TrialLayer

The platform is not a static data wrapper. It depends on scheduled ingestion, normalization, snapshots, and recalculated scores.

export const ctgovSync = task({
  id: "ctgov-full-sync",
  cron: "0 7 * * *",
  run: async () => {
    const trials = await fetchClinicalTrialsGov();
    const normalized = trials.map(parseTrialRecord);

    await upsertTrials(normalized);
    await createHistoricalSnapshots(normalized);
    await recalculateTrialHealthScores(normalized);
  },
});

Clerk, Stripe, usage limits, and protected routes turn the intelligence layer into an actual SaaS product.

const { userId } = await auth();
if (!userId) redirect("/sign-in");

const access = await getUserAccess(userId);
if (!access.canUseAiQuery) {
  return NextResponse.json(
    { error: "AI queries require a paid subscription" },
    { status: 402 },
  );
}

AI is used where it belongs: summarizing, querying, briefing, and explaining dense clinical context.

const response = await anthropic.messages.create({
  model: AI_MODEL,
  max_tokens: MAX_TOKENS_SYNTHESIZE,
  messages: [
    {
      role: "user",
      content: buildClinicalIntelligencePrompt(context),
    },
  ],
});