pengjunjie/ragflow
1
0
Fork 1
mirror of https://github.com/infiniflow/ragflow.git synced 2025-12-08 20:42:30 +08:00
Code Issues Packages Projects Releases Wiki Activity

Refa: cleanup synchronous functions in chat_model and implement synchronization for conversation and dialog chats (#11779)

Browse Source 
### What problem does this PR solve?

Cleanup synchronous functions in chat_model and implement
synchronization for conversation and dialog chats.

### Type of change

- [x] Refactoring
- [x] Performance Improvement
This commit is contained in:
Yongteng Lei
2025-12-08 09:43:03 +08:00
committed by GitHub
parent 9b8971a9de
commit 51ec708c58
10 changed files with 421 additions and 843 deletions
Download Patch File Download Diff File Expand all files Collapse all files

33
api/db/services/dialog_service.py
View File

@ -178,7 +178,8 @@ class DialogService(CommonService):
offset += limit
return res
def chat_solo(dialog, messages, stream=True):
async def async_chat_solo(dialog, messages, stream=True):
attachments = ""
if "files" in messages[-1]:
attachments = "\n\n".join(FileService.get_files(messages[-1]["files"]))
@ -197,7 +198,8 @@ def chat_solo(dialog, messages, stream=True):
if stream:
last_ans = ""
delta_ans = ""
for ans in chat_mdl.chat_streamly(prompt_config.get("system", ""), msg, dialog.llm_setting):
answer = ""
async for ans in chat_mdl.async_chat_streamly(prompt_config.get("system", ""), msg, dialog.llm_setting):
answer = ans
delta_ans = ans[len(last_ans):]
if num_tokens_from_string(delta_ans) < 16:
@ -208,7 +210,7 @@ def chat_solo(dialog, messages, stream=True):
if delta_ans:
yield {"answer": answer, "reference": {}, "audio_binary": tts(tts_mdl, delta_ans), "prompt": "", "created_at": time.time()}
else:
answer = chat_mdl.chat(prompt_config.get("system", ""), msg, dialog.llm_setting)
answer = await chat_mdl.async_chat(prompt_config.get("system", ""), msg, dialog.llm_setting)
user_content = msg[-1].get("content", "[content not available]")
logging.debug("User: {}|Assistant: {}".format(user_content, answer))
yield {"answer": answer, "reference": {}, "audio_binary": tts(tts_mdl, answer), "prompt": "", "created_at": time.time()}
@ -347,13 +349,12 @@ def meta_filter(metas: dict, filters: list[dict], logic: str = "and"):
return []
return list(doc_ids)
def chat(dialog, messages, stream=True, **kwargs):
async def async_chat(dialog, messages, stream=True, **kwargs):
assert messages[-1]["role"] == "user", "The last content of this conversation is not from user."
if not dialog.kb_ids and not dialog.prompt_config.get("tavily_api_key"):
for ans in chat_solo(dialog, messages, stream):
async for ans in async_chat_solo(dialog, messages, stream):
yield ans
return None
return
chat_start_ts = timer()
@ -400,7 +401,7 @@ def chat(dialog, messages, stream=True, **kwargs):
ans = use_sql(questions[-1], field_map, dialog.tenant_id, chat_mdl, prompt_config.get("quote", True), dialog.kb_ids)
if ans:
yield ans
return None
return
for p in prompt_config["parameters"]:
if p["key"] == "knowledge":
@ -508,7 +509,8 @@ def chat(dialog, messages, stream=True, **kwargs):
empty_res = prompt_config["empty_response"]
yield {"answer": empty_res, "reference": kbinfos, "prompt": "\n\n### Query:\n%s" % " ".join(questions),
"audio_binary": tts(tts_mdl, empty_res)}
return {"answer": prompt_config["empty_response"], "reference": kbinfos}
yield {"answer": prompt_config["empty_response"], "reference": kbinfos}
return
kwargs["knowledge"] = "\n------\n" + "\n\n------\n\n".join(knowledges)
gen_conf = dialog.llm_setting
@ -612,7 +614,7 @@ def chat(dialog, messages, stream=True, **kwargs):
if stream:
last_ans = ""
answer = ""
for ans in chat_mdl.chat_streamly(prompt + prompt4citation, msg[1:], gen_conf):
async for ans in chat_mdl.async_chat_streamly(prompt + prompt4citation, msg[1:], gen_conf):
if thought:
ans = re.sub(r"^.*</think>", "", ans, flags=re.DOTALL)
answer = ans
@ -626,19 +628,19 @@ def chat(dialog, messages, stream=True, **kwargs):
yield {"answer": thought + answer, "reference": {}, "audio_binary": tts(tts_mdl, delta_ans)}
yield decorate_answer(thought + answer)
else:
answer = chat_mdl.chat(prompt + prompt4citation, msg[1:], gen_conf)
answer = await chat_mdl.async_chat(prompt + prompt4citation, msg[1:], gen_conf)
user_content = msg[-1].get("content", "[content not available]")
logging.debug("User: {}|Assistant: {}".format(user_content, answer))
res = decorate_answer(answer)
res["audio_binary"] = tts(tts_mdl, answer)
yield res
return None
return
def use_sql(question, field_map, tenant_id, chat_mdl, quota=True, kb_ids=None):
sys_prompt = """
You are a Database Administrator. You need to check the fields of the following tables based on the user's list of questions and write the SQL corresponding to the last question.
You are a Database Administrator. You need to check the fields of the following tables based on the user's list of questions and write the SQL corresponding to the last question.
Ensure that:
1. Field names should not start with a digit. If any field name starts with a digit, use double quotes around it.
2. Write only the SQL, no explanations or additional text.
@ -805,8 +807,7 @@ def tts(tts_mdl, text):
return None
return binascii.hexlify(bin).decode("utf-8")
def ask(question, kb_ids, tenant_id, chat_llm_name=None, search_config={}):
async def async_ask(question, kb_ids, tenant_id, chat_llm_name=None, search_config={}):
doc_ids = search_config.get("doc_ids", [])
rerank_mdl = None
kb_ids = search_config.get("kb_ids", kb_ids)
@ -880,7 +881,7 @@ def ask(question, kb_ids, tenant_id, chat_llm_name=None, search_config={}):
return {"answer": answer, "reference": refs}
answer = ""
for ans in chat_mdl.chat_streamly(sys_prompt, msg, {"temperature": 0.1}):
async for ans in chat_mdl.async_chat_streamly(sys_prompt, msg, {"temperature": 0.1}):
answer = ans
yield {"answer": answer, "reference": {}}
yield decorate_answer(answer)