fix: prevent abuse of preview chat endpoints via rate limiting by utsav1213 · Pull Request #2463 · baptisteArno/typebot.io

utsav1213 · 2026-04-16T09:57:21Z

🐛 Problem

Preview chat endpoints (/preview/startChat and /sessions/continueChat) bypass usage tracking and do not increment totalChatsUsed.

This allows users to abuse preview sessions as production chats without hitting plan limits.

✅ Solution

Introduced a Redis-backed rate limiter to restrict abuse of preview endpoints without affecting legitimate builder usage.

Applied rate limiting to preview session creation (startChatPreview)
Uses @upstash/ratelimit with a limit of 20 requests per minute
Rate limiting is applied per authenticated user ID, or per IP address as a fallback
Throws a TOO_MANY_REQUESTS error when the limit is exceeded
Designed to prevent automated abuse while preserving normal developer experience

⚠️ Note

This does not affect normal builder usage, as limits are high enough for manual testing but prevent high-frequency automated abuse.

🧪 Testing

Run local environment (docker compose up -d + bun run dev)
Open preview repeatedly or hit /preview/startChat endpoint directly
After ~20 requests/minute, API returns TOO_MANY_REQUESTS

📌 Future Improvements

Extend rate limiting to /sessions/continueChat to fully prevent bypass via existing preview sessions
Make limits configurable per environment

vercel · 2026-04-16T09:57:27Z

@utsav1213 is attempting to deploy a commit to the Typebot Team on Vercel.

A member of the Team first needs to authorize it.

Copilot

Pull request overview

Adds server-side rate limiting intended to prevent abuse of preview chat endpoints, addressing a gap where preview sessions could be used without impacting plan usage/limits.

Changes:

Passes the incoming Request through ORPC contexts so handlers can access request metadata (e.g., headers/IP).
Introduces a Redis-backed Upstash Ratelimit helper and applies it to startChat preview handling.
Adds dependencies/lockfile updates for @upstash/ratelimit and ioredis.

Reviewed changes

Copilot reviewed 8 out of 9 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
packages/config/src/orpc/viewer/context.ts	Exposes `req` on viewer ORPC context.
packages/config/src/orpc/builder/context.ts	Adds optional `req` to builder ORPC context.
packages/bot-engine/src/lib/rateLimiter.ts	New Redis-backed rate limiter singleton for preview start.
packages/bot-engine/src/api/handleStartChatPreview.ts	Enforces rate limiting before starting preview sessions.
apps/builder/src/app/api/orpc/[[...rest]]/route.ts	Passes request into builder ORPC context.
apps/builder/src/app/api/[[...rest]]/route.ts	Passes request into builder API context.
packages/bot-engine/package.json	Adds `@upstash/ratelimit` dependency.
package.json	Adds `ioredis` (root).
bun.lock	Lockfile updates for new/updated dependencies.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-04-16T10:01:38Z

  context: Context;
 }) => {
+  if (startPreviewRateLimiter) {
+    const ip = req ? getIp(req) : undefined;


getIp expects an object with the x-forwarded-for / cf-connecting-ip header values, but this code passes the Request object directly. As a result ip will always be null/undefined and the limiter key will fall back to user?.id or "unknown", which defeats the intended per-IP limiting. Build the header object from req.headers (like other call sites do) and then pass that into getIp.

Suggested change

const ip = req ? getIp(req) : undefined;

const headers = req?.headers

? {

"x-forwarded-for":

typeof req.headers.get === "function"

? (req.headers.get("x-forwarded-for") ?? undefined)

: req.headers["x-forwarded-for"],

"cf-connecting-ip":

typeof req.headers.get === "function"

? (req.headers.get("cf-connecting-ip") ?? undefined)

: req.headers["cf-connecting-ip"],

}

: undefined;

const ip = headers ? getIp(headers) : undefined;

Copilot · 2026-04-16T10:01:38Z

+    const id = user?.id ?? ip ?? "unknown";
+    const { success } = await startPreviewRateLimiter.limit(id);


The new rate limit key uses user?.id before ip, but the PR description says preview requests are limited per IP. If per-IP behavior is intended, prefer using IP as the primary identifier (and only fall back to user ID when IP is unavailable), or combine both into a stable key.

Copilot · 2026-04-16T10:01:39Z

+  if (startPreviewRateLimiter) {
+    const ip = req ? getIp(req) : undefined;
+    const id = user?.id ?? ip ?? "unknown";
+    const { success } = await startPreviewRateLimiter.limit(id);
+    if (!success) {
+      throw new ORPCError("TOO_MANY_REQUESTS", {
+        message: "Rate limit exceeded. Please try again later.",
+      });
+    }
+  }


The PR description mentions rate limiting both /preview/startChat and /sessions/continueChat, but the limiter is only enforced here in startChatPreview. continueChat can still be used indefinitely once a preview session is created, which leaves a large portion of the abuse path unprotected. Consider applying a similar limiter in the /v1/sessions/{sessionId}/continueChat handler for preview sessions (e.g., detect preview sessions and rate limit by IP).

