Update Google Gemini model versions (#176)

feat(env): ✨ update Google Gemini model versions

* Changed `GOOGLE_GEMINI` to `2.0_flash` in `.env.example`.
* Updated model names to `gemini-2.5-pro` in `eval_main.py` and `vertex_ai.py`.
* Modified model instantiation to use `gemini-2.0-flash` and `gemini-2.5-pro` in `gemini_model.py`.

Signed-off-by: Jack Luar <[email protected]>

Author: luarss

backend/.env.example

@@ -17,8 +17,8 @@ LLM_MODEL=gemini
 # Specify model name if using Ollama
 OLLAMA_MODEL=
 
-# Set Google Gemini model version
-GOOGLE_GEMINI=1.5_flash
+# Set Google Gemini model version (options: 2.0_flash, 2.5_flash, 2.5_pro)
+GOOGLE_GEMINI=2.0_flash
 
 LLM_TEMP=1
 

evaluation/auto_evaluation/eval_main.py

@@ -45,7 +45,7 @@ def __init__(self, base_url: str, dataset: str, reranker_base_url: str = ""):
         self.reranker_base_url = reranker_base_url
         self.qns = preprocess.read_data(self.dataset)
         self.eval_model = GeminiModel(
-            model_name="gemini-1.5-pro-002",
+            model_name="gemini-2.5-pro",
             project=os.getenv("GOOGLE_PROJECT_ID", ""),
             location=os.getenv("GOOGLE_CLOUD_LOCATION", "us-central1"),
         )

evaluation/auto_evaluation/src/models/vertex_ai.py

@@ -76,15 +76,15 @@ def get_model_name(self):
 
 
 def main():
-    model = GoogleVertexAILangChain(model_name="gemini-1.5-pro-002")
+    model = GoogleVertexAILangChain(model_name="gemini-2.5-pro")
     prompt = "Write me a joke"
     print(f"Prompt: {prompt}")
     response = model.generate(prompt, schema=Response)
     print(f"Response: {response}")
 
 
 async def main_async():
-    model = GoogleVertexAILangChain(model_name="gemini-1.5-pro-002")
+    model = GoogleVertexAILangChain(model_name="gemini-2.5-pro")
     prompt = "Write me a joke"
     print(f"Prompt: {prompt}")
     response = await model.a_generate(prompt, schema=Response)

evaluation/script_based_evaluation/models/gemini_model.py

@@ -31,7 +31,7 @@ def base_gemini_1_5_flash(query: str) -> tuple[str, float]:
     ]
     while True:
         try:
-            model = genai.GenerativeModel("gemini-1.5-flash")
+            model = genai.GenerativeModel("gemini-2.0-flash")
             start_time = time.time()
             query = " " + query
             response = model.generate_content(query, safety_settings=safety_config)
@@ -73,7 +73,7 @@ def base_gemini_1_5_pro(query: str) -> tuple[str, float]:
     ]
     while True:
         try:
-            model = genai.GenerativeModel("gemini-1.5-pro")
+            model = genai.GenerativeModel("gemini-2.5-pro")
             start_time = time.time()
             query = " " + query
             response = model.generate_content(