chore: stop tracking generated model files

.gitignore

@@ -101,6 +101,9 @@ RELEASE-GUIDE.md
 TODOS.md
 AGENTS.md
 
+# Generated model registries
+packages/pi-ai/src/*.generated.ts
+
 # opencode local config
 .opencode/
 test-results/

packages/pi-ai/README.md

@@ -1326,7 +1326,9 @@ Create a new provider file (for example `amazon-bedrock.ts`) that exports:
 - Map chat/tool-capable provider model data to the standardized `Model` interface via `scripts/generate-models.ts`
 - Map image-generation provider model data to the standardized `ImagesModel` interface via `scripts/generate-image-models.ts`
 - Handle provider-specific quirks (pricing format, capability flags, model ID transformations)
-- Run `pnpm run generate` from `packages/pi-ai` when changing model generation; `pnpm run build` compiles the committed generated catalogs without refreshing them.
+- The generated catalogs (`src/*.generated.ts`) are ignored by git. Run `pnpm run generate` from `packages/pi-ai` when changing model generation or when you need to inspect the generated output locally.
+- `pnpm run build` and `pnpm test` regenerate the catalogs before compiling or running Vitest, so a fresh clone does not need committed generated files.
+- Generation fetches live model metadata from models.dev, OpenRouter, and Vercel AI Gateway. Run those commands with network access; for image-model generation, set `OPENROUTER_MODELS_URL` to point the OpenRouter image model fetch at a fixture or mirror when needed.
 
 #### 5. Tests (`test/`)
 

packages/pi-ai/package.json

@@ -73,7 +73,8 @@
     "generate-models": "node scripts/generate-models.ts",
     "generate-image-models": "node scripts/generate-image-models.ts",
     "generate": "pnpm run generate-models && pnpm run generate-image-models",
-    "build": "pnpm run clean && tsc -p tsconfig.json --incremental false",
+    "build": "pnpm run generate && pnpm run clean && tsc -p tsconfig.json --incremental false",
+    "pretest": "pnpm run generate",
     "test": "vitest --run",
     "prepublishOnly": "pnpm run clean && pnpm run build"
   },

packages/pi-ai/scripts/generate-image-models.ts

@@ -29,7 +29,9 @@ interface OpenRouterModelRecord {
 async function fetchOpenRouterImageModels(): Promise<ImagesModel<"openrouter-images">[]> {
 	try {
 		console.log("Fetching image models from OpenRouter API...");
-		const response = await fetch(`${OPENROUTER_BASE_URL}/models?output_modalities=image`);
+		const url =
+			process.env["OPENROUTER_MODELS_URL"] ?? `${OPENROUTER_BASE_URL}/models?output_modalities=image`;
+		const response = await fetch(url);
 		const data = (await response.json()) as { data?: OpenRouterModelRecord[] };
 		const models: ImagesModel<"openrouter-images">[] = [];
 
@@ -76,7 +78,7 @@ async function fetchOpenRouterImageModels(): Promise<ImagesModel<"openrouter-ima
 	}
 }
 
-function generateImageModelsFile(models: ImagesModel<"openrouter-images">[]): string {
+export function generateImageModelsFile(models: ImagesModel<"openrouter-images">[]): string {
 	const imageModelsByProvider = {
 		openrouter: Object.fromEntries(
 			models
@@ -119,13 +121,24 @@ ${providerEntries}
 
 async function main(): Promise<void> {
 	const models = await fetchOpenRouterImageModels();
+	if (models.length === 0) {
+		console.error(
+			"OpenRouter image-models fetch returned no models; refusing to overwrite src/image-models.generated.ts.",
+		);
+		console.error(
+			"Retry when network is available, or build with the committed catalog (skip generate-image-models).",
+		);
+		process.exit(1);
+	}
 	const output = generateImageModelsFile(models);
 	const outputPath = join(packageRoot, "src", "image-models.generated.ts");
 	writeFileSync(outputPath, output, "utf-8");
 	console.log(`Generated ${outputPath}`);
 }
 
-main().catch((error) => {
-	console.error(error);
-	process.exit(1);
-});
+if (process.argv[1] === __filename) {
+	main().catch((error) => {
+		console.error(error);
+		process.exit(1);
+	});
+}