Spaces:
Running
Running
Matthew Schulz
commited on
Commit
·
003becb
1
Parent(s):
91759b9
refactor: updated drive upload to use browser localstorage to store id for uploading chats
Browse files- .gitignore +3 -4
- app.py +43 -29
- utils/chatLogger.py +56 -26
.gitignore
CHANGED
@@ -1,9 +1,8 @@
|
|
1 |
venv
|
2 |
.env
|
3 |
-
|
4 |
-
|
5 |
-
utils/__pycache__
|
6 |
test.py
|
7 |
data/old_KB.py
|
8 |
creds.json
|
9 |
-
utils/__pycache__/google_drive_utils.cpython-312.pyc
|
|
|
1 |
venv
|
2 |
.env
|
3 |
+
.gradio
|
4 |
+
__pycache__
|
5 |
+
utils/__pycache__
|
6 |
test.py
|
7 |
data/old_KB.py
|
8 |
creds.json
|
|
app.py
CHANGED
@@ -1,6 +1,7 @@
|
|
1 |
import os
|
2 |
import json
|
3 |
import base64
|
|
|
4 |
import logging
|
5 |
|
6 |
from typing import Generator
|
@@ -16,7 +17,8 @@ from utils.utils import (
|
|
16 |
get_messages,
|
17 |
load_knowledge_base,
|
18 |
)
|
19 |
-
|
|
|
20 |
|
21 |
|
22 |
def initialize():
|
@@ -40,9 +42,6 @@ def initialize():
|
|
40 |
logger.info("Loading env variables...")
|
41 |
if not os.getenv("SPACE_ID"):
|
42 |
load_dotenv()
|
43 |
-
drive_creds_encoded = os.getenv(
|
44 |
-
"GOOGLE_DRIVE_SERVICE_ACCOUNT_CREDENTIALS_BASE64"
|
45 |
-
).strip()
|
46 |
|
47 |
logger.info("Initializing OpenAI client...")
|
48 |
openAI_client = OpenAI(
|
@@ -51,10 +50,13 @@ def initialize():
|
|
51 |
)
|
52 |
|
53 |
logger.info("Loading Drive service account details...")
|
|
|
|
|
|
|
54 |
service_account_json = json.loads(base64.b64decode(drive_creds_encoded).decode())
|
55 |
|
56 |
-
logger.info("Initializing
|
57 |
-
|
58 |
|
59 |
logger.info("Ready for user query...")
|
60 |
return (
|
@@ -63,7 +65,7 @@ def initialize():
|
|
63 |
knowledge_base,
|
64 |
openAI_client,
|
65 |
logger,
|
66 |
-
|
67 |
)
|
68 |
|
69 |
|
@@ -73,13 +75,14 @@ def initialize():
|
|
73 |
knowledge_base,
|
74 |
openAI_client,
|
75 |
logger,
|
76 |
-
|
77 |
) = initialize()
|
78 |
|
79 |
|
80 |
def rag_chatbot(
|
81 |
user_message: str,
|
82 |
-
chat_history: list
|
|
|
83 |
) -> Generator[list, None, None]:
|
84 |
"""
|
85 |
Retrieves relevant documents to user query and streams LLM response catching errors along the way.
|
@@ -163,33 +166,44 @@ def rag_chatbot(
|
|
163 |
|
164 |
try:
|
165 |
logger.info("Trying to upload chat history to Drive...")
|
166 |
-
|
167 |
except Exception as e:
|
168 |
-
logger.warning(f"
|
169 |
|
170 |
logger.info("Returning chat history...")
