2.4 KiB

Raw Blame History

ML-API: Metrics Analysis & Recommendations Service

ML-API is a high-performance Rust service designed to process system metrics, analyze them using machine learning models, and provide actionable recommendations. The service transforms raw metric arrays into structured prompts, sends them to configured ML models, and returns system insights.

Key Features

Dual Interface Support: Choose between WebSocket or REST API for your integration needs
Flexible ML Backend Integration: Easily connect to any ML model API (Ollama, OpenAI, etc.)
Smart Prompt Generation: Automatically transforms metrics into optimized ML prompts
Comprehensive Analysis: Returns system health insights and improvement recommendations
Configurable Logging: Adjustable log levels for optimal debugging and monitoring

Quick Start

Prerequisites

Rust 1.70 or later
Configured ML backend (Ollama, OpenAI, etc.)

Configuration

Create a .env file in your project root:

ML_TARGET_URL="http://url.to/ml/api"
ML_MODEL_NAME="kis-test"
ML_REQUEST_TIMEOUT="10" 
ML_API_LOG_LEVEL="INFO"
ML_API_PORT="5134"

Installation

git clone https://github.com/your-repo/ml-api.git
cd ml-api
cargo build --release

Running the Service

cargo run --release

API Endpoints

WebSocket Interface

Endpoint: ws://your-server/api/method/ws
Protocol: Real-time metric processing with continuous feedback

REST API

POST Endpoint: http://your-server/api/method/rest
GET Endpoint: http://your-server/api/swagger (API documentation)

Request Format

Send your metrics as a JSON array in the following format:

{
  "service_name": "zvks",
  "metrics": [
    {
      "id": "10001",
      "name": "cpu_utilization",
      "description": "cpu_utilization",
      "value": 1.23,
      "status": 0,
      "device": 18,
      "source": "module$11"
    },
    // ... metrics ...
  ]
}

Response Format

The API returns comprehensive analysis and recommendations in PLAIN TEXT

Advanced Configuration

Logging Levels

Adjust the log granularity by changing ML_LOG_LEVEL in your .env:

TRACE: Full debug information (verbose)
DEBUG: Development-level logging
INFO: Standard operational messages (default)
WARN: Only warnings and errors
ERROR: Critical errors only
OFF: Disable logging completely

2.4 KiB Raw Blame History