nforge 1.1.1

The ML API Framework - FastAPI for Machine Learning

🚀 NeuralForge

The AI-Native ML API Framework

Deploy ML models to production in minutes, not weeks.

---

🎯 What is NeuralForge?

NeuralForge is the modern Python framework for deploying ML models to production.

Built specifically for ML Engineers and Data Scientists, NeuralForge provides everything you need to serve ML models at scale:

• ✅ Model Registry - Version and manage all your models
• ✅ A/B Testing - Safely roll out model improvements
• ✅ Monitoring - Track predictions and detect drift
• ✅ Auto-scaling - Handle traffic spikes automatically
• ✅ Production-ready - Security, logging, metrics built-in

Why NeuralForge?

FastAPI is great for general APIs. NeuralForge is built specifically for ML.

Feature	FastAPI	NeuralForge
Model Versioning	❌ Manual	✅ Built-in
A/B Testing	❌ Manual	✅ Built-in
Drift Detection	❌ Manual	✅ Built-in
Batch Prediction	❌ Manual	✅ Built-in
GPU Management	❌ Manual	✅ Built-in
ML Monitoring	❌ Manual	✅ Built-in

---

🚀 Quick Start

Installation

Basic Installation:

pip install nforge

With ML Optimization (PyTorch):

pip install nforge[pytorch]

Full Installation:

pip install nforge[all]

Note: PyTorch is optional. The core framework works perfectly without it. PyTorch is only needed for model quantization and GPU acceleration features. See docs/guides/OPTIONAL_DEPENDENCIES.md for details.

Your First ML API (5 minutes)

from neuralforge import NeuralForge
from pydantic import BaseModel

# Create app
app = NeuralForge()

# Define input/output
class PredictionInput(BaseModel):
    text: str

class PredictionOutput(BaseModel):
    sentiment: str
    confidence: float

# Create endpoint
@app.post("/predict")
async def predict(data: PredictionInput) -> PredictionOutput:
    # Your ML model here
    sentiment = "positive" if "good" in data.text.lower() else "negative"
    
    return PredictionOutput(
        sentiment=sentiment,
        confidence=0.95
    )

# Run
if __name__ == "__main__":
    app.run(port=8000)

Test it:

curl -X POST http://localhost:8000/predict \
  -H "Content-Type: application/json" \
  -d '{"text": "This is great!"}'

---

🎨 Key Features

1. Model Registry & Versioning

Manage multiple model versions with ease:

from neuralforge.ml import ModelRegistry

registry = ModelRegistry()

# Register models
@registry.register(name="sentiment", version="1.0.0")
class SentimentModelV1:
    def predict(self, text: str):
        return {"sentiment": "positive", "confidence": 0.9}

@registry.register(name="sentiment", version="2.0.0")
class SentimentModelV2:
    def predict(self, text: str):
        # Improved model
        return {"sentiment": "positive", "confidence": 0.95}

# Load specific version
model = await registry.load("sentiment", version="1.0.0")

2. Built-in A/B Testing

Test model improvements safely:

from neuralforge.ml import ABTest

# Create A/B test
ab_test = ABTest(
    name="sentiment-v2-test",
    variants={
        "control": {"version": "1.0.0", "traffic": 0.7},
        "treatment": {"version": "2.0.0", "traffic": 0.3}
    }
)

@app.post("/predict")
@ab_test.route()
async def predict(data: Input):
    # Automatically selects variant
    model = await ab_test.get_model()
    return model.predict(data)

3. Prediction Monitoring

Track model performance in production:

from neuralforge.ml import PredictionMonitor

monitor = PredictionMonitor(
    model_name="sentiment",
    track_latency=True,
    alert_on_degradation=True
)

@app.post("/predict")
@monitor.track()
async def predict(data: Input):
    # Automatically logged and monitored
    return model.predict(data)

4. Production Features Built-in

Everything you need for production:

• ✅ Authentication - API keys, JWT, OAuth2
• ✅ Rate Limiting - Protect your API
• ✅ CORS - Cross-origin requests
• ✅ Logging - Structured logging with masking
• ✅ Metrics - Prometheus-compatible
• ✅ Health Checks - Liveness and readiness probes

---

📚 Examples

Sentiment Analysis

# See examples/ml_api/main.py
# Full working sentiment analysis API with:
# - User authentication
# - Prediction history
# - Analytics
# - Rate limiting

Text Generation with Streaming

from neuralforge import NeuralForge
from neuralforge.streaming import SSEResponse, TokenStreamBuilder

app = NeuralForge()

@app.stream("/generate")
async def generate(prompt: str):
    builder = TokenStreamBuilder()
    
    # Stream tokens as they're generated
    async for token in llm.generate(prompt):
        yield builder.add_token(token)
    
    yield builder.finish()

gRPC Server

from neuralforge.grpc import GRPCServer, GRPCServicer, grpc_method

# Enable gRPC alongside HTTP
grpc_server = app.enable_grpc(port=50051)

class InferenceService(GRPCServicer):
    @grpc_method(response_streaming=True)
    async def StreamPredict(self, request, context):
        async for token in model.generate(request.prompt):
            yield TokenResponse(token=token)

grpc_server.register_service(InferenceService)