|
171 |
return chat_history
|
172 |
|
173 |
|
174 |
# Gradio app code
|
175 |
-
|
176 |
-
|
177 |
-
|
178 |
-
|
179 |
-
|
180 |
-
|
181 |
-
|
182 |
-
|
183 |
-
|
184 |
-
|
185 |
-
|
186 |
-
|
187 |
-
|
188 |
-
|
189 |
-
|
190 |
-
|
191 |
-
|
192 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
193 |
|
194 |
|
195 |
if __name__ == "__main__":
|
|
|
1 |
import os
|
2 |
import json
|
3 |
import base64
|
4 |
+
import uuid
|
5 |
import logging
|
6 |
|
7 |
from typing import Generator
|
|
|
17 |
get_messages,
|
18 |
load_knowledge_base,
|
19 |
)
|
20 |
+
|
21 |
+
from utils.chatLogger import ChatUploader
|
22 |
|
23 |
|
24 |
def initialize():
|
|
|
42 |
logger.info("Loading env variables...")
|
43 |
if not os.getenv("SPACE_ID"):
|
44 |
load_dotenv()
|
|
|
|
|
|
|
45 |
|
46 |
logger.info("Initializing OpenAI client...")
|
47 |
openAI_client = OpenAI(
|
|
|
50 |
)
|
51 |
|
52 |
logger.info("Loading Drive service account details...")
|
53 |
+
drive_creds_encoded = os.getenv(
|
54 |
+
"GOOGLE_DRIVE_SERVICE_ACCOUNT_CREDENTIALS_BASE64"
|
55 |
+
).strip()
|
56 |
service_account_json = json.loads(base64.b64decode(drive_creds_encoded).decode())
|
57 |
|
58 |
+
logger.info("Initializing ChatUploader instance...")
|
59 |
+
chat_uploader = ChatUploader(service_account_json)
|
60 |
|
61 |
logger.info("Ready for user query...")
|
62 |
return (
|
|
|
65 |
knowledge_base,
|
66 |
openAI_client,
|
67 |
logger,
|
68 |
+
chat_uploader,
|
69 |
)
|
70 |
|
71 |
|
|
|
75 |
knowledge_base,
|
76 |
openAI_client,
|
77 |
logger,
|
78 |
+
chat_uploader,
|
79 |
) = initialize()
|
80 |
|
81 |
|
82 |
def rag_chatbot(
|
83 |
user_message: str,
|
84 |
+
chat_history: list,
|
85 |
+
browser_id: str,
|
86 |
) -> Generator[list, None, None]:
|
87 |
"""
|
88 |
Retrieves relevant documents to user query and streams LLM response catching errors along the way.
|
|
|
166 |
|
167 |
try:
|
168 |
logger.info("Trying to upload chat history to Drive...")
|
169 |
+
chat_uploader.upload_chat_history(chat_history, browser_id)
|
170 |
except Exception as e:
|
171 |
+
logger.warning(f"Warning: error during Google Drive upload: {e}")
|
172 |
|
173 |
logger.info("Returning chat history...")
|
174 |
return chat_history
|
175 |
|
176 |
|
177 |
# Gradio app code
|
178 |
+
with gr.Blocks() as demo:
|
179 |
+
browser_id_state = gr.BrowserState(default_value=None)
|
180 |
+
|
181 |
+
@demo.load(inputs=browser_id_state, outputs=browser_id_state)
|
182 |
+
def load_browser_id(current_id):
|
183 |
+
if current_id is None or current_id == "":
|
184 |
+
new_id = str(uuid.uuid4())
|
185 |
+
return new_id
|
186 |
+
return current_id
|
187 |
+
|
188 |
+
gr.ChatInterface(
|
189 |
+
fn=rag_chatbot,
|
190 |
+
title="Matthew Schulz's RAG Chatbot 💬🤖",
|
191 |
+
additional_inputs=browser_id_state,
|
192 |
+
type="messages",
|
193 |
+
examples=[
|
194 |
+
["What is Matthew's educational background?"],
|
195 |
+
["What machine learning projects has Matthew worked on?"],
|
196 |
+
