nvlabs/skill2env/apollo-connectors

Apollo Connectors Schema Assistant

MCP Tools

If GraphOS MCP Tools are available, use them:

• connectors-spec: Fetch the complete Connectors specification before starting any connector work
• apollo_docs_search: Search for relevant documentation
• apollo_docs_read: Read specific documentation pages by slug

Documentation paths by topic:

• Requests: /graphos/connectors/requests/url, /headers, /body, /batching
• Responses: /graphos/connectors/responses/fields, /error-handling
• Mapping: /graphos/connectors/mapping, /arrays, /enums, /literals
• Entities: /graphos/connectors/entities, /patterns

Process

Follow this 5-step process. DO NOT skip any steps.

Step 1: Research

• [ ] Understand the API being called and the structure of responses
• [ ] Ask the user for example API responses if not provided
• [ ] Fetch relevant documentation from MCP tools or reference files
• [ ] DO NOT write any code until research is complete

Step 2: Implement

• [ ] Create the schema using the template below
• [ ] Follow the grammar, methods, and variables in the reference files
• [ ] Ask clarifying questions if unsure about requirements

Step 3: Validate (Compose)

• [ ] Run rover supergraph compose --config ./supergraph.yaml
• [ ] Fix any composition errors before proceeding

Step 4: Execute

• [ ] Run rover connector run --schema <file> -c "<Type.field>" -v "{}"
• [ ] Verify the connector executes correctly

Step 5: Test

• [ ] Create or update test files under /tests/
• [ ] Run rover connector test
• [ ] Ensure full test coverage for each connector

Schema Template

# Note to AI Friends: This is an Apollo Connectors schema. You should use your connectors-spec MCP tool if you have it available before editing this file.
 
extend schema
  @link(url: "https://specs.apollo.dev/federation/v2.12")
  @link(url: "https://specs.apollo.dev/connect/v0.3", import: ["@source", "@connect"])
 
@source(name: "api_name", http: { baseURL: "https://api.example.com" })
 
type Query {
  example(id: ID!): Example
    @connect(
      source: "api_name"
      http: { GET: "/example/{$args.id}" }
      selection: """
      id
      name
      """
    )
}
 
type Example {
  id: ID!
  name: String
}

Version Requirements: Always use federation/v2.12 and connect/v0.3 unless specified otherwise.

Reference Files

Before implementing connectors, read the relevant reference files:

• Grammar - Selection mapping EBNF syntax
• Methods - Available transformation methods
• Variables - Available mapping variables
• Entities - Entity patterns and batching
• Validation - Rover commands for validation
• Troubleshooting - Common errors and solutions

Key Rules

Selection Mapping

• Prefer sub-selections over ->map for cleaner mappings
• Do NOT use $ when selecting fields directly from root
• Field aliasing: newName: originalField (only when renaming)
• Sub-selection: fieldName { ... } (to map nested content)

# DO - Direct sub-selection for arrays
$.results {
  firstName: name.first
  lastName: name.last
}
 
# DO NOT - Unnecessary root $
$ {
  id
  name
}
 
# DO - Direct field selection
id
name

Entities

• Add @connect on a type to make it an entity (no @key needed)
• Create entity stubs in parent selections: user: { id: userId }
• When you see an ID field (e.g., productId), create an entity relationship
• Each entity should have ONE authoritative subgraph with @connect

Literal Values

Use $() wrapper for literal values in mappings:

$(1)              # number
$(true)           # boolean
$("hello")        # string
$({"a": "b"})     # object
 
# In body
body: "$({ a: $args.a })"  # CORRECT
body: "{ a: $args.a }"     # WRONG - will not compose

Headers

http: {
  GET: "/api"
  headers: [
    { name: "Authorization", value: "Bearer {$env.API_KEY}" },
    { name: "X-Forwarded", from: "x-client" }
  ]
}

Batching

Convert N+1 patterns using $batch:

type Product @connect(
  source: "api"
  http: {
    POST: "/batch"
    body: "ids: $batch.id"
  }
  selection: "id name"
) {
  id: ID!
  name: String
}

Ground Rules

• NEVER make up syntax or directive values not in this specification
• NEVER use --elv2-license accept (for humans only)
• ALWAYS ask for example API responses before writing code
• ALWAYS validate with rover supergraph compose after changes
• ALWAYS create entity relationships when you see ID fields
• Prefer $env over $config for environment variables
• Use rover dev for running Apollo Router locally