fix: Detailed API error reporting for agent analysis

backend/app/services/llm.py

@@ -120,11 +120,24 @@ def generate_analysis(tender: Tender, company: CompanyProfile, document_text: st
         return analysis
 
     prompt = _build_analysis_prompt(tender, company, document_text)
-    output = call_gemini(prompt)
     
+    output = ""
+    error_detail = ""
+    try:
+        model = get_gemini_model()
+        response = model.generate_content(prompt)
+        output = response.text
+    except Exception as e:
+        error_detail = str(e)
+        print(f"Error calling Gemini: {e}")
+
     if not output:
         analysis = generate_mock_analysis(tender, company)
-        analysis.audit_log = ["❌ Error: Gemini API returned an empty response.", "Check quota or region restrictions.", "Fallback: Using mock analysis."]
+        analysis.audit_log = [
+            f"❌ API Error: {error_detail or 'Empty response'}",
+            "Possible causes: Quota limit, Invalid Key, or Region restriction.",
+            "Fallback: Using system default analysis."
+        ]
         return analysis
 
     parse_result = _parse_gemini_response(output)