API Ingestion Connectors

Ingest data from external APIs into graph-ready formats for knowledge graph construction.

This skill retrieves data from diverse API sources and prepares it for transformation into graph-ready structures such as nodes, relationships, and triples.

Quick Start

Use When

• Ingesting data from REST APIs
• Querying GraphQL endpoints
• Integrating external services into data pipelines
• Pulling data from SaaS platforms
• Transforming API responses into graph datasets
• Building real-time knowledge graph updates

Inputs

• API endpoint URLs
• Authentication credentials
• Request parameters and headers
• Pagination configuration
• Response format specifications
• Transformation mappings

Outputs

• JSON/CSV datasets
• Graph-ready node/edge structures
• RDF triples
• Connector configurations
• ETL pipeline definitions

Example

Input API Configuration:

Endpoint: https://api.example.com/users
Method: GET
Auth: Bearer Token
Pagination: page-based, 30 items per page

Generated Output:

{
  "nodes": [
    {"id": "user_1", "type": "Person", "name": "Alice", "email": "alice@example.com"},
    {"id": "org_1", "type": "Organization", "name": "Acme Corp"}
  ],
  "edges": [
    {"source": "user_1", "target": "org_1", "relation": "WORKS_AT"}
  ]
}

Supported API Types

1. REST APIs

Connect to standard HTTP REST endpoints with flexible authentication and pagination

type: rest
endpoint: https://api.example.com/resource
method: GET|POST|PUT|DELETE
response_format: json|xml|csv

2. GraphQL APIs

Query GraphQL endpoints with structured query definitions

query {
  users {
    id
    name
    email
    organization {
      name
    }
  }
}

3. OAuth-Protected APIs

Authenticate using OAuth 2.0 flows (authorization code, client credentials)

auth_type: oauth2
client_id: ${CLIENT_ID}
client_secret: ${CLIENT_SECRET}
token_endpoint: https://api.example.com/oauth/token

4. API Key Authentication

Simple API key-based authentication

auth_type: api_key
key_param: X-API-Key
key_value: ${API_KEY}

5. Bearer Token Authentication

OAuth 2.0 bearer token authentication

auth_type: bearer
token: ${ACCESS_TOKEN}

Pagination Strategies

Offset/Limit Pagination

type: offset
param_offset: offset
param_limit: limit
start_at: 0
page_size: 20

Page-Based Pagination

type: page
param_page: page
page_size: 30
start_at: 1

Cursor-Based Pagination

type: cursor
cursor_param: after
next_cursor_field: pageInfo.endCursor
has_next_field: pageInfo.hasNextPage

Execution Steps

1. Validate Configuration – Check endpoint, auth, and parameters
2. Authenticate – Obtain credentials and tokens
3. Make Request – Execute HTTP/GraphQL request
4. Handle Pagination – Fetch all pages/results
5. Parse Response – Extract and validate response data
6. Transform Data – Convert to graph-ready format
7. Generate Output – Create nodes, edges, or triples
8. Feed to Pipeline – Pass to downstream transformation skills

Output Formats

Node-Edge Structure

{
  "nodes": [{"id": "...", "type": "...", "properties": {...}}],
  "edges": [{"source": "...", "target": "...", "type": "...", "properties": {...}}]
}

Graph Triples (RDF)

:entity1 :relationType :entity2 .
:entity1 :property "value" .

CSV Export

node_id,node_type,node_name,property1,property2

Error Handling

The connector should handle:

• Network Errors – Retry logic with exponential backoff
• Authentication Errors – Token refresh, credential validation
• Rate Limiting – Backoff and request throttling
• Malformed Responses – Schema validation and error reporting
• Timeouts – Connection and read timeout handling

Example retry strategy:

retry:
  max_attempts: 3
  backoff_factor: 2
  initial_delay: 1s
  max_delay: 60s
  retryable_status_codes: [429, 500, 502, 503, 504]

Recommended Libraries

• HTTP Clients: requests, httpx, aiohttp
• GraphQL: gql, graphene, strawberry
• OAuth: authlib, oauthlib
• Data Validation: pydantic, jsonschema
• Data Transformation: pandas, polars

Best Practices

✓ Respect API rate limits and terms of service
✓ Implement exponential backoff for retries
✓ Validate response schemas before processing
✓ Handle and log errors appropriately
✓ Cache results when possible
✓ Normalize and deduplicate entities
✓ Secure credentials (use environment variables)
✓ Monitor API changes and versioning
✓ Implement request timeout handling
✓ Document API assumptions and requirements

Integration with Downstream Skills

The ingested data feeds into:

• JSON → Triples Converter – Transform to RDF
• CSV → Graph Loader – Load into graph database
• Text → Entity/Relation Extractor – Extract structured knowledge
• ETL Pipeline Generator – Orchestrate full workflows
• Schema Validation – Validate against graph schema

References

See connector-patterns.md for detailed API connector patterns and example-connectors.md for complete connector examples.

Version: 1.0.0