Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -21,10 +21,69 @@ llm = HuggingFaceHub(repo_id=model ,
|
|
21 |
"temperature":0.2})
|
22 |
langchain.llm_cache = InMemoryCache()
|
23 |
|
24 |
-
def
|
25 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
26 |
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
"temperature":0.2})
|
22 |
langchain.llm_cache = InMemoryCache()
|
23 |
|
24 |
+
def build_context(files,urls):
|
25 |
+
documents = []
|
26 |
+
if files is not None:
|
27 |
+
for idx, file in enumerate(files):
|
28 |
+
if file.name.endswith('.pdf'):
|
29 |
+
pdf_path = file.name
|
30 |
+
loader = PyPDFLoader(pdf_path)
|
31 |
+
documents.extend(loader.load())
|
32 |
+
elif file.name.endswith('.docx'):
|
33 |
+
doc_path = file.name
|
34 |
+
loader = Docx2txtLoader(doc_path)
|
35 |
+
documents.extend(loader.load())
|
36 |
+
elif file.name.endswith('.ppt') or file.name.endswith('.pptx'):
|
37 |
+
ppt_path = file.name
|
38 |
+
loader = UnstructuredPowerPointLoader(ppt_path)
|
39 |
+
documents.extend(loader.load())
|
40 |
+
if (urls is not None) | (urls != ""):
|
41 |
+
list_urls = urls.split(sep=",")
|
42 |
+
for url in list_urls:
|
43 |
+
loader = OnlinePDFLoader(url)
|
44 |
+
documents.extend(loader.load())
|
45 |
+
text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
|
46 |
+
chunked_documents = text_splitter.split_documents(documents)
|
47 |
+
vectordb = Chroma.from_documents(
|
48 |
+
documents=chunked_documents,
|
49 |
+
embedding=embeddings
|
50 |
+
)
|
51 |
+
global qa_chain
|
52 |
+
qa_chain = RetrievalQA.from_chain_type(
|
53 |
+
llm=llm,
|
54 |
+
retriever=vectordb.as_retriever(search_kwargs={'k': 7}),
|
55 |
+
chain_type="stuff",
|
56 |
+
#return_source_documents=True
|
57 |
+
)
|
58 |
+
return "ready"
|
59 |
|
60 |
+
def loading():
|
61 |
+
return "Loading..."
|
62 |
+
|
63 |
+
def respond(message, chat_history):
|
64 |
+
result = qa_chain({"query": message})["result"]
|
65 |
+
chat_history.append((message, result))
|
66 |
+
time.sleep(2)
|
67 |
+
return "", chat_history
|
68 |
+
|
69 |
+
with gr.Blocks() as demo:
|
70 |
+
with gr.Row():
|
71 |
+
with gr.Column():
|
72 |
+
pdf_docs = gr.Files(label="Load pdf files", file_types=['.pdf'], type="file")
|
73 |
+
with gr.Column():
|
74 |
+
urls = gr.Textbox(label="Enter one of multiple online pdf urls (comma separated if multiple)", value=None)
|
75 |
+
with gr.Row():
|
76 |
+
load_context = gr.Button("Load documents and urls")
|
77 |
+
with gr.Row():
|
78 |
+
loading_status = gr.Textbox(label="Status", placeholder="", interactive=False)
|
79 |
+
with gr.Row():
|
80 |
+
with gr.Column():
|
81 |
+
hg_chatbot = gr.Chatbot()
|
82 |
+
msg = gr.Textbox(label="User message")
|
83 |
+
clear = gr.ClearButton([msg, hg_chatbot])
|
84 |
+
load_context.click(loading, None, loading_status, queue=False)
|
85 |
+
load_context.click(build_context, inputs=[pdf_docs, urls], outputs=[loading_status], queue=False)
|
86 |
+
msg.submit(respond, [msg, hg_chatbot], [msg, hg_chatbot])
|
87 |
+
|
88 |
+
demo.queue(concurrency_count=3)
|
89 |
+
demo.launch(debug=True)
|