Benchmarking

# Run latency benchmarks
neuralforge benchmark run --scenario latency

# Compare against FastAPI/BentoML
neuralforge benchmark compare --against fastapi,bentoml

# Generate HTML report
neuralforge benchmark report --format html --output report.html

OpenTelemetry Tracing

from neuralforge.observability import TracingManager, configure_structured_logging

# Configure tracing
tracer = TracingManager(service_name="ml-api")
tracer.configure(exporter="otlp", endpoint="localhost:4317")

# Structured JSON logging
configure_structured_logging(
    service_name="ml-api",
    format_type="json"
)

---

🐳 Docker Deployment

One-Command Deployment

# Copy environment template
cp env.template .env

# Start everything (API + PostgreSQL + Redis)
docker compose up -d

# Check health
curl http://localhost:8000/health

See docs/deployment/DOCKER_QUICKSTART.md for details.

---

📚 Documentation

Quick Start

• Quick Start Guide - Get started in 5 minutes
• Production Quickstart - Production-ready setup

Core Documentation

• Production Features - All production features explained
• Core Functionalities - Framework capabilities
• Code Structure - Project organization
• Code Quality - Quality standards and best practices

Deployment

• Docker Deployment - Complete Docker guide
• Docker Quickstart - Quick Docker setup

Guides

• Testing Examples - Testing patterns
• Optional Dependencies - ML framework options

Contributing

• Contributing Guide - How to contribute
• Changelog - Version history

📖 Complete Documentation Index

---

🏗️ Architecture

NeuralForge
├── Core Framework
│   ├── Async Routing (FastAPI-like)
│   ├── Dependency Injection
│   ├── Database Integration (SQLAlchemy)
│   └── Middleware Stack
│
├── ML Features (v0.8+)
│   ├── Model Registry
│   ├── A/B Testing
│   ├── Prediction Monitoring
│   └── Drift Detection
│
└── Production Features
    ├── Authentication & Authorization
    ├── Rate Limiting
    ├── Logging & Metrics
    └── Docker Deployment

---

🧪 Development

Setup

git clone https://github.com/rockstream/neuralforge.git
cd neuralforge

# Create virtual environment
python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt
pip install -r requirements-dev.txt

# Run tests
pytest

Running Tests

# All tests
pytest

# With coverage
pytest --cov=neuralforge --cov-report=html

# Specific test
pytest tests/test_routing.py -v

Current Status: ✅ 117/117 tests passing

---

🗺️ Roadmap

✅ v0.5 (Current) - Core Framework

• Async routing with validation
• Dependency injection
• Database integration (SQLAlchemy)
• Middleware (CORS, Security, Logging, Rate Limiting)
• Docker deployment
• 117 tests passing

🔄 v0.8 (6 weeks) - ML Essentials

• Model Registry & Versioning
• A/B Testing Framework
• Prediction Monitoring
• 3+ ML Examples

🎯 v1.0 (12 weeks) - Production ML Framework

• Drift Detection
• Model Optimization
• GPU Support
• Auto-scaling
• Enterprise Features

---

🤝 Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

Areas we need help:

• 🐛 Bug fixes
• 📝 Documentation
• 🎨 ML examples
• 🔌 ML tool integrations
• ⚡ Performance optimization

---

📊 Comparison

NeuralForge vs FastAPI

Use FastAPI when:

• Building general-purpose APIs
• Need maximum flexibility
• Not ML-specific

Use NeuralForge when:

• Deploying ML models
• Need model versioning
• Want A/B testing built-in
• Need drift detection
• ML-specific features matter

NeuralForge vs BentoML

NeuralForge:

• ✅ General-purpose + ML-specific
• ✅ FastAPI-like developer experience
• ✅ More flexible

BentoML:

• ✅ ML-only focus
• ✅ More opinionated
• ✅ Mature ecosystem

---

🚀 Enterprise Agent System

Production-ready code quality and security analysis powered by NeuralForge Enterprise Agents.

🎯 Features

• Automated Quality Checks: Enterprise-grade code analysis on every PR
• Security Scanning: SAST, dependency scanning, and secret detection
• CI/CD Integration: GitHub Actions workflows with automated reporting
• Metrics & Monitoring: Prometheus-compatible metrics collection

📊 Current Status

• Enterprise Coverage: 75% (Code Quality + Security + Observability + Integration)
• Test Coverage: 71% (View Report)
• Security Scans: Automated via Security Agent v2

📚 Documentation

• Deployment Guide - Production deployment instructions
• Quick Reference - Daily CLI commands
• Enterprise Agent Docs - Technical documentation

🔧 Quick Start

# Run quality analysis
python .agent/cli.py analyze --quick

# Run security scan
python .agent/cli.py security

# View metrics
python .agent/cli.py metrics

Note: Replace YOUR_USERNAME in badge URLs with your GitHub username for live status badges.

---

📄 License

MIT License - see LICENSE for details.

---

🙏 Acknowledgments

• Inspired by FastAPI - Amazing DX
• Inspired by BentoML - ML serving patterns
• Built with Pydantic, SQLAlchemy, and Uvicorn

---

📞 Support

• Documentation: [Coming Soon]
• GitHub Issues: Report bugs
• Discord: Join community (Coming Soon)
• Email: [e-mail]

---

Built with ❤️ for ML Engineers

Get Started • Documentation • Examples