From 740f6548a25f2deccd4d2ec5e7e4a3312435d050 Mon Sep 17 00:00:00 2001
From: Cohee <18619528+Cohee1207@users.noreply.github.com>
Date: Wed, 8 Nov 2023 12:07:14 +0200
Subject: [PATCH] Increase timeouts of OAI out of quota requests

---
 public/scripts/openai.js | 2 +-
 server.js                | 3 ++-
 2 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/public/scripts/openai.js b/public/scripts/openai.js
index d42436b08..9420eb8c6 100644
--- a/public/scripts/openai.js
+++ b/public/scripts/openai.js
@@ -110,7 +110,7 @@ const max_4k = 4095;
 const max_8k = 8191;
 const max_16k = 16383;
 const max_32k = 32767;
-const max_128k = 128 * 1024 - 1;
+const max_128k = 128 * 1000;
 const scale_max = 8191;
 const claude_max = 9000; // We have a proper tokenizer, so theoretically could be larger (up to 9k)
 const palm2_max = 7500; // The real context window is 8192, spare some for padding due to using turbo tokenizer
diff --git a/server.js b/server.js
index 6ebe925b9..02aa7a1bd 100644
--- a/server.js
+++ b/server.js
@@ -3222,7 +3222,8 @@ app.post("/generate_openai", jsonParser, function (request, response_generate_op
             } else if (fetchResponse.status === 429 && retries > 0) {
                 console.log(`Out of quota, retrying in ${Math.round(timeout / 1000)}s`);
                 setTimeout(() => {
-                    makeRequest(config, response_generate_openai, request, retries - 1);
+                    timeout *= 2;
+                    makeRequest(config, response_generate_openai, request, retries - 1, timeout);
                 }, timeout);
             } else {
                 await handleErrorResponse(fetchResponse);