Function main

packages/core/script/generate-databricks.ts:179–289

()

];

async function main() {
  console.log(
    `${dryRun ? "[DRY RUN] " : ""}${newOnly ? "[NEW ONLY] " : ""}Fetching Databricks foundation-models...`,
  );

  const url = `https://${workspace}/api/2.0/serving-endpoints:foundation-models`;
  const res = await fetch(url, {
    headers: { Authorization: `Bearer ${token}` },
  });
  if (!res.ok) {
    console.error(`Failed to fetch API: ${res.status} ${res.statusText}`);
    console.error(await res.text().catch(() => ""));
    process.exit(1);
  }

  const json = await res.json();
  const parsed = FoundationModelsResponse.safeParse(json);
  if (!parsed.success) {
    console.error("Invalid API response:", parsed.error.errors);
    process.exit(1);
  }

  const endpoints = parsed.data.endpoints.filter(
    (e) =>
      !IGNORE_PREFIXES.some((p) => e.name.startsWith(p)) &&
      e.config?.served_entities?.some(
        (se) =>
          se.foundation_model?.ai_gateway_v2_supported === true &&
          se.foundation_model?.api_types?.includes("mlflow/v1/chat/completions"),
      ),
  );

  const existingFiles = new Set<string>();
  try {
    for await (const f of new Bun.Glob("*.toml").scan({ cwd: MODELS_DIR })) {
      existingFiles.add(f);
    }
  } catch {
    // directory may not exist yet
  }

  console.log(
    `Found ${endpoints.length} models in API, ${existingFiles.size} existing files\n`,
  );

  const apiModelIds = new Set<string>();
  let created = 0;
  let updated = 0;
  let unchanged = 0;

  for (const ep of endpoints) {
    const filename = `${ep.name}.toml`;
    apiModelIds.add(filename);
    const filePath = path.join(MODELS_DIR, filename);

    const resolution = await resolveCanonical(ep.name);
    const newContent = formatToml(resolution, ep.name);
    const tag = resolution?.type === "base_model" ? `base_model ${resolution.from}` : resolution?.type ?? "stub";