Spaces:
Running
Running
Keane Moraes
commited on
Commit
•
2abcb58
1
Parent(s):
2b58524
changes to q&a and mindmap varaibles
Browse files
app.py
CHANGED
@@ -42,42 +42,8 @@ takeaways = []
|
|
42 |
folder_name = "./tests"
|
43 |
input_accepted = False
|
44 |
is_completed_analysis = False
|
45 |
-
|
46 |
-
|
47 |
-
messages=[
|
48 |
-
{"role": "system", "content": "You are a helpful AI Tutor. Who anwers brief questions about AI."},
|
49 |
-
{"role": "user", "content": "I want to learn AI"},
|
50 |
-
{"role": "assistant", "content": "Thats awesome, what do you want to know aboout AI"}
|
51 |
-
]
|
52 |
-
return messages
|
53 |
-
|
54 |
-
nodes = []
|
55 |
-
edges = []
|
56 |
-
|
57 |
-
nodes.append( Node(id="Spiderman",
|
58 |
-
label="Peter Parker",
|
59 |
-
size=25,
|
60 |
-
shape="circularImage",
|
61 |
-
image="http://marvel-force-chart.surge.sh/marvel_force_chart_img/top_spiderman.png")
|
62 |
-
) # includes **kwargs
|
63 |
-
nodes.append( Node(id="Captain_Marvel",
|
64 |
-
size=25,
|
65 |
-
shape="circularImage",
|
66 |
-
image="http://marvel-force-chart.surge.sh/marvel_force_chart_img/top_captainmarvel.png")
|
67 |
-
)
|
68 |
-
edges.append( Edge(source="Captain_Marvel",
|
69 |
-
label="friend_of",
|
70 |
-
target="Spiderman",
|
71 |
-
)
|
72 |
-
)
|
73 |
-
|
74 |
-
config = Config(width=750,
|
75 |
-
height=950,
|
76 |
-
directed=True,
|
77 |
-
physics=True,
|
78 |
-
hierarchical=False,
|
79 |
-
)
|
80 |
-
|
81 |
|
82 |
user_secret = os.getenv("OPENAI_API_KEY")
|
83 |
|
@@ -151,10 +117,6 @@ with st.sidebar:
|
|
151 |
else:
|
152 |
st.error("Please type in your youtube link or upload the PDF")
|
153 |
st.experimental_rerun()
|
154 |
-
|
155 |
-
# Save the transcript information
|
156 |
-
with open(f"{folder_name}/data_transcription.json", "w") as f:
|
157 |
-
json.dump(data_transcription, f, indent=4)
|
158 |
|
159 |
# Generate embeddings
|
160 |
if not os.path.exists(f"{folder_name}/word_embeddings.csv"):
|
@@ -227,11 +189,11 @@ tab1, tab2, tab3, tab4, tab5, tab6 = st.tabs(["Introduction", "Summary", "Transc
|
|
227 |
|
228 |
# =========== INTRODUCTION ===========
|
229 |
with tab1:
|
230 |
-
st.subheader("Introduction")
|
231 |
st.markdown("## How do I use this?")
