LLM-Driven Extraction of Unstructured Data — Built for API Deployments & ETL Pipeline Workflows
Python ai-agentsdata-engineeringdocument-aigenerative-aiidpjson-extractionllmmcp-serverocrpdf-extractionprompt-engineeringstructured-output
UN-4021 [MISC] Let the Vite dev server run behind a TLS ingress (#2249)
Jaseem Jasda1a9e5ef860
diff --git a/frontend/vite.config.js b/frontend/vite.config.jsindex 12b3356f0..395fee7b5 100644--- a/frontend/vite.config.js+++ b/frontend/vite.config.js@@ -113,6 +113,18 @@ export default defineConfig(({ mode }) => {server: {host: "0.0.0.0",port: Number(env.PORT) || 3000,+ // Vite refuses requests whose Host header is not localhost or an IP, as a+ // DNS-rebinding guard. That is the right default locally, but when the dev+ // server runs inside a cluster pod behind an ingress the browser sends the+ // real domain and every request comes back "Blocked request". Opt in per+ // environment with a comma-separated list, e.g. ".example.com".+ // Empty list == Vite's default (localhost/IPs only), so local dev and the+ // production build are unaffected.+ allowedHosts: env.VITE_DEV_ALLOWED_HOSTS+ ? env.VITE_DEV_ALLOWED_HOSTS.split(",")+ .map((host) => host.trim())+ .filter(Boolean)+ : [],// Docker-specific: Enable polling for file watchingwatch: {usePolling: true,@@ -124,6 +136,13 @@ export default defineConfig(({ mode }) => {clientPort: env.WDS_SOCKET_PORT? Number(env.WDS_SOCKET_PORT): Number(env.PORT) || 3000,+ // Behind a TLS-terminating ingress the page is served over https, so the+ // HMR socket has to be wss on the ingress port (443) rather than the+ // port the dev server itself listens on. Left unset, Vite derives the+ // protocol from its own (plain http) server and the socket never opens.+ ...(env.VITE_DEV_HMR_PROTOCOL+ ? { protocol: env.VITE_DEV_HMR_PROTOCOL }+ : {}),},// Proxy configuration (similar to setupProxy.js in CRA)proxy: