Custom Model Resolver

How to configure AI model resolution for your generators.

How Model Resolution Works

Generator blocks specify a model string (e.g., "gpt-5-mini"). At runtime, the server resolves this string to an actual AI SDK model instance via a model resolver.

const chatGen = generator({
  name: "chat",
  model: "gpt-5-mini",  // Resolved at runtime
  // ...
});

Default Resolver

By default, the server uses the Vercel AI Gateway:

const router = createFlowApiRouter({ registry });
// Uses AI_GATEWAY_API_KEY or Vercel OIDC for model resolution

Custom Resolver with AI SDK

Use createAiSdkModelResolver to provide your own model mapping:

import { createFlowApiRouter, createAiSdkModelResolver } from "@flow-state-dev/server";
import { openai } from "@ai-sdk/openai";
import { anthropic } from "@ai-sdk/anthropic";

const router = createFlowApiRouter({
  registry,
  modelResolver: createAiSdkModelResolver((modelId) => {
    switch (modelId) {
      case "gpt-5-mini":
        return openai("gpt-4o-mini");
      case "claude-sonnet":
        return anthropic("claude-sonnet-4-20250514");
      default:
        return openai(modelId);
    }
  }),
});

Multiple Providers

Route models to different providers based on the model ID pattern:

const resolver = createAiSdkModelResolver((modelId) => {
  if (modelId.startsWith("claude-")) {
    return anthropic(modelId);
  }
  if (modelId.startsWith("gpt-")) {
    return openai(modelId);
  }
  // Default provider
  return openai(modelId);
});

Environment-Based Configuration

const resolver = createAiSdkModelResolver((modelId) => {
  const provider = process.env.AI_PROVIDER ?? "openai";

  switch (provider) {
    case "anthropic":
      return anthropic(modelId);
    case "openai":
    default:
      return openai(modelId);
  }
});

Testing

In tests, use createMockModelResolver to avoid real API calls:

import { createMockModelResolver } from "@flow-state-dev/testing";

const mockResolver = createMockModelResolver({
  models: {
    "gpt-5-mini": { output: "Mocked response" },
  },
});

Or use generator mocks in test harnesses (preferred):

const result = await testFlow({
  flow: myFlow,
  action: "chat",
  input: { message: "Hello" },
  generators: {
    "chat": { output: "Mocked!" },
  },
});