Enhanced ChatGPT Clone: Features Agents, MCP, Skills, DeepSeek, Anthropic, AWS, OpenAI, Responses API, Azure, Groq, o1, GPT-5, Mistral, OpenRouter, Vertex AI, Gemini, Artifacts, AI model switching, message search, Code Interpreter, langchain, DALL-E-3, OpenAPI Actions, Functions, Secure Multi-User Auth, Presets, open-source for self-hosting. Active
TypeScript aianthropicartifactsawsazurechatgptchatgpt-cloneclaudeclonedeepseekgeminigooglegpt-5librechatmcpo1openairesponses-apivisionwebui
import { logger } from '@librechat/data-schemas';
import type { Request, Response, NextFunction } from 'express';
import { buildRequestContext, runWithTenantContext } from './tenant';
import { rejectTenantHeader } from './tenantHeader';
import { isEnabled } from '~/utils';
/**
* Pre-authentication tenant context middleware for unauthenticated routes.
*
* Reads the tenant identifier from the `X-Tenant-Id` request header and wraps
* downstream handlers in `tenantStorage.run()` so that Mongoose queries and
* config resolution run within the correct tenant scope.
*
* **Where to use**: Mount on routes that must be tenant-aware before
* authentication has occurred:
* - `GET /api/config` — login page needs tenant-specific config (social logins, registration)
* - `/api/auth/*` — login, register, password reset
* - `/oauth/*` — OAuth callback flows
* - `GET /api/share/:shareId` — public shared conversation links
*
* **How the header gets set**: The deployment's reverse proxy, auth gateway,
* or OpenID strategy sets `X-Tenant-Id` based on subdomain, path, or OIDC claim.
* This middleware does NOT resolve tenants from subdomains or tokens — that is
* the responsibility of the deployment layer. Header-based resolution is disabled
* unless the operator explicitly sets `TRUST_TENANT_HEADER=true`.
*
* **Design**: Intentionally minimal. No subdomain parsing, no OIDC claim
* extraction, no YAML-driven strategy. Multi-tenant deployments can:
* 1. Set the header in the reverse proxy / ingress (simplest),
* 2. Replace this middleware's resolver logic entirely, or
* 3. Layer additional resolution on top (e.g., OpenID `tenant` claim → header).
*
* If no header is present, downstream runs without tenant ALS context (same as
* single-tenant mode), while request logging context can still propagate.
*/
export function preAuthTenantMiddleware(req: Request, res: Response, next: NextFunction): void {
const raw = req.headers['x-tenant-id'];
const requestContext = buildRequestContext(req);
if (!raw || typeof raw !== 'string' || !isEnabled(process.env.TRUST_TENANT_HEADER)) {
runWithTenantContext(requestContext, next);
return;
}
const tenantId = raw.trim();
if (!tenantId) {
runWithTenantContext(requestContext, next);
return;
}
const rejection = rejectTenantHeader(tenantId);
if (rejection === 'system') {
runWithTenantContext(requestContext, () => {
logger.warn('[preAuthTenant] Rejected __SYSTEM__ sentinel in X-Tenant-Id header', {
ip: req.ip,
path: req.path,
});
next();
});
return;
}
if (rejection === 'malformed') {
runWithTenantContext(requestContext, () => {
logger.warn('[preAuthTenant] Rejected malformed X-Tenant-Id header', {
ip: req.ip,
length: tenantId.length,
path: req.path,
});
next();
});
return;
}
runWithTenantContext({ ...requestContext, tenantId }, next);
}