|
232 |
st.markdown("Do one of the following")
|
233 |
st.markdown('* Type in your youtube URL that you want worked on')
|
234 |
st.markdown('* Place the PDF file that you want worked on')
|
|
|
235 |
st.markdown("**Once the file / url has finished saving, a 'Start Analysis' button will appear. Click on this button to begin the note generation**")
|
236 |
st.warning("NOTE: This is just a demo product in alpha testing. Any and all bugs will soon be fixed")
|
237 |
st.warning("After the note taking is done, you will see multiple tabs for more information")
|
@@ -278,88 +240,89 @@ with tab5:
|
|
278 |
st.warning("Please wait for the analysis to finish")
|
279 |
|
280 |
# =========== QUERY BOT ===========
|
281 |
-
with tab6:
|
282 |
-
if 'generated' not in st.session_state:
|
283 |
-
st.session_state['generated'] = []
|
284 |
-
|
285 |
-
if 'past' not in st.session_state:
|
286 |
-
st.session_state['past'] = []
|
287 |
-
|
288 |
-
def get_text():
|
289 |
-
st.header("Ask me something about the video:")
|
290 |
-
input_text = st.text_input("You: ", key="prompt")
|
291 |
-
return input_text
|
292 |
-
|
293 |
-
|
294 |
-
def get_embedding_text(prompt):
|
295 |
-
response = openai.Embedding.create(
|
296 |
-
input= prompt.strip(),
|
297 |
-
model="text-embedding-ada-002"
|
298 |
-
)
|
299 |
-
q_embedding = response['data'][0]['embedding']
|
300 |
-
print("the folder name at got here 1.5 is ", folder_name)
|
301 |
-
df = pd.read_csv(f'{folder_name}/word_embeddings.csv', index_col=0)
|
302 |
-
df['embedding'] = df['embedding'].apply(eval).apply(np.array)
|
303 |
-
|
304 |
-
df['distances'] = distances_from_embeddings(q_embedding, df['embedding'].values, distance_metric='cosine')
|
305 |
-
returns = []
|
306 |
-
|
307 |
-
# Sort by distance with 2 hints
|
308 |
-
for i, row in df.sort_values('distances', ascending=True).head(4).iterrows():
|
309 |
-
# Else add it to the text that is being returned
|
310 |
-
returns.append(row["text"])
|
311 |
-
|
312 |
-
# Return the context
|
313 |
-
return "\n\n###\n\n".join(returns)
|
314 |
-
|
315 |
-
def generate_response(prompt):
|
316 |
-
one_shot_prompt = '''
|
317 |
-
I am YoutubeGPT, a highly intelligent question answering bot.
|
318 |
-
If you ask me a question that is rooted in truth, I will give you the answer.
|
319 |
-
Q: What is human life expectancy in the United States?
|
320 |
-
A: Human life expectancy in the United States is 78 years.
|
321 |
-
Q: '''+prompt+'''
|
322 |
-
A:
|
323 |
-
'''
|
324 |
-
completions = openai.Completion.create(
|
325 |
-
engine = "text-davinci-003",
|
326 |
-
prompt = one_shot_prompt,
|
327 |
-
max_tokens = 1024,
|
328 |
-
n = 1,
|
329 |
-
stop=["Q:"],
|
330 |
-
temperature=0.5,
|
331 |
-
)
|
332 |
-
message = completions.choices[0].text
|
333 |
-
return message
|
334 |
-
|
335 |
if is_completed_analysis:
|
336 |
-
|
337 |
-
|
338 |
-
|
339 |
-
|
340 |
-
|
341 |
-
|
342 |
-
|
343 |
-
|
344 |
-
|
345 |
-
|
346 |
-
|
347 |
-
|
348 |
-
|
349 |
-
|
350 |
-
|
351 |
-
|
352 |
-
|
353 |
-
|
354 |
-
|
355 |
-
|
356 |
-
|
357 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
358 |
|
359 |
-
|
360 |
-
|
361 |
-
|
362 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
363 |
|
364 |
|
365 |
# st.header("What else")
|
|
|
42 |
folder_name = "./tests"
|
43 |
input_accepted = False
|
44 |
is_completed_analysis = False
|
45 |
+
if not os.path.exists(folder_name):
|
46 |
+
os.mkdir(folder_name)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
47 |
|
48 |
user_secret = os.getenv("OPENAI_API_KEY")
|
49 |
|
|
|
117 |
else:
|
118 |
st.error("Please type in your youtube link or upload the PDF")
|
119 |
st.experimental_rerun()
|
|
|
|
|
|
|
|
|
120 |
|
121 |
# Generate embeddings
|
122 |
if not os.path.exists(f"{folder_name}/word_embeddings.csv"):
|
|
|
189 |
|
190 |
# =========== INTRODUCTION ===========
|
191 |
with tab1:
|
|
|
192 |
st.markdown("## How do I use this?")