Copilot · 2026-04-16T10:01:39Z


 type Context = {
  user: { id: string } | null;
+  req?: any;


req is typed as any in the handler context, which hides type issues (and contributed to the incorrect getIp(req) usage). Since createContext now passes a Fetch Request, type this as req?: Request (or a minimal { headers: Headers } shape) so header access and IP extraction stay type-safe.

Suggested change

req?: any;

req?: Request;

Copilot · 2026-04-16T10:01:39Z

+declare const window: { startPreviewRateLimiter: Ratelimit | undefined };
+
+const windowAny = globalThis as any;
+
+if (!windowAny.startPreviewRateLimiter && env.REDIS_URL) {
+  windowAny.startPreviewRateLimiter = createRateLimiter({
+    requests: 20,
+    window: "1 m",
+    prefix: "start-preview-ratelimit",
+  });
+}
+
+export const startPreviewRateLimiter = windowAny.startPreviewRateLimiter as Ratelimit | undefined;


The global caching pattern here is inconsistent with other rate limiter modules in the repo (they use declare const global: { ... } and assign on global). Declaring window in a server-side module is also confusing and may conflict with DOM typings in some TS configs. Consider switching to the existing global pattern and dropping the window declaration entirely.

Suggested change

declare const window: { startPreviewRateLimiter: Ratelimit | undefined };

const windowAny = globalThis as any;

if (!windowAny.startPreviewRateLimiter && env.REDIS_URL) {

windowAny.startPreviewRateLimiter = createRateLimiter({

requests: 20,

window: "1 m",

prefix: "start-preview-ratelimit",

});

}

export const startPreviewRateLimiter = windowAny.startPreviewRateLimiter as Ratelimit | undefined;

declare const global: {

startPreviewRateLimiter: Ratelimit | undefined;

};

if (!global.startPreviewRateLimiter && env.REDIS_URL) {

global.startPreviewRateLimiter = createRateLimiter({

requests: 20,

window: "1 m",

prefix: "start-preview-ratelimit",

});

}

export const startPreviewRateLimiter = global.startPreviewRateLimiter;

Copilot · 2026-04-16T10:01:39Z

+  if (startPreviewRateLimiter) {
+    const ip = req ? getIp(req) : undefined;
+    const id = user?.id ?? ip ?? "unknown";
+    const { success } = await startPreviewRateLimiter.limit(id);
+    if (!success) {
+      throw new ORPCError("TOO_MANY_REQUESTS", {
+        message: "Rate limit exceeded. Please try again later.",
+      });
+    }
+  }


There are Playwright API tests covering preview startChat/continueChat, but none validating the new rate limiting behavior. Adding a test that the preview endpoint returns a 429/TOO_MANY_REQUESTS after the threshold (with the limiter enabled) would prevent regressions and ensure the mitigation works as intended.

fix(bot-engine): enforce rate limits on preview chat endpoints

dac94f0

Copilot AI review requested due to automatic review settings April 16, 2026 09:57

Copilot started reviewing on behalf of utsav1213 April 16, 2026 09:57 View session

Copilot AI reviewed Apr 16, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix: prevent abuse of preview chat endpoints via rate limiting#2463

fix: prevent abuse of preview chat endpoints via rate limiting#2463
utsav1213 wants to merge 1 commit intobaptisteArno:mainfrom
utsav1213:fix/preview-rate-limit

utsav1213 commented Apr 16, 2026 •

edited

Loading

Uh oh!

vercel bot commented Apr 16, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 16, 2026

Uh oh!

Copilot AI Apr 16, 2026

Uh oh!

Copilot AI Apr 16, 2026

Uh oh!

Copilot AI Apr 16, 2026

Uh oh!

Copilot AI Apr 16, 2026

Uh oh!

Copilot AI Apr 16, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

-    const ip = req ? getIp(req) : undefined;
+    const headers = req?.headers
+      ? {
+          "x-forwarded-for":
+            typeof req.headers.get === "function"
+              ? (req.headers.get("x-forwarded-for") ?? undefined)
+              : req.headers["x-forwarded-for"],
+          "cf-connecting-ip":
+            typeof req.headers.get === "function"
+              ? (req.headers.get("cf-connecting-ip") ?? undefined)
+              : req.headers["cf-connecting-ip"],
+        }
+      : undefined;
+    const ip = headers ? getIp(headers) : undefined;

		const id = user?.id ?? ip ?? "unknown";
		const { success } = await startPreviewRateLimiter.limit(id);

Uh oh!

Conversation

utsav1213 commented Apr 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🐛 Problem

✅ Solution

⚠️ Note

🧪 Testing

📌 Future Improvements

Uh oh!

vercel bot commented Apr 16, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 16, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 16, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 16, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 16, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 16, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 16, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

utsav1213 commented Apr 16, 2026 •

edited

Loading