API Cloud Architecture

Technical proposal for API gateway infrastructure

Overview

This document outlines the architecture for an API cloud platform using Kong as the gateway. The platform enables organizations to expose APIs with authentication, rate limiting, and intelligent routing to backend ML pipelines.

Figure 1: API Cloud Architecture Blueprint

Flow Configuration

The following attributes define a customer API rollout:

Organization Attributes

{
    "organization": "ITU",
    "zone": "swatantra",
    "theme": "sales"
}

Credential Map

{
    "apiuser": "vinay",
    "apikey": "ENTER_KEY_HERE"
}

Pipeline Map

Routes requests to appropriate ML pipelines:

{
    "pmap": [
        {
            "pipeuse": "csim",
            "pipeip": "192.168.5.98",
            "pipeakey": "abcdefghijklmnopqrstu",
            "pipeport": "17092"
        },
        {
            "pipeuse": "ureco",
            "pipeip": "192.168.5.99",
            "pipeakey": "klfmnopqrstabcdefghij",
            "pipeport": "17093"
        }
    ]
}

Policies

{
    "policies": {
        "ratelimit": 200,
        "quota": 500000
    }
}

Complete Configuration Example

{
    "organization": "ITU",
    "zone": "swatantra",
    "subject": "sales",
    "cmap": {
        "apiuser": "vinay",
        "apikey": "ENTER_KEY_HERE"
    },
    "pmap": [
        {
            "pipeuse": "csim",
            "pipeip": "192.168.5.98",
            "pipeakey": "abcdefghijklmnopqrstu",
            "pipeport": "17092"
        }
    ],
    "dlink": "api.organisation.one/docs/sales/ingest",
    "policies": {
        "ratelimit": 200,
        "quota": 500000
    },
    "service": {
        "servicename": "itusales",
        "serviceurl": "h.one/cerebro/itu/swatantra/sales",
        "upstreamurl": "192.168.5.100:8080/cerebro/sales",
        "routehost": "itu.organisation.one",
        "routepaths": "h.one/cerebro/itu/swatantra/sales"
    }
}

API Routing Patterns

Pattern	Example	Use Case
Common Public API	api.h.one/ingest	Publish friendly
Granular Path	api.h.one/itu/swantantra/sales/ingest	Developer friendly
Subdomain Based	itu.h.one/sales/ingest	Documentation friendly
Organization Based	h.one/itu/ingest	Deployment friendly

Setup Procedure

Step 1: Create Service and Route

# Add service
curl -i -X POST \
  --url http://localhost:8001/services/ \
  --data 'name=itu-service' \
  --data 'url=http://mockbin.org/request'

# Add route
curl -i -X POST \
  --url http://localhost:8001/services/itu-service/routes \
  --data 'paths[]=/ingest'

# Verify configuration
curl -i -X GET --url http://localhost:8000/ingest

Step 2: Create Consumer

# Enable key-auth plugin
curl -i -X POST \
  --url http://localhost:8001/services/itu-service/plugins/ \
  --data 'name=key-auth'

# Create consumer
curl -i -X POST \
  --url http://localhost:8001/consumers/ \
  --data "username=motoko"

# Provision credentials
curl -i -X POST \
  --url http://localhost:8001/consumers/motoko/key-auth/ \
  --data 'key=ENTER_KEY_HERE'

# Verify
curl -i -X GET \
  --url http://localhost:8000/ingest \
  --header "apikey: ENTER_KEY_HERE"

Step 3: Enable Request Transformer

Append pipeline information to requests:

curl -X POST http://localhost:8001/plugins \
  --data "name=request-transformer" \
  --data "consumer.id={consumer_id}" \
  --data "config.append.body=pipeuse:csim, pipeip:192.168.5.98, pipeakey:abcdefgh, pipeport:17092"

Step 4: Apply Policies

# Add rate limiting
curl -X POST http://localhost:8001/plugins \
  --data "name=rate-limiting" \
  --data "consumer.id={consumer_id}" \
  --data "config.minute=10"

Step 5: Declarative Configuration

Configuration can be applied through:

• REST API calls
• JSON with hot reload
• YAML with hot reload

API Matrix

Interface Details

API	Name	Description
a_csim	Content Similarity	Discover similar documents within vector space
a_preureco	Pre-determined User Recommendations	Cold-start solution for user recommendations
a_postureco	Post-determined User Recommendations	Collaborative filtering for user discovery
a_wnext	What Next?	Recommended next steps for efficient outcomes
a_creco	Context Recommendations	Discover relevant contexts for users

Physical Pipelines

Pipeline	APIs Served	Component	Purpose
p_csim	a_csim, a_preureco	content-similarity	Word2Vec text similarity
p_ureco	a_preureco, a_postureco, a_wnext, a_creco	recommend-users-for-contexts	User recommendations

Installation

Standalone Setup

# Install Kong
sudo apt-get update
sudo apt-get install -y apt-transport-https curl lsb-core
echo "deb https://kong.bintray.com/kong-deb `lsb_release -sc` main" | sudo tee -a /etc/apt/sources.list
curl -o bintray.key https://bintray.com/user/downloadSubjectPublicKey?username=bintray
sudo apt-key add bintray.key
sudo apt-get update
sudo apt-get install -y kong

Cassandra Configuration

-- Create keyspace
CREATE KEYSPACE kong WITH REPLICATION = {
    'class': 'SimpleStrategy',
    'replication_factor': 1
};

-- Create user and role
CREATE USER kong WITH PASSWORD = 'kong' AND LOGIN = true;
CREATE ROLE kong_admin;
GRANT ALL PERMISSIONS ON KEYSPACE kong TO kong_admin;
GRANT CREATE ON ALL KEYSPACES TO kong_admin;
GRANT kong_admin TO kong;

Kong Configuration

Edit /etc/kong/kong.conf:

database = cassandra
cassandra_contact_points = xxx.xxx.xxx.xxx
cassandra_keyspace = kong
cassandra_username = kong
cassandra_password = xxxxxxxxxx
db_update_propagation = 60

Start Kong

kong migrations bootstrap [-c /path/to/kong.conf]
kong start [-c /path/to/kong.conf]
curl -i http://localhost:8001/

References

• Kong Quick Start
• Kong Ubuntu Installation

---

This proposal outlines the foundation for a scalable API gateway. Cluster deployment extensions are available for production environments.