Spaces:
Sleeping
Sleeping
FauziIsyrinApridal
commited on
Commit
·
4d1c64a
1
Parent(s):
847786a
clean code
Browse files- app.py +2 -61
- app/document_processor.py +45 -2
app.py
CHANGED
|
@@ -2,12 +2,9 @@ import streamlit as st
|
|
| 2 |
import os
|
| 3 |
from dotenv import load_dotenv
|
| 4 |
from langsmith import traceable
|
| 5 |
-
from datetime import datetime
|
| 6 |
-
from typing import List, Dict, Optional
|
| 7 |
-
|
| 8 |
from app.chat import initialize_session_state, display_chat_history
|
| 9 |
-
from app.data_loader import get_data,
|
| 10 |
-
from app.document_processor import process_documents, save_vector_store_to_supabase, load_vector_store_from_supabase
|
| 11 |
from app.prompts import sahabat_prompt
|
| 12 |
from app.db import supabase
|
| 13 |
from langchain_community.llms import Replicate
|
|
@@ -17,58 +14,9 @@ from langchain_community.document_transformers import LongContextReorder
|
|
| 17 |
|
| 18 |
load_dotenv()
|
| 19 |
|
| 20 |
-
# ---------------------------------------------------------
|
| 21 |
-
# ⚡️ CONFIG
|
| 22 |
-
# ---------------------------------------------------------
|
| 23 |
BUCKET_NAME = "pnp-bot-storage-archive"
|
| 24 |
VECTOR_STORE_PREFIX = "vector_store"
|
| 25 |
|
| 26 |
-
# ---------------------------------------------------------
|
| 27 |
-
# ⚡️ UTILITY
|
| 28 |
-
# ---------------------------------------------------------
|
| 29 |
-
def get_latest_data_timestamp_from_files(bucket_name: str) -> float:
|
| 30 |
-
"""Get the latest timestamp from files in a Supabase storage bucket."""
|
| 31 |
-
files = list_all_files(bucket_name)
|
| 32 |
-
latest_time = 0.0
|
| 33 |
-
for file in files:
|
| 34 |
-
iso_time = file.get("updated_at") or file.get("created_at")
|
| 35 |
-
if iso_time:
|
| 36 |
-
try:
|
| 37 |
-
timestamp = datetime.fromisoformat(iso_time.replace('Z', '+00:00')).timestamp()
|
| 38 |
-
latest_time = max(latest_time, timestamp)
|
| 39 |
-
except Exception as e:
|
| 40 |
-
print(f"Gagal parsing waktu dari {file