Ananke.Orchestration.Google

Google Gemini provider for Ananke.Orchestration — IStreamingAgentModel implementation with tool calling, structured output, and token-level streaming via the official Google GenAI SDK. Supports both the Gemini Developer API and Vertex AI.

Install

dotnet add package Ananke.Orchestration.Google

Quick start

using Ananke.Orchestration.Google; // Gemini Developer API (API key) IStreamingAgentModel model = GeminiAgentModel.Create(apiKey, "gemini-2.5-flash"); // Vertex AI (project + location, uses Application Default Credentials) IStreamingAgentModel model = GeminiAgentModel.CreateVertexAI(project, location, "gemini-2.5-flash"); // Or from an existing Google.GenAI.Client var client = new Google.GenAI.Client(apiKey: apiKey); IStreamingAgentModel model = new GeminiAgentModel(client, "gemini-2.5-flash");

Use in a workflow agent job

var agentJob = AgentJobFactory .Create<MyState, MyResponse>("analyze", model) .WithSystemPrompt("You are a research analyst.") .WithTools(searchTools) .WithPrompt(state => $"Analyze: {state.Query}") .MapResult((state, response) => state with { Analysis = response.Text }) .Build();

YAML manifest (Ananke.Design)

models: gemini: provider: gemini model: gemini-2.5-flash var models = new ModelResolver() .Register("gemini", "Gemini", GeminiAgentModel.Create) .Resolve(manifest, key => config[key]);

Features

• Full IStreamingAgentModel implementation (streaming + non-streaming)
• Tool calling with automatic function dispatch
• Structured JSON output via ResponseSchema
• Automatic JSON Schema to Google Schema conversion
• Gemini Developer API (API key) and Vertex AI (ADC) support

Requirements

• Ananke.Orchestration (transitive)
• Google.GenAI SDK ≥ 1.1.0

Documentation

Full docs, demos, and architecture: github.com/sevensamurai/Ananke

License

Apache 2.0