["What experience does Matthew have in software engineering?"],
|
197 |
+
["Why did Matthew choose to pursue a degree in computer science?"],
|
198 |
+
["Does Matthew have any leadership experience?"],
|
199 |
+
["Has Matthew completed any Summer internships?"],
|
200 |
+
["Tell me about some real-world projects Matthew has worked on."],
|
201 |
+
["What is Matthew's greatest strength and weakness?"],
|
202 |
+
],
|
203 |
+
save_history=True,
|
204 |
+
run_examples_on_click=False,
|
205 |
+
cache_examples=False,
|
206 |
+
)
|
207 |
|
208 |
|
209 |
if __name__ == "__main__":
|
utils/chatLogger.py
CHANGED
@@ -1,64 +1,94 @@
|
|
1 |
import io
|
2 |
import json
|
3 |
-
import uuid
|
4 |
from googleapiclient.discovery import build
|
5 |
-
from googleapiclient.http import MediaIoBaseUpload
|
6 |
from google.oauth2 import service_account
|
7 |
|
8 |
|
9 |
-
class
|
10 |
def __init__(
|
11 |
self,
|
12 |
service_account_json: dict,
|
13 |
root_folder_id: str = "1KtfVgL1Rg1iX-ZMHH4Im__ss-pgbaDM9",
|
14 |
):
|
15 |
"""
|
16 |
-
Initializes a new chat
|
17 |
"""
|
18 |
credentials = service_account.Credentials.from_service_account_info(
|
19 |
service_account_json, scopes=["https://www.googleapis.com/auth/drive"]
|
20 |
)
|
21 |
self.drive_service = build("drive", "v3", credentials=credentials)
|
22 |
self.root_folder_id = root_folder_id
|
23 |
-
self.session_id = str(uuid.uuid4())
|
24 |
-
self.folder_id = self._create_session_folder()
|
25 |
|
26 |
-
def
|
27 |
"""
|
28 |
-
|
29 |
"""
|
30 |
-
|
31 |
-
|
32 |
-
"
|
33 |
-
"
|
34 |
-
|
35 |
-
|
36 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
37 |
|
38 |
def upload_chat_history(
|
39 |
-
self, chat_history: list, filename: str = "chat_log.json"
|
40 |
) -> None:
|
41 |
"""
|
42 |
-
Uploads
|
43 |
"""
|
|
|
|
|
44 |
query = (
|
45 |
-
f"name = '{filename}' and '{
|
46 |
-
|
47 |
)
|
48 |
results = self.drive_service.files().list(q=query, fields="files(id)").execute()
|
49 |
files = results.get("files", [])
|
50 |
|
51 |
-
content = json.dumps(chat_history, indent=2)
|
52 |
-
|
53 |
-
media = MediaIoBaseUpload(
|
54 |
-
io.BytesIO(content.encode()), mimetype="application/json"
|
55 |
-
)
|
56 |
-
|
57 |
if files:
|
58 |
file_id = files[0]["id"]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
59 |
self.drive_service.files().update(
|
60 |
fileId=file_id, media_body=media
|
61 |
).execute()
|
62 |
else:
|
63 |
-
|
|
|
|
|
|
|
|
|
64 |
self.drive_service.files().create(body=metadata, media_body=media).execute()
|
|
|
1 |
import io
|
2 |
import json
|
|
|
3 |
from googleapiclient.discovery import build
|
4 |
+
from googleapiclient.http import MediaIoBaseUpload, MediaIoBaseDownload
|
5 |
from google.oauth2 import service_account
|
6 |
|
7 |
|
8 |
+
class ChatUploader:
|
9 |
def __init__(
|
10 |
self,
|
11 |
service_account_json: dict,
|
12 |
root_folder_id: str = "1KtfVgL1Rg1iX-ZMHH4Im__ss-pgbaDM9",
|
13 |
):
|
14 |
"""
|
15 |
+
Initializes a new chat uploader instance using a service account JSON dict.
