From fe32952825a92c31d25ff19b4209632e0877fd5c Mon Sep 17 00:00:00 2001
From: Yongteng Lei <yongtengrey@outlook.com>
Date: Mon, 18 Aug 2025 14:51:10 +0800
Subject: [PATCH] Fix: Gemini parameters error (#9520)

### What problem does this PR solve?

Fix Gemini parameters error.

### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)

---------

Co-authored-by: Kevin Hu <kevinhu.sh@gmail.com>
---
 conf/llm_factories.json | 41 ++++++++---------------------------------
 rag/llm/cv_model.py     | 10 +++++-----
 2 files changed, 13 insertions(+), 38 deletions(-)

diff --git a/conf/llm_factories.json b/conf/llm_factories.json
index 78f100130..072185a9c 100644
--- a/conf/llm_factories.json
+++ b/conf/llm_factories.json
@@ -1146,60 +1146,35 @@
                     "llm_name": "gemini-2.5-flash",
                     "tags": "LLM,CHAT,1024K,IMAGE2TEXT",
                     "max_tokens": 1048576,
-                    "model_type": "image2text",
+                    "model_type": "chat",
                     "is_tools": true
                 },
                 {
                     "llm_name": "gemini-2.5-pro",
                     "tags": "LLM,CHAT,IMAGE2TEXT,1024K",
                     "max_tokens": 1048576,
-                    "model_type": "image2text",
+                    "model_type": "chat",
                     "is_tools": true
                 },
                 {
-                    "llm_name": "gemini-2.5-flash-preview-05-20",
+                    "llm_name": "gemini-2.5-flash-lite",
                     "tags": "LLM,CHAT,1024K,IMAGE2TEXT",
                     "max_tokens": 1048576,
-                    "model_type": "image2text",
+                    "model_type": "chat",
                     "is_tools": true
                 },
                 {
-                    "llm_name": "gemini-2.0-flash-001",
-                    "tags": "LLM,CHAT,1024K",
-                    "max_tokens": 1048576,
-                    "model_type": "image2text",
-                    "is_tools": true
-                },
-                {
-                    "llm_name": "gemini-2.0-flash-thinking-exp-01-21",
+                    "llm_name": "gemini-2.0-flash",
                     "tags": "LLM,CHAT,1024K",
                     "max_tokens": 1048576,
                     "model_type": "chat",
                     "is_tools": true
                 },
                 {
-                    "llm_name": "gemini-1.5-flash",
-                    "tags": "LLM,IMAGE2TEXT,1024K",
+                    "llm_name": "gemini-2.0-flash-lite",
+                    "tags": "LLM,CHAT,1024K",
                     "max_tokens": 1048576,
-                    "model_type": "image2text"
-                },
-                {
-                    "llm_name": "gemini-2.5-pro-preview-05-06",
-                    "tags": "LLM,IMAGE2TEXT,1024K",
-                    "max_tokens": 1048576,
-                    "model_type": "image2text"
-                },
-                {
-                    "llm_name": "gemini-1.5-pro",
-                    "tags": "LLM,IMAGE2TEXT,2048K",
-                    "max_tokens": 2097152,
-                    "model_type": "image2text"
-                },
-                {
-                    "llm_name": "gemini-1.5-flash-8b",
-                    "tags": "LLM,IMAGE2TEXT,1024K",
-                    "max_tokens": 1048576,
-                    "model_type": "image2text",
+                    "model_type": "chat",
                     "is_tools": true
                 },
                 {
diff --git a/rag/llm/cv_model.py b/rag/llm/cv_model.py
index a04ded780..4ca4ee190 100644
--- a/rag/llm/cv_model.py
+++ b/rag/llm/cv_model.py
@@ -539,24 +539,24 @@ class GeminiCV(Base):
         return res.text, res.usage_metadata.total_token_count
 
     def chat(self, system, history, gen_conf, images=[]):
-        from transformers import GenerationConfig
+        generation_config = dict(temperature=gen_conf.get("temperature", 0.3), top_p=gen_conf.get("top_p", 0.7))
         try:
             response = self.model.generate_content(
                 self._form_history(system, history, images),
-                generation_config=GenerationConfig(temperature=gen_conf.get("temperature", 0.3), top_p=gen_conf.get("top_p", 0.7)))
+                generation_config=generation_config)
             ans = response.text
             return ans, response.usage_metadata.total_token_count
         except Exception as e:
             return "**ERROR**: " + str(e), 0
 
     def chat_streamly(self, system, history, gen_conf, images=[]):
-        from transformers import GenerationConfig
         ans = ""
         response = None
         try:
+            generation_config = dict(temperature=gen_conf.get("temperature", 0.3), top_p=gen_conf.get("top_p", 0.7))
             response = self.model.generate_content(
                 self._form_history(system, history, images),
-                generation_config=GenerationConfig(temperature=gen_conf.get("temperature", 0.3), top_p=gen_conf.get("top_p", 0.7)),
+                generation_config=generation_config,
                 stream=True,
             )
 
@@ -572,7 +572,7 @@ class GeminiCV(Base):
             yield response.usage_metadata.total_token_count
         else:
             yield 0
-            
+
 
 class NvidiaCV(Base):
     _FACTORY_NAME = "NVIDIA"