Update app.py

app.py

@@ -8,23 +8,7 @@ import re
 import requests
 import transformers
 import chardet
-#import deepeval
-try:
-    import deepeval
-    from deepeval import evaluate
-    from deepeval.metrics import AnswerRelevancyMetric
-    from deepeval.test_case import LLMTestCase
-    DEEPEVAL_AVAILABLE = True
-except Exception as e:
-    print(f"DeepEval import failed: {e}, continuing without evaluation features")
-    DEEPEVAL_AVAILABLE = False
-    # 創建備用函數
-    def evaluate(*args, **kwargs):
-        return []
-    class AnswerRelevancyMetric:
-        def __init__(self, *args, **kwargs): pass
-    class LLMTestCase:
-        def __init__(self, *args, **kwargs): pass
+import deepeval
 import difflib
 from transformers import AutoModelForCausalLM, AutoTokenizer
 from transformers.models.llama.configuration_llama import LlamaConfig
@@ -71,19 +55,15 @@ tempfile.tempdir = "/tmp"
 # 在此處加入 DeepEval 的 monkey-patch，避免全域更改工作目錄
 original_evaluate = deepeval.evaluate
 
-# 在此處加入 DeepEval 的 monkey-patch，避免全域更改工作目錄
-if DEEPEVAL_AVAILABLE:
-    original_evaluate = deepeval.evaluate
-
-    def patched_evaluate(*args, **kwargs):
-        current_dir = os.getcwd()
-        try:
-            os.chdir("/tmp")
-            return original_evaluate(*args, **kwargs)
-        finally:
-            os.chdir(current_dir)
+def patched_evaluate(*args, **kwargs):
+    current_dir = os.getcwd()
+    try:
+        os.chdir("/tmp")
+        return original_evaluate(*args, **kwargs)
+    finally:
+        os.chdir(current_dir)
 
-    deepeval.evaluate = patched_evaluate
+deepeval.evaluate = patched_evaluate
 
 
 SHOW_EVAL = os.getenv("SHOW_EVAL", "false").lower() == "true"
@@ -169,7 +149,7 @@ print(f"Using device => {device}")
 
 hf_token = os.environ.get("HF_TOKEN")
 openai_api_key = os.environ.get("OPENAI_API_KEY")
-model_id = "meta-llama/Llama-3.2-3B-Instruct"
+model_id = "ChienChung/my-llama-1b"
 
 config_path = hf_hub_download(
     repo_id=model_id,
@@ -294,7 +274,7 @@ def rag_llama_qa(query):
 def rag_gpt4_qa(query):
     raw_answer = qa_gpt.run(query)
 
-    if SHOW_EVAL and DEEPEVAL_AVAILABLE:
+    if SHOW_EVAL:
         try:
             top_docs = retriever.get_relevant_documents(query)
             test_case = LLMTestCase(
@@ -336,7 +316,7 @@ def upload_and_chat(file, query):
         chain_type_kwargs={"prompt": custom_prompt}
     )
     raw_answer = qa_temp.run(query)
-    if SHOW_EVAL and DEEPEVAL_AVAILABLE:
+    if SHOW_EVAL:
         try:
             test_case = LLMTestCase(
                 input=query,
@@ -1232,7 +1212,7 @@ def multi_agent_chat_advanced(query: str, file=None) -> str:
                         #answer = session_graph_chain(query)["answer"]
 
                         # ✅ DeepEval 評估僅在 Tab1 文件 QA 的情況下觸發
-                        if SHOW_EVAL and DEEPEVAL_AVAILABLE:
+                        if SHOW_EVAL:
                             try:
                                 test_case = LLMTestCase(
                                     input=query,