Spaces:
Build error
Build error
XThomasBU
commited on
Commit
·
2719f21
1
Parent(s):
9207578
fixed follow up logging
Browse files
code/main.py
CHANGED
|
@@ -395,11 +395,20 @@ class Chatbot:
|
|
| 395 |
|
| 396 |
if self.config["llm_params"]["generate_follow_up"]:
|
| 397 |
start_time = time.time()
|
| 398 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 399 |
query=user_query_dict["input"],
|
| 400 |
response=answer,
|
| 401 |
chat_history=res.get("chat_history"),
|
| 402 |
context=res.get("context"),
|
|
|
|
| 403 |
)
|
| 404 |
|
| 405 |
for question in list_of_questions:
|
|
|
|
| 395 |
|
| 396 |
if self.config["llm_params"]["generate_follow_up"]:
|
| 397 |
start_time = time.time()
|
| 398 |
+
config = {
|
| 399 |
+
"callbacks": (
|
| 400 |
+
[cl.LangchainCallbackHandler()]
|
| 401 |
+
if cl_data._data_layer and self.config["chat_logging"]["callbacks"]
|
| 402 |
+
else None
|
| 403 |
+
)
|
| 404 |
+
}
|
| 405 |
+
|
| 406 |
+
list_of_questions = await self.question_generator.generate_questions(
|
| 407 |
query=user_query_dict["input"],
|
| 408 |
response=answer,
|
| 409 |
chat_history=res.get("chat_history"),
|
| 410 |
context=res.get("context"),
|
| 411 |
+
config=config,
|
| 412 |
)
|
| 413 |
|
| 414 |
for question in list_of_questions:
|
code/modules/chat/langchain/langchain_rag.py
CHANGED
|
@@ -100,8 +100,8 @@ class QuestionGenerator:
|
|
| 100 |
def __init__(self):
|
| 101 |
pass
|
| 102 |
|
| 103 |
-
def generate_questions(self, query, response, chat_history, context):
|
| 104 |
-
questions = return_questions(query, response, chat_history, context)
|
| 105 |
return questions
|
| 106 |
|
| 107 |
|
|
@@ -204,7 +204,7 @@ class Langchain_RAG_V2(BaseRAG):
|
|
| 204 |
is_shared=True,
|
| 205 |
),
|
| 206 |
],
|
| 207 |
-
)
|
| 208 |
|
| 209 |
if callbacks is not None:
|
| 210 |
self.rag_chain = self.rag_chain.with_config(callbacks=callbacks)
|
|
|
|
| 100 |
def __init__(self):
|
| 101 |
pass
|
| 102 |
|
| 103 |
+
def generate_questions(self, query, response, chat_history, context, config):
|
| 104 |
+
questions = return_questions(query, response, chat_history, context, config)
|
| 105 |
return questions
|
| 106 |
|
| 107 |
|
|
|
|
| 204 |
is_shared=True,
|
| 205 |
),
|
| 206 |
],
|
| 207 |
+
).with_config(run_name="Langchain_RAG_V2")
|
| 208 |
|
| 209 |
if callbacks is not None:
|
| 210 |
self.rag_chain = self.rag_chain.with_config(callbacks=callbacks)
|
code/modules/chat/langchain/utils.py
CHANGED
|
@@ -280,7 +280,8 @@ def create_retrieval_chain(
|
|
| 280 |
return retrieval_chain
|
| 281 |
|
| 282 |
|
| 283 |
-
|
|
|
|
| 284 |
|
| 285 |
system = (
|
| 286 |
"You are someone that suggests a question based on the student's input and chat history. "
|
|
@@ -303,13 +304,17 @@ def return_questions(query, response, chat_history_str, context):
|
|
| 303 |
)
|
| 304 |
llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)
|
| 305 |
question_generator = prompt | llm | StrOutputParser()
|
| 306 |
-
|
|
|
|
|
|
|
|
|
|
| 307 |
{
|
| 308 |
"chat_history_str": chat_history_str,
|
| 309 |
"context": context,
|
| 310 |
"query": query,
|
| 311 |
"response": response,
|
| 312 |
-
}
|
|
|
|
| 313 |
)
|
| 314 |
|
| 315 |
list_of_questions = new_questions.split("...")
|
|
|
|
| 280 |
return retrieval_chain
|
| 281 |
|
| 282 |
|
| 283 |
+
# TODO: Remove Hard-coded values
|
| 284 |
+
async def return_questions(query, response, chat_history_str, context, config):
|
| 285 |
|
| 286 |
system = (
|
| 287 |
"You are someone that suggests a question based on the student's input and chat history. "
|
|
|
|
| 304 |
)
|
| 305 |
llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)
|
| 306 |
question_generator = prompt | llm | StrOutputParser()
|
| 307 |
+
question_generator = question_generator.with_config(
|
| 308 |
+
run_name="follow_up_question_generator"
|
| 309 |
+
)
|
| 310 |
+
new_questions = await question_generator.ainvoke(
|
| 311 |
{
|
| 312 |
"chat_history_str": chat_history_str,
|
| 313 |
"context": context,
|
| 314 |
"query": query,
|
| 315 |
"response": response,
|
| 316 |
+
},
|
| 317 |
+
config=config,
|
| 318 |
)
|
| 319 |
|
| 320 |
list_of_questions = new_questions.split("...")
|