Skip to Content
DocsAckAdvancedEcosystem Integrations

Ecosystem Integrations

Reuse an AckSchema (or a @Schemable() model’s Model.schema) across Ack’s official adapter packages. Each adapter has its own supported schema shapes; MCP tool inputs, for example, must render as a non-null object root.

Adapter PackageTarget EcosystemPrimary Entry Point
ack_mcp_dartModel Context Protocol (mcp_dart)server.registerAckTool(...), schema.toMcpToolInputSchema()
ack_firebase_aiFirebase AI Logic (firebase_ai / Gemini)schema.toFirebaseAiResponseJsonSchema()
ack_json_schema_builderjson_schema_builderschema.toJsonSchemaBuilder(), jsbSchema.toAckSchema()

1. Define a Model or Schema Once

import 'package:ack/ack.dart'; part 'weather_models.g.dart'; @Schemable(description: 'Parameters for querying current weather.') final class const WeatherQuery({ /// @description City and country, e.g. "Tokyo, JP". @HasMinLength(2) required final String city, final String units = 'celsius', }) with _$WeatherQueryAck { static final schema = WeatherQuerySchema.schema; static final fromJson = WeatherQuerySchema.fromJson; } @Schemable(description: 'Current weather report.') final class const WeatherReport({ required final String city, required final double temperatureC, }) with _$WeatherReportAck { static final schema = WeatherReportSchema.schema; static final fromJson = WeatherReportSchema.fromJson; }

2. MCP Tool Servers (ack_mcp_dart)

ack_mcp_dart bridges AckSchema with McpServer from package:mcp_dart/mcp_dart.dart. It exports the tool’s input JSON Schema, validates and decodes incoming arguments before your callback runs, and formats Ack validation failures as MCP CallToolResult(isError: true) responses. MCP’s structural validation runs first; its failures do not reach the Ack callback.

Install

dart pub add ack:^2.0.0-beta.3 ack_mcp_dart:^2.0.0-beta.3 mcp_dart

Register a typed input tool

Pass WeatherQuery.schema through input. The callback receives a WeatherQuery and returns a CallToolResult; output validation and serialization belong to the callback:

import 'dart:convert'; import 'package:ack_mcp_dart/ack_mcp_dart.dart'; import 'package:mcp_dart/mcp_dart.dart'; import 'weather_models.dart'; void registerWeatherTools(McpServer server) { server.registerAckTool<WeatherQuery>( 'get_weather', description: 'Fetch the current weather for a city.', input: WeatherQuery.schema, callback: (query, extra) { final report = WeatherReport( city: query.city, temperatureC: 21.5, ); // Generated toJson validates and encodes the report. final json = report.toJson(); return CallToolResult( content: [TextContent(text: jsonEncode(json))], structuredContent: json, ); }, ); }

For manual registration, use schema.toMcpToolInputSchema(), parseMcpToolArguments(schema, args), and ackValidationErrorToCallToolResult(error, toolName: name). On the client, validate structured output explicitly with WeatherReport.schema.safeParse(result.structuredContent). The adapter does not register or advertise output schemas; use mcp_dart directly when those are required.


3. Firebase AI / Gemini Structured Outputs (ack_firebase_ai)

ack_firebase_ai renders an Ack schema as the JSON-compatible map accepted by GenerationConfig.responseJsonSchema. Use a model that supports that field, initialize Firebase in your app, and validate the response with the same Ack schema. This is not a converter to Firebase’s native Schema type or a function declaration helper.

Install

flutter pub add ack:^2.0.0-beta.3 ack_firebase_ai:^2.0.0-beta.3 firebase_ai firebase_core

Generate and validate structured output

import 'dart:convert'; import 'package:ack_firebase_ai/ack_firebase_ai.dart'; import 'package:firebase_ai/firebase_ai.dart'; import 'weather_models.dart'; Future<WeatherReport> askGeminiWeather( String prompt, { required String modelName, }) async { final model = FirebaseAI.googleAI().generativeModel( model: modelName, generationConfig: GenerationConfig( responseMimeType: 'application/json', responseJsonSchema: WeatherReport.schema.toFirebaseAiResponseJsonSchema(), ), ); final response = await model.generateContent([Content.text(prompt)]); final decoded = jsonDecode(response.text ?? 'null'); // Validate and decode Gemini's response into WeatherReport return WeatherReportSchema.parse(decoded); }

The adapter returns schema.toJsonSchema() unchanged. It does not append constraint descriptions or adapt keywords to a provider-specific subset. Response schemas guide generation; Ack validation remains authoritative for constraints and refinements that the provider does not enforce.


4. json_schema_builder Bridge (ack_json_schema_builder)

ack_json_schema_builder provides bidirectional conversion between AckSchema and package:json_schema_builder:

Install

dart pub add ack:^2.0.0-beta.3 ack_json_schema_builder:^2.0.0-beta.3 json_schema_builder

Convert in both directions

import 'package:ack/ack.dart'; import 'package:ack_json_schema_builder/ack_json_schema_builder.dart'; import 'package:json_schema_builder/json_schema_builder.dart' as jsb; import 'weather_models.dart'; void main() { // Native Ack schemas render as Draft-7; imported 2020-12 resources are preserved. final jsb.Schema builderSchema = WeatherQuery.schema.toJsonSchemaBuilder(); // Need legacy Draft-7 definitions lowering instead? final jsb.Schema draft7Schema = WeatherQuery.schema.toJsonSchemaBuilderDraft7(); // 2. json_schema_builder Schema -> executable AckSchema<Object, Object> final AckSchema<Object, Object> importedValidator = jsb.Schema.object( properties: {'city': jsb.Schema.string(minLength: 2)}, required: ['city'], ).toAckSchema(assertFormats: true); print(builderSchema.value); print(draft7Schema.value); final valid = importedValidator.parse({'city': 'Lisbon'}); print(valid); }

Building Your Own Adapter Package

Want to convert Ack schemas to another target format (OpenAPI, Protobuf, GraphQL, or another LLM SDK)? Traverse schema.toSchemaModel() (AckSchemaModel). See:

Last updated on