Spaces:
Paused
Paused
Daniel Marques
commited on
Commit
•
3c4d8f2
1
Parent(s):
42d6b9c
fix: add tokenWs
Browse files
main.py
CHANGED
@@ -14,7 +14,7 @@ from langchain.chains import RetrievalQA
|
|
14 |
from langchain.embeddings import HuggingFaceInstructEmbeddings
|
15 |
from langchain.prompts import PromptTemplate
|
16 |
from langchain.memory import ConversationBufferMemory
|
17 |
-
from langchain.callbacks.base import BaseCallbackHandler
|
18 |
from langchain.schema import LLMResult
|
19 |
|
20 |
# from langchain.embeddings import HuggingFaceEmbeddings
|
@@ -35,9 +35,9 @@ tokenWS = ''
|
|
35 |
|
36 |
class MyCustomHandler(BaseCallbackHandler):
|
37 |
def on_llm_new_token(self, token: str, **kwargs) -> None:
|
38 |
-
global
|
39 |
|
40 |
-
|
41 |
|
42 |
print(f" token: {tokenWS}")
|
43 |
|
@@ -73,11 +73,14 @@ RETRIEVER = DB.as_retriever()
|
|
73 |
|
74 |
LLM = load_model(device_type=DEVICE_TYPE, model_id=MODEL_ID, model_basename=MODEL_BASENAME, stream=True, callbacks=[MyCustomHandler()])
|
75 |
|
76 |
-
template = """you are a helpful, respectful and honest assistant.
|
77 |
-
You should only
|
78 |
-
|
79 |
-
|
80 |
-
Always
|
|
|
|
|
|
|
81 |
Context: {history} \n {context}
|
82 |
Question: {question}
|
83 |
"""
|
@@ -245,11 +248,15 @@ async def create_upload_file(file: UploadFile):
|
|
245 |
@api_app.websocket("/ws")
|
246 |
async def websocket_endpoint(websocket: WebSocket):
|
247 |
global QA
|
|
|
248 |
|
249 |
await websocket.accept()
|
250 |
while True:
|
|
|
|
|
|
|
251 |
data = await websocket.receive_text()
|
252 |
|
253 |
-
|
254 |
|
255 |
-
await websocket.send_text(f"Message text was:
|
|
|
14 |
from langchain.embeddings import HuggingFaceInstructEmbeddings
|
15 |
from langchain.prompts import PromptTemplate
|
16 |
from langchain.memory import ConversationBufferMemory
|
17 |
+
from langchain.callbacks.base import BaseCallbackHandler, AsyncCallbackHandler
|
18 |
from langchain.schema import LLMResult
|
19 |
|
20 |
# from langchain.embeddings import HuggingFaceEmbeddings
|
|
|
35 |
|
36 |
class MyCustomHandler(BaseCallbackHandler):
|
37 |
def on_llm_new_token(self, token: str, **kwargs) -> None:
|
38 |
+
global websocketClient
|
39 |
|
40 |
+
websocketClient.send_text(f"Message text was: {tokenWS}")
|
41 |
|
42 |
print(f" token: {tokenWS}")
|
43 |
|
|
|
73 |
|
74 |
LLM = load_model(device_type=DEVICE_TYPE, model_id=MODEL_ID, model_basename=MODEL_BASENAME, stream=True, callbacks=[MyCustomHandler()])
|
75 |
|
76 |
+
template = """you are a helpful, respectful and honest assistant.
|
77 |
+
Your name is Katara llma. You should only use the source documents provided to answer the questions.
|
78 |
+
You should only respond only topics that contains in documents use to training.
|
79 |
+
Use the following pieces of context to answer the question at the end.
|
80 |
+
Always answer in the most helpful and safe way possible.
|
81 |
+
If you don't know the answer to a question, just say that you don't know, don't try to make up an answer, don't share false information.
|
82 |
+
Use 15 sentences maximum. Keep the answer as concise as possible.
|
83 |
+
Always say "thanks for asking!" at the end of the answer.
|
84 |
Context: {history} \n {context}
|
85 |
Question: {question}
|
86 |
"""
|
|
|
248 |
@api_app.websocket("/ws")
|
249 |
async def websocket_endpoint(websocket: WebSocket):
|
250 |
global QA
|
251 |
+
global websocketClient
|
252 |
|
253 |
await websocket.accept()
|
254 |
while True:
|
255 |
+
global websocketClient
|
256 |
+
websocketClient = websocket;
|
257 |
+
|
258 |
data = await websocket.receive_text()
|
259 |
|
260 |
+
QA(data)
|
261 |
|
262 |
+
await websocket.send_text(f"Message text was:")
|