|
16 |
"""
|
17 |
credentials = service_account.Credentials.from_service_account_info(
|
18 |
service_account_json, scopes=["https://www.googleapis.com/auth/drive"]
|
19 |
)
|
20 |
self.drive_service = build("drive", "v3", credentials=credentials)
|
21 |
self.root_folder_id = root_folder_id
|
|
|
|
|
22 |
|
23 |
+
def _get_or_create_browser_folder(self, browser_id: str) -> str:
|
24 |
"""
|
25 |
+
Searches for an existing folder for the given browser_id. If not found, creates a folder named 'browser_{browser_id}' and returns its ID.
|
26 |
"""
|
27 |
+
folder_name = f"browser_{browser_id}"
|
28 |
+
query = (
|
29 |
+
f"name = '{folder_name}' and '{self.root_folder_id}' in parents and "
|
30 |
+
"mimeType = 'application/vnd.google-apps.folder' and trashed = false"
|
31 |
+
)
|
32 |
+
results = self.drive_service.files().list(q=query, fields="files(id)").execute()
|
33 |
+
folders = results.get("files", [])
|
34 |
+
|
35 |
+
if folders:
|
36 |
+
return folders[0]["id"]
|
37 |
+
else:
|
38 |
+
metadata = {
|
39 |
+
"name": folder_name,
|
40 |
+
"mimeType": "application/vnd.google-apps.folder",
|
41 |
+
"parents": [self.root_folder_id],
|
42 |
+
}
|
43 |
+
folder = (
|
44 |
+
self.drive_service.files().create(body=metadata, fields="id").execute()
|
45 |
+
)
|
46 |
+
return folder["id"]
|
47 |
|
48 |
def upload_chat_history(
|
49 |
+
self, chat_history: list, browser_id: str, filename: str = "chat_log.json"
|
50 |
) -> None:
|
51 |
"""
|
52 |
+
Uploads the chat log file inside the browser's folder. If the folder and/or file exists, it appends the new chat entries to the current log. Otherwise, it creates them.
|
53 |
"""
|
54 |
+
folder_id = self._get_or_create_browser_folder(browser_id)
|
55 |
+
|
56 |
query = (
|
57 |
+
f"name = '{filename}' and '{folder_id}' in parents and "
|
58 |
+
"mimeType = 'application/json' and trashed = false"
|
59 |
)
|
60 |
results = self.drive_service.files().list(q=query, fields="files(id)").execute()
|
61 |
files = results.get("files", [])
|
62 |
|
|
|
|
|
|
|
|
|
|
|
|
|
63 |
if files:
|
64 |
file_id = files[0]["id"]
|
65 |
+
|
66 |
+
request = self.drive_service.files().get_media(fileId=file_id)
|
67 |
+
existing_stream = io.BytesIO()
|
68 |
+
downloader = MediaIoBaseDownload(existing_stream, request)
|
69 |
+
done = False
|
70 |
+
while not done:
|
71 |
+
_, done = downloader.next_chunk()
|
72 |
+
|
73 |
+
existing_stream.seek(0)
|
74 |
+
try:
|
75 |
+
existing_chat_history = json.loads(existing_stream.read())
|
76 |
+
except json.JSONDecodeError:
|
77 |
+
existing_chat_history = []
|
78 |
+
|
79 |
+
updated_chat_history = existing_chat_history + chat_history
|
80 |
+
|
81 |
+
content = json.dumps(updated_chat_history, indent=2)
|
82 |
+
media = MediaIoBaseUpload(
|
83 |
+
io.BytesIO(content.encode()), mimetype="application/json"
|
84 |
+
)
|
85 |
self.drive_service.files().update(
|
86 |
fileId=file_id, media_body=media
|
87 |
).execute()
|
88 |
else:
|
89 |
+
content = json.dumps(chat_history, indent=2)
|
90 |
+
media = MediaIoBaseUpload(
|
91 |
+
io.BytesIO(content.encode()), mimetype="application/json"
|
92 |
+
)
|
93 |
+
metadata = {"name": filename, "parents": [folder_id]}
|
94 |
self.drive_service.files().create(body=metadata, media_body=media).execute()
|