Refine title generation prompt and parameters

- Update title prompt to be more specific about single title generation
- Reduce max_tokens from 50 to 20 for shorter titles
- Improve system message to emphasize single title output
- Add word limit guidance in prompt

src/summarization.py

@@ -206,7 +206,7 @@ def summarize_transcript_langchain(transcript: str, selected_gguf_model: str, pr
 
 def create_title_prompt() -> PromptTemplate:
     """Prompt for generating a document title"""
-    template = """Generate a concise, descriptive title for this transcript. The title should capture the main topic or theme.
+    template = """Generate a single, concise title for this transcript that captures the main topic or theme. Keep it under 10 words.
 
 Transcript:
 {text}
@@ -243,11 +243,11 @@ def generate_title(transcript: str, selected_gguf_model: str) -> str:
         # Generate title
         response = llm.create_chat_completion(
             messages=[
-                {"role": "system", "content": "You are an expert at creating concise, descriptive titles for documents and transcripts."},
+                {"role": "system", "content": "You are an expert at creating single, concise titles for documents and transcripts. Always provide exactly one title, nothing else."},
                 {"role": "user", "content": formatted_prompt}
             ],
             stream=False,
-            max_tokens=50,  # Limit title length
+            max_tokens=20,  # Very short for titles
         )
 
         title = response['choices'][0]['message']['content'].strip()