Backend Services Overview

This section provides detailed documentation for each of Inwire's backend services. Each service handles a specific domain within the ML lifecycle and exposes REST APIs that the frontend and other services consume.

---

Service Architecture

Inwire follows a microservices architecture where each service is:

• Independently deployable — Services can be updated without affecting others
• Domain-focused — Each service owns a specific capability
• API-first — All functionality exposed via documented REST APIs
• Scalable — Services can scale horizontally based on load

Service Communication

┌─────────────────────────────────────────────────────────────────────────────┐
│                              Frontend (Next.js)                              │
└─────────────────────────────────────────────────────────────────────────────┘
                                      │
                                      ▼
┌─────────────────────────────────────────────────────────────────────────────┐
│                              API Gateway                                     │
│                     (Authentication, Routing, Rate Limiting)                 │
└─────────────────────────────────────────────────────────────────────────────┘
         │              │              │              │              │
         ▼              ▼              ▼              ▼              ▼
    ┌─────────┐   ┌─────────┐   ┌─────────┐   ┌─────────┐   ┌─────────┐
    │  Core   │   │ Synthex │   │ Model   │   │ModelOps │   │   RAG   │
    │         │   │         │   │Training │   │         │   │         │
    └─────────┘   └─────────┘   └─────────┘   └─────────┘   └─────────┘
         │              │              │              │              │
         └──────────────┴──────────────┴──────────────┴──────────────┘
                                      │
                                      ▼
                        ┌─────────────────────────────┐
                        │     Shared Infrastructure    │
                        │  PostgreSQL │ Redis │ S3    │
                        └─────────────────────────────┘

---

Service Catalog

Core Services

Service	Purpose	Port	Documentation
Core	Authentication, users, teams, organizations	18000	Core Service
Synthex	Synthetic data generation & data management	18090	Synthex User Guide
Model Training	Experiment tracking & training orchestration	18052	Model Training
ModelOps	Model deployment & lifecycle management	18087	ModelOps User Guide

AI/ML Services

Service	Purpose	Port	Documentation
RAG	Retrieval-Augmented Generation pipelines	8186	RAG Service
RAG Index Manager	Knowledge base optimization	8187	RAG Index
RAG Sentinel	AI security & compliance	8188	RAG Sentinel
PromptScope	Prompt engineering & testing	8082	PromptScope
Agent Studio	AI agent development	8089	Agent Studio

Data & Operations Services

Service	Purpose	Port	Documentation
Stream	Real-time data processing	8083	Stream Service
FlowForge	Workflow automation	8084	FlowForge
Inwire Observe	Platform monitoring & observability	8085	Observe
Vault Agent	Secrets management	8086	Vault Agent

---

Core Service

The Core Service is the foundation of the Inwire platform, handling:

• Authentication — JWT-based auth, SSO integration, API keys
• Authorization — Role-based access control (RBAC)
• User Management — User profiles, preferences, settings
• Team Management — Team creation, membership, permissions
• Organization Management — Multi-tenant organization support
• Integrations — External service connections (cloud, VCS, registries)

Key Endpoints

Endpoint	Method	Description
/api/v1/auth/login	POST	User authentication
/api/v1/auth/token/refresh	POST	Refresh JWT token
/api/v1/users	GET/POST	User management
/api/v1/teams	GET/POST	Team management
/api/v1/organizations	GET/POST	Organization management
/api/v1/integrations	GET/POST	Integration configuration

Authentication Flow

┌──────────┐         ┌──────────┐         ┌──────────┐
│  Client  │         │   Core   │         │ Database │
└────┬─────┘         └────┬─────┘         └────┬─────┘
     │                    │                    │
     │  POST /auth/login  │                    │
     │───────────────────>│                    │
     │                    │  Verify credentials│
     │                    │───────────────────>│
     │                    │<───────────────────│
     │   JWT + Refresh    │                    │
     │<───────────────────│                    │
     │                    │                    │
     │  API Request + JWT │                    │
     │───────────────────>│                    │
     │                    │  Validate token    │
     │   Response         │                    │
     │<───────────────────│                    │

---

Synthex Service

Synthex is the data management and synthetic data generation service — the "data twin" of the platform. It handles everything related to data preparation, profiling, transformation, and synthetic data generation.

Full documentation: Synthex User Guide

Capabilities

• Dataset import and management
• Data profiling and schema detection
• Data cleaning and transformation recipes
• Synthetic data generation (multiple methods)
• Data quality evaluation
• Privacy-preserving data handling

Key Endpoints

Endpoint	Method	Description
/api/v1/datasets	GET/POST	Dataset management
/api/v1/profiles	GET/POST	Data profile management
/api/v1/recipes	GET/POST	Transformation recipes
/api/v1/generators	GET/POST	Generator configurations
/api/v1/jobs	GET/POST	Generation job management

---

Model Training Service

The Model Training Service orchestrates ML experiments and training jobs:

• Experiment Tracking — Organize and track ML experiments
• Run Management — Execute and monitor training runs
• Metric Logging — Track metrics, parameters, and artifacts
• Dataset Integration — Connect with Synthex for data
• Hyperparameter Tuning — Automated parameter optimization

Key Endpoints

Endpoint	Method	Description
/api/v1/experiments	GET/POST	Experiment management
/api/v1/runs	GET/POST	Training run management
/api/v1/metrics	GET/POST	Metric logging
/api/v1/artifacts	GET/POST	Artifact storage

Training Workflow

┌─────────────┐     ┌─────────────┐     ┌─────────────┐
│  Synthex    │────>│   Model     │────>│  ModelOps   │
│  (Dataset)  │     │  Training   │     │  (Deploy)   │
└─────────────┘     └─────────────┘     └─────────────┘
                          │
                          ▼
                    ┌───────────┐
                    │ Experiment│
                    │  Tracking │
                    │  - Runs   │
                    │  - Metrics│
                    │  - Models │
                    └───────────┘

---

ModelOps Service

The ModelOps Service manages the model lifecycle from registration to production:

Full documentation: ModelOps User Guide

• Model Registry — Version and catalog trained models
• Deployment Management — Deploy models to various targets
• Scaling — Auto-scaling based on traffic
• Monitoring — Track model performance in production
• Governance — Approval workflows and audit trails

Key Endpoints

Endpoint	Method	Description
/api/v1/models	GET/POST	Model registry
/api/v1/deployments	GET/POST	Deployment management
/api/v1/builds	GET/POST	Container builds
/api/v1/clusters	GET	Available clusters

Deployment Strategies

Strategy	Description	Use Case
Rolling	Gradual pod replacement	Low-risk updates
Blue/Green	Instant traffic switch	Zero-downtime
Canary	Partial traffic routing	Production testing
Shadow	Mirror traffic without affecting users	Validation

---

RAG Service

The RAG (Retrieval-Augmented Generation) Service builds knowledge-grounded AI pipelines:

• Document Ingestion — Import and process documents
• Chunking — Split documents into searchable segments
• Embedding — Generate vector embeddings
• Retrieval — Similarity and keyword search
• Generation — LLM-powered response generation

Key Endpoints

Endpoint	Method	Description
/api/v1/pipelines	GET/POST	Pipeline management
/api/v1/documents	GET/POST	Document ingestion
/api/v1/query	POST	Query pipeline
/api/v1/embeddings	POST	Generate embeddings

---

RAG Index Manager

Optimizes and manages RAG knowledge bases:

• Index Optimization — Tune vector indexes for performance
• Reindexing — Rebuild indexes when needed
• Analytics — Query performance metrics
• Maintenance — Cleanup and consolidation

---

RAG Sentinel

AI security and compliance for RAG pipelines:

• Content Filtering — Detect and block harmful content
• PII Detection — Identify personal information
• Compliance Checking — Ensure regulatory compliance
• Audit Logging — Track all queries and responses

---

PromptScope Service

Prompt engineering and testing platform:

• Prompt Editor — Create and edit prompts
• Playground — Test prompts with different models
• Version Control — Track prompt versions
• A/B Testing — Compare prompt performance
• Analytics — Measure prompt effectiveness

---

Agent Studio

AI agent development platform:

• Agent Builder — Visual agent construction
• Tool Integration — Connect agents to tools
• Testing — Simulate agent interactions
• Deployment — Deploy agents to production

---

Stream Service

Real-time data processing:

• Ingestion — Receive data from various sources
• Processing — Transform and enrich in real-time
• Routing — Send to destinations
• Monitoring — Track throughput and latency

---

FlowForge Service

Workflow automation:

• Workflow Builder — Visual workflow design
• Triggers — Event and schedule-based triggers
• Actions — Connect to services and APIs
• Monitoring — Track workflow executions

---

Inwire Observe

Platform monitoring and observability:

• Metrics — System and application metrics
• Logs — Centralized log aggregation
• Traces — Distributed tracing
• Alerts — Configurable alerting

---

Vault Agent

Secrets management:

• Secret Storage — Secure secret storage
• Dynamic Secrets — Generate secrets on demand
• Rotation — Automatic secret rotation
• Audit — Secret access logging

---

Common Patterns

API Response Format

All services use consistent response formats:

Success Response:

{
  "data": { ... },
  "meta": {
    "page": 1,
    "per_page": 20,
    "total": 100
  }
}

Error Response:

{
  "error": {
    "code": "VALIDATION_ERROR",
    "message": "Invalid input",
    "details": [ ... ]
  }
}

Pagination

List endpoints support pagination:

GET /api/v1/datasets?page=1&per_page=20&sort=created_at&order=desc

Filtering

Most list endpoints support filtering:

GET /api/v1/jobs?status=running&type=generation

---

Health Checks

All services expose health endpoints:

Endpoint	Description
/health	Basic liveness check
/health/ready	Readiness check (includes dependencies)
/health/live	Kubernetes liveness probe

---

API Documentation

Each service provides OpenAPI documentation:

• Swagger UI: http://service-url/docs
• ReDoc: http://service-url/redoc
• OpenAPI JSON: http://service-url/openapi.json

---

Next Steps

For detailed service documentation:

• Synthex User Guide — Comprehensive synthetic data generation guide
• Return to User Guide for platform overview