From 8939d8af379b920b32c02a227b8a4c9bbe3824ce Mon Sep 17 00:00:00 2001
From: Danny Avila <danny@librechat.ai>
Date: Fri, 31 May 2024 12:16:08 -0400
Subject: [PATCH] =?UTF-8?q?=F0=9F=A6=99=20feat:=20Add=20LLama=203=20System?=
 =?UTF-8?q?=20Context=20Length=20(#2938)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 api/utils/tokens.js      |  2 ++
 api/utils/tokens.spec.js | 12 ++++++++++++
 2 files changed, 14 insertions(+)

diff --git a/api/utils/tokens.js b/api/utils/tokens.js
index b0d9098821..3b973c40cf 100644
--- a/api/utils/tokens.js
+++ b/api/utils/tokens.js
@@ -59,6 +59,8 @@ const openAIModels = {
   'gpt-3.5-turbo-1106': 16375, // -10 from max
   'gpt-3.5-turbo-0125': 16375, // -10 from max
   'mistral-': 31990, // -10 from max
+  llama3: 8187, // -5 from max
+  'llama-3': 8187, // -5 from max
 };
 
 const cohereModels = {
diff --git a/api/utils/tokens.spec.js b/api/utils/tokens.spec.js
index c51610c917..228e298605 100644
--- a/api/utils/tokens.spec.js
+++ b/api/utils/tokens.spec.js
@@ -20,6 +20,18 @@ describe('getModelMaxTokens', () => {
     );
   });
 
+  test('should return correct tokens for LLama 3 models', () => {
+    expect(getModelMaxTokens('meta-llama/llama-3-8b')).toBe(
+      maxTokensMap[EModelEndpoint.openAI]['llama-3'],
+    );
+    expect(getModelMaxTokens('meta-llama/llama-3-8b')).toBe(
+      maxTokensMap[EModelEndpoint.openAI]['llama3'],
+    );
+    expect(getModelMaxTokens('llama-3-500b')).toBe(maxTokensMap[EModelEndpoint.openAI]['llama-3']);
+    expect(getModelMaxTokens('llama3-70b')).toBe(maxTokensMap[EModelEndpoint.openAI]['llama3']);
+    expect(getModelMaxTokens('llama3:latest')).toBe(maxTokensMap[EModelEndpoint.openAI]['llama3']);
+  });
+
   test('should return undefined for no match', () => {
     expect(getModelMaxTokens('unknown-model')).toBeUndefined();
   });