feat: auto-generate OpenRouter model registry from API + add missing models (#1407) (#1426)

Added scripts/generate-openrouter-models.mjs that fetches the full model
list from OpenRouter's API and generates TypeScript entries matching the
existing models.generated.ts format. Run with:
  node scripts/generate-openrouter-models.mjs > /tmp/openrouter.ts

Updated the OpenRouter section in models.generated.ts from 241 → 350
models, including all nvidia/nemotron variants requested in the issue.

Fixes #1407

scripts/generate-openrouter-models.mjs

@@ -0,0 +1,93 @@
+#!/usr/bin/env node
+/**
+ * Generate OpenRouter model entries for models.generated.ts
+ *
+ * Fetches the full model list from OpenRouter's API and generates
+ * TypeScript model entries matching the existing registry format.
+ *
+ * Usage: node scripts/generate-openrouter-models.mjs > /tmp/openrouter-models.ts
+ *
+ * The output is a partial TypeScript object that can be merged into
+ * packages/pi-ai/src/models.generated.ts under the "openrouter" key.
+ */
+
+const API_URL = "https://openrouter.ai/api/v1/models";
+
+async function fetchModels() {
+  const resp = await fetch(API_URL);
+  if (!resp.ok) throw new Error(`API returned ${resp.status}`);
+  const data = await resp.json();
+  return data.data || [];
+}
+
+function inferApi(model) {
+  // Models that support the responses API
+  if (model.id.startsWith("openai/") || model.id.startsWith("anthropic/")) {
+    return "openai-completions";
+  }
+  return "openai-completions";
+}
+
+function inferReasoning(model) {
+  const id = model.id.toLowerCase();
+  return id.includes("o1") || id.includes("o3") || id.includes("o4") ||
+    id.includes("reasoning") || id.includes("think");
+}
+
+function inferInput(model) {
+  const arch = model.architecture || {};
+  const modality = (arch.input_modalities || []).join(",").toLowerCase();
+  if (modality.includes("image")) return '["text", "image"]';
+  return '["text"]';
+}
+
+function formatCost(pricing) {
+  if (!pricing) return { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 };
+  // OpenRouter pricing is per-token in dollars; our format is per-million-tokens
+  const toPerMillion = (v) => Math.round(parseFloat(v || "0") * 1_000_000 * 100) / 100;
+  return {
+    input: toPerMillion(pricing.prompt),
+    output: toPerMillion(pricing.completion),
+    cacheRead: 0,
+    cacheWrite: 0,
+  };
+}
+
+async function main() {
+  const models = await fetchModels();
+
+  console.log('\t"openrouter": {');
+
+  for (const m of models.sort((a, b) => a.id.localeCompare(b.id))) {
+    const cost = formatCost(m.pricing);
+    const contextWindow = m.context_length || 128000;
+    const maxOutput = m.top_provider?.max_completion_tokens || Math.min(contextWindow, 16384);
+    const reasoning = inferReasoning(m);
+    const input = inferInput(m);
+
+    console.log(`\t\t"${m.id}": {`);
+    console.log(`\t\t\tid: "${m.id}",`);
+    console.log(`\t\t\tname: ${JSON.stringify(m.name || m.id)},`);
+    console.log(`\t\t\tapi: "${inferApi(m)}",`);
+    console.log(`\t\t\tprovider: "openrouter",`);
+    console.log(`\t\t\tbaseUrl: "https://openrouter.ai/api/v1",`);
+    console.log(`\t\t\treasoning: ${reasoning},`);
+    console.log(`\t\t\tinput: ${input},`);
+    console.log(`\t\t\tcost: {`);
+    console.log(`\t\t\t\tinput: ${cost.input},`);
+    console.log(`\t\t\t\toutput: ${cost.output},`);
+    console.log(`\t\t\t\tcacheRead: ${cost.cacheRead},`);
+    console.log(`\t\t\t\tcacheWrite: ${cost.cacheWrite},`);
+    console.log(`\t\t\t},`);
+    console.log(`\t\t\tcontextWindow: ${contextWindow},`);
+    console.log(`\t\t\tmaxOutput: ${maxOutput},`);
+    console.log(`\t\t},`);
+  }
+
+  console.log("\t},");
+}
+
+main().catch(err => {
+  console.error(err);
+  process.exit(1);
+});