refine markdown prompt (#2551)

### What problem does this PR solve?


### Type of change

- [x] Performance Improvement

api/db/services/dialog_service.py

@@ -180,7 +180,7 @@ def chat(dialog, messages, stream=True, **kwargs):
         yield {"answer": empty_res, "reference": kbinfos, "audio_binary": tts(tts_mdl, empty_res)}
         return {"answer": prompt_config["empty_response"], "reference": kbinfos}
 
-    kwargs["knowledge"] = "\n------\n".join(knowledges)
+    kwargs["knowledge"] = "\n\n------\n\n".join(knowledges)
     gen_conf = dialog.llm_setting
 
     msg = [{"role": "system", "content": prompt_config["system"].format(**kwargs)}]
@@ -221,7 +221,7 @@ def chat(dialog, messages, stream=True, **kwargs):
         if answer.lower().find("invalid key") >= 0 or answer.lower().find("invalid api") >= 0:
             answer += " Please set LLM API-Key in 'User Setting -> Model Providers -> API-Key'"
         done_tm = timer()
-        prompt += "\n### Elapsed\n  - Retrieval: %.1f ms\n  - LLM: %.1f ms"%((retrieval_tm-st)*1000, (done_tm-st)*1000)
+        prompt += "\n\n### Elapsed\n  - Retrieval: %.1f ms\n  - LLM: %.1f ms"%((retrieval_tm-st)*1000, (done_tm-st)*1000)
         return {"answer": answer, "reference": refs, "prompt": prompt}
 
     if stream: