AWS AppSync

AWS AppSync (appsync) on fakecloud: the complete 74-operation surface (100% conformance) — GraphQL APIs, data sources, resolvers, functions, schema types, API caches, domain names, Event APIs, and source-API associations. restJson1.

fakecloud implements AWS AppSync as a restJson1 service. All 74 operations ship with 100% conformance against AWS's own Smithy model, backed by account-partitioned state that persists across restarts in persistent mode. AppSync signs SigV4 with the appsync scope; every operation is a RESTful <METHOD> /v1|/v2/... route with path labels (e.g. POST /v1/apis, GET /v1/apis/{apiId}, POST /v1/apis/{apiId}/types/{typeName}/resolvers), routed by its HTTP method and @http URI template.

This is a faithful control plane + schema state, not a set of stubs: every resource is created, stored, and read back with the exact wire shapes AWS returns, and every sub-resource operation validates its parent API.

GraphQL APIs

  • CreateGraphqlApi mints an apiId, an arn:aws:appsync:<region>:<account>:apis/<apiId> ARN, the GRAPHQL + REALTIME endpoint uris (and matching dns), and echoes the auth configuration for any of the five authentication types — API_KEY, AWS_IAM, AMAZON_COGNITO_USER_POOLS, OPENID_CONNECT, AWS_LAMBDA — including additionalAuthenticationProviders, userPoolConfig, openIDConnectConfig, lambdaAuthorizerConfig, logConfig, and the introspection / query-depth / resolver-count limits.
  • GetGraphqlApi / ListGraphqlApis / UpdateGraphqlApi / DeleteGraphqlApi round-trip the stored API. Deleting an API cascades to its keys, data sources, resolvers, functions, types, cache, schema, and environment variables.

Sub-resources

  • API keys (CreateApiKey / ListApiKeys / UpdateApiKey / DeleteApiKey) with expires / deletes epoch bounds.
  • Data sources (CreateDataSource / Get / List / Update / Delete) for every DataSourceType — AWS_LAMBDA, AMAZON_DYNAMODB, AMAZON_OPENSEARCH_SERVICE, AMAZON_ELASTICSEARCH, HTTP, RELATIONAL_DATABASE, AMAZON_EVENTBRIDGE, AMAZON_BEDROCK_RUNTIME, NONE — with the type-specific config (dynamodbConfig, lambdaConfig, httpConfig, ...) echoed verbatim.
  • Resolvers (CreateResolver / Get / List / ListResolversByFunction / Update / Delete) attached to a type.field, UNIT or PIPELINE kind, with VTL request/response mapping templates or APPSYNC_JS code + runtime.
  • Functions (CreateFunction / Get / List / Update / Delete).
  • Schema types (CreateType / Get / List / Update / Delete), SDL or JSON format.
  • API caches (CreateApiCache / Get / Update / Delete / FlushApiCache) and environment variables (PutGraphqlApiEnvironmentVariables / GetGraphqlApiEnvironmentVariables).
  • Custom domain names (CreateDomainName / Get / List / Update / Delete) and API associations (AssociateApi / GetApiAssociation / DisassociateApi).
  • Event APIs (CreateApi / GetApi / ListApis / UpdateApi / DeleteApi) and channel namespaces (CreateChannelNamespace and its Get/List/Update/Delete).
  • Source-API associations (AssociateSourceGraphqlApi / AssociateMergedGraphqlApi / GetSourceApiAssociation / ListSourceApiAssociations / UpdateSourceApiAssociation / DisassociateSourceGraphqlApi / DisassociateMergedGraphqlApi / StartSchemaMerge / ListTypesByAssociation).

Schema lifecycle

StartSchemaCreation ingests a base64-encoded SDL schema document and stores it. GetSchemaCreationStatus settles the async lifecycle from PROCESSING to SUCCESS on read (the same state-machine-on-read pattern other async fakecloud operations use). GetIntrospectionSchema returns a representation derived from the stored SDL: the SDL verbatim for format=SDL, or a deterministic JSON projection listing the declared type names for format=JSON.

Validation and errors

Model-derived validation rejects contract violations with the error codes each operation declares — BadRequestException universally (missing required members, out-of-range integers, invalid enums, over-length strings), plus NotFoundException (a sub-resource op whose parent API — or the addressed resource — does not exist), ConcurrentModificationException, and ConflictException (a duplicate channel namespace). State is account-partitioned and persists across restarts in persistent mode.

Honest limitations

  • GraphQL query execution is not implemented. fakecloud models the AppSync control plane and schema state; it does not resolve GraphQL queries or subscriptions against the configured data sources. This is the faithful control plane such an execution engine would build on, and it is a candidate for a later batch.
  • EvaluateCode / EvaluateMappingTemplate run a documented evaluation subset. They validate that the code/template and the context are well-formed and deterministically render the context's arguments (the overwhelmingly common $util.toJson($ctx.arguments) shape), returning the model's error output member on a malformed context. They do not run a full VTL or APPSYNC_JS interpreter.
  • Data-source introspection (StartDataSourceIntrospection / GetDataSourceIntrospection) returns job metadata; it does not introspect a live relational database.