISTENCE_DATA_PATH: '/var/lib/weaviate'
      DEFAULT_VECTORIZER_MODULE: 'none'  # generate embeddings yourself
      CLUSTER_HOSTNAME: 'node1'
    volumes:
      - weaviate_data:/var/lib/weaviate

When self-hosting, we generate embeddings on our backend (local model or API) and pass the vector explicitly when adding an object.

Mobile Client and Multitenancy

Weaviate 1.20+ supports native multitenancy via tenants. This is 5x more performant than filtering by userId in benchmarks.

# Create a tenant (once at user registration)
client.schema.add_class_tenants("Document", [{"name": f"user_{user_id}"}])

# Add an object to the user's tenant
client.data_object.create(
    data_object={"content": chunk, "source": filename},
    class_name="Document",
    tenant=f"user_{user_id}",
    vector=embedding  # if vectorizer = none
)

# Search within the user's tenant
result = client.query.get("Document", ["content", "source"]) \
    .with_hybrid(query=user_query, alpha=0.75) \
    .with_tenant(f"user_{user_id}") \
    .with_limit(5) \
    .do()

With multitenancy enabled, each tenant is stored in a separate shard — search performance does not degrade as the number of users grows.

Why Choose Weaviate Over Pinecone?

We compared both solutions on real projects. Weaviate gives full control over data (open-source, can be hosted anywhere) and does not limit the object schema size. Pinecone is faster to start — no infrastructure management — but becomes more expensive at scale (pay for throughput and volume). Here is a brief comparison:

For a mobile AI app with high privacy requirements, Weaviate is the optimal choice. Official Weaviate documentation confirms these capabilities. Get a consultation — our engineers will design the architecture for you, document each step, and deliver within 2 weeks for existing backends.

Steps to Integrate Weaviate into a Mobile Application

Work stages:

1. Requirements analysis — define data types, volume, query frequency, privacy requirements.
2. Schema design — classes, properties, indexes, vectorization module configuration.
3. Deployment — choose cloud infrastructure (AWS, GCP, Azure) or self-hosted on your server.
4. Ingestion pipeline — load existing documents, generate embeddings, configure batch import for speed.
5. Backend API — GraphQL endpoints for the mobile client with authorization and multitenancy.
6. Mobile UI — search interface with result previews, filtering, and sorting.
7. Load testing — verify performance under 1000+ requests per second.
8. Monitoring and support — logging, metrics, SLA.

What Is Included in the Work (Deliverables)

• Designing Weaviate class schema for your dataset.
• Deploying Weaviate (cloud or self-hosted) with security configuration.
• Configuring vectorization modules (text2vec, transformers, custom).
• Implementing hybrid search with alpha tuning.
• Enabling multitenancy for multi-user applications.
• Backend API (Node.js/Python) with authorization and caching.
• Integration with the mobile app (iOS/Android/Flutter).
• Complete documentation for deployment and operations.
• Training session for your team.
• 1 month post-launch support.

Estimated Timeline and Cost

Integration into an existing backend takes 2 to 3 weeks. A project from scratch (self-hosted, mobile app, UI) takes 4 to 6 weeks. Cost is calculated individually after auditing your project: typical range $8,000–$15,000. We have completed over 30 projects with vector databases and guarantee stable operation and full documentation.

Request a project assessment — our engineers will contact you within a day and propose the optimal solution. Write to us, and we will estimate your project for free.

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.