Spaces:
Runtime error
Runtime error
Upload GPT-4_PDF_summary.py
Browse files- GPT-4_PDF_summary.py +18 -19
GPT-4_PDF_summary.py
CHANGED
@@ -1,16 +1,11 @@
|
|
1 |
#!/usr/bin/env python
|
2 |
# coding: utf-8
|
3 |
|
|
|
4 |
# In[ ]:
|
5 |
|
6 |
|
7 |
-
|
8 |
-
|
9 |
-
|
10 |
-
# In[ ]:
|
11 |
-
|
12 |
-
|
13 |
-
import os
|
14 |
from langchain.chains import RetrievalQA
|
15 |
from langchain.llms import OpenAI
|
16 |
from langchain.document_loaders import TextLoader
|
@@ -50,7 +45,7 @@ select_k = pn.widgets.IntSlider(
|
|
50 |
name="Number of relevant chunks", start=1, end=5, step=1, value=2
|
51 |
)
|
52 |
select_chain_type = pn.widgets.RadioButtonGroup(
|
53 |
-
name='Chain type',
|
54 |
options=['stuff', 'map_reduce', "refine", "map_rerank"]
|
55 |
)
|
56 |
|
@@ -79,8 +74,9 @@ def qa(file, query, chain_type, k):
|
|
79 |
# create the vectorestore to use as the index
|
80 |
db = Chroma.from_documents(texts, embeddings)
|
81 |
# expose this index in a retriever interface
|
82 |
-
retriever = db.as_retriever(
|
83 |
-
|
|
|
84 |
qa = RetrievalQA.from_chain_type(
|
85 |
llm=OpenAI(), chain_type=chain_type, retriever=retriever, return_source_documents=True)
|
86 |
result = qa({"query": query})
|
@@ -93,16 +89,18 @@ def qa(file, query, chain_type, k):
|
|
93 |
|
94 |
convos = [] # store all panel objects in a list
|
95 |
|
|
|
96 |
def qa_result(_):
|
97 |
os.environ["OPENAI_API_KEY"] = openaikey.value
|
98 |
-
|
99 |
-
# save pdf file to a temp file
|
100 |
if file_input.value is not None:
|
101 |
file_input.save("/.cache/temp.pdf")
|
102 |
-
|
103 |
prompt_text = prompt.value
|
104 |
if prompt_text:
|
105 |
-
result = qa(file="/.cache/temp.pdf", query=prompt_text,
|
|
|
106 |
convos.extend([
|
107 |
pn.Row(
|
108 |
pn.panel("\U0001F60A", width=10),
|
@@ -114,11 +112,12 @@ def qa_result(_):
|
|
114 |
pn.Column(
|
115 |
result["result"],
|
116 |
"Relevant source text:",
|
117 |
-
pn.pane.Markdown('\n--------------------------------------------------------------------\n'.join(
|
|
|
118 |
)
|
119 |
)
|
120 |
])
|
121 |
-
#return convos
|
122 |
return pn.Column(*convos, margin=15, width=575, min_height=400)
|
123 |
|
124 |
|
@@ -134,7 +133,8 @@ qa_interactive = pn.panel(
|
|
134 |
# In[8]:
|
135 |
|
136 |
|
137 |
-
output = pn.WidgetBox('*Output will show up here:*',
|
|
|
138 |
|
139 |
|
140 |
# In[9]:
|
@@ -148,9 +148,8 @@ pn.Column(
|
|
148 |
1) Upload a PDF. 2) Enter OpenAI API key. This costs $. Set up billing at [OpenAI](https://platform.openai.com/account). 3) Type a question and click "Run".
|
149 |
|
150 |
"""),
|
151 |
-
pn.Row(file_input,openaikey),
|
152 |
output,
|
153 |
widgets
|
154 |
|
155 |
).servable()
|
156 |
-
|
|
|
1 |
#!/usr/bin/env python
|
2 |
# coding: utf-8
|
3 |
|
4 |
+
# !pip install langchain openai chromadb tiktoken pypdf panel
|
5 |
# In[ ]:
|
6 |
|
7 |
|
8 |
+
import os
|
|
|
|
|
|
|
|
|
|
|
|
|
9 |
from langchain.chains import RetrievalQA
|
10 |
from langchain.llms import OpenAI
|
11 |
from langchain.document_loaders import TextLoader
|
|
|
45 |
name="Number of relevant chunks", start=1, end=5, step=1, value=2
|
46 |
)
|
47 |
select_chain_type = pn.widgets.RadioButtonGroup(
|
48 |
+
name='Chain type',
|
49 |
options=['stuff', 'map_reduce', "refine", "map_rerank"]
|
50 |
)
|
51 |
|
|
|
74 |
# create the vectorestore to use as the index
|
75 |
db = Chroma.from_documents(texts, embeddings)
|
76 |
# expose this index in a retriever interface
|
77 |
+
retriever = db.as_retriever(
|
78 |
+
search_type="similarity", search_kwargs={"k": k})
|
79 |
+
# create a chain to answer questions
|
80 |
qa = RetrievalQA.from_chain_type(
|
81 |
llm=OpenAI(), chain_type=chain_type, retriever=retriever, return_source_documents=True)
|
82 |
result = qa({"query": query})
|
|
|
89 |
|
90 |
convos = [] # store all panel objects in a list
|
91 |
|
92 |
+
|
93 |
def qa_result(_):
|
94 |
os.environ["OPENAI_API_KEY"] = openaikey.value
|
95 |
+
|
96 |
+
# save pdf file to a temp file
|
97 |
if file_input.value is not None:
|
98 |
file_input.save("/.cache/temp.pdf")
|
99 |
+
|
100 |
prompt_text = prompt.value
|
101 |
if prompt_text:
|
102 |
+
result = qa(file="/.cache/temp.pdf", query=prompt_text,
|
103 |
+
chain_type=select_chain_type.value, k=select_k.value)
|
104 |
convos.extend([
|
105 |
pn.Row(
|
106 |
pn.panel("\U0001F60A", width=10),
|
|
|
112 |
pn.Column(
|
113 |
result["result"],
|
114 |
"Relevant source text:",
|
115 |
+
pn.pane.Markdown('\n--------------------------------------------------------------------\n'.join(
|
116 |
+
doc.page_content for doc in result["source_documents"]))
|
117 |
)
|
118 |
)
|
119 |
])
|
120 |
+
# return convos
|
121 |
return pn.Column(*convos, margin=15, width=575, min_height=400)
|
122 |
|
123 |
|
|
|
133 |
# In[8]:
|
134 |
|
135 |
|
136 |
+
output = pn.WidgetBox('*Output will show up here:*',
|
137 |
+
qa_interactive, width=630, scroll=True)
|
138 |
|
139 |
|
140 |
# In[9]:
|
|
|
148 |
1) Upload a PDF. 2) Enter OpenAI API key. This costs $. Set up billing at [OpenAI](https://platform.openai.com/account). 3) Type a question and click "Run".
|
149 |
|
150 |
"""),
|
151 |
+
pn.Row(file_input, openaikey),
|
152 |
output,
|
153 |
widgets
|
154 |
|
155 |
).servable()
|
|