|
193 |
st.markdown("Do one of the following")
|
194 |
st.markdown('* Type in your youtube URL that you want worked on')
|
195 |
st.markdown('* Place the PDF file that you want worked on')
|
196 |
+
st.markdown('* Place the audio file that you want worked on')
|
197 |
st.markdown("**Once the file / url has finished saving, a 'Start Analysis' button will appear. Click on this button to begin the note generation**")
|
198 |
st.warning("NOTE: This is just a demo product in alpha testing. Any and all bugs will soon be fixed")
|
199 |
st.warning("After the note taking is done, you will see multiple tabs for more information")
|
|
|
240 |
st.warning("Please wait for the analysis to finish")
|
241 |
|
242 |
# =========== QUERY BOT ===========
|
243 |
+
with tab6:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
244 |
if is_completed_analysis:
|
245 |
+
if 'generated' not in st.session_state:
|
246 |
+
st.session_state['generated'] = []
|
247 |
+
|
248 |
+
if 'past' not in st.session_state:
|
249 |
+
st.session_state['past'] = []
|
250 |
+
|
251 |
+
def get_text():
|
252 |
+
st.header("Ask me something about the video:")
|
253 |
+
input_text = st.text_input("You: ", key="prompt")
|
254 |
+
return input_text
|
255 |
+
|
256 |
+
|
257 |
+
def get_embedding_text(prompt):
|
258 |
+
response = openai.Embedding.create(
|
259 |
+
input= prompt.strip(),
|
260 |
+
model="text-embedding-ada-002"
|
261 |
+
)
|
262 |
+
q_embedding = response['data'][0]['embedding']
|
263 |
+
print("the folder name at got here 1.5 is ", folder_name)
|
264 |
+
# df = pd.read_csv(f'{folder_name}/word_embeddings.csv', index_col=0)
|
265 |
+
data['embedding'] = data['embedding'].apply(eval).apply(np.array)
|
266 |
+
|
267 |
+
data['distances'] = distances_from_embeddings(q_embedding, data['embedding'].values, distance_metric='cosine')
|
268 |
+
returns = []
|
269 |
+
|
270 |
+
# Sort by distance with 2 hints
|
271 |
+
for i, row in data.sort_values('distances', ascending=True).head(4).iterrows():
|
272 |
+
# Else add it to the text that is being returned
|
273 |
+
returns.append(row["text"])
|
274 |
+
|
275 |
+
# Return the context
|
276 |
+
return "\n\n###\n\n".join(returns)
|
277 |
+
|
278 |
+
def generate_response(prompt):
|
279 |
+
one_shot_prompt = '''
|
280 |
+
I am YoutubeGPT, a highly intelligent question answering bot.
|
281 |
+
If you ask me a question that is rooted in truth, I will give you the answer.
|
282 |
+
Q: What is human life expectancy in the United States?
|
283 |
+
A: Human life expectancy in the United States is 78 years.
|
284 |
+
Q: '''+prompt+'''
|
285 |
+
A:
|
286 |
+
'''
|
287 |
+
completions = openai.Completion.create(
|
288 |
+
engine = "text-davinci-003",
|
289 |
+
prompt = one_shot_prompt,
|
290 |
+
max_tokens = 1024,
|
291 |
+
n = 1,
|
292 |
+
stop=["Q:"],
|
293 |
+
temperature=0.5,
|
294 |
+
)
|
295 |
+
message = completions.choices[0].text
|
296 |
+
return message
|
297 |
+
|
298 |
+
if is_completed_analysis:
|
299 |
+
user_input = get_text()
|
300 |
+
print("user input is ", user_input)
|
301 |
+
print("the folder name at got here 0.5 is ", folder_name)
|
302 |
+
else:
|
303 |
+
user_input = None
|
304 |
|
305 |
+
if 'messages' not in st.session_state:
|
306 |
+
st.session_state['messages'] = get_initial_message()
|
307 |
+
|
308 |
+
if user_input:
|
309 |
+
print("got here 1")
|
310 |
+
print("the folder name at got here 1.5 is ", folder_name)
|
311 |
+
text_embedding = get_embedding_text(user_input)
|
312 |
+
print("the folder name at got here 1.5 is ", folder_name)
|
313 |
+
print("got here 2")
|
314 |
+
title = data_transcription['title']
|
315 |
+
string_title = "\n\n###\n\n".join(title)
|
316 |
+
user_input_embedding = 'Using this context: "'+string_title+'. '+text_embedding+'", answer the following question. \n'+user_input
|
317 |
+
print("got here 3")
|
318 |
+
output = generate_response(user_input_embedding)
|
319 |
+
st.session_state.past.append(user_input)
|
320 |
+
st.session_state.generated.append(output)
|
321 |
+
|
322 |
+
if st.session_state['generated']:
|
323 |
+
for i in range(len(st.session_state['generated'])-1, -1, -1):
|
324 |
+
message(st.session_state["generated"][i], key=str(i))
|
325 |
+
message(st.session_state['past'][i], is_user=True, key=str(i) + '_user')
|
326 |
|
327 |
|
328 |
# st.header("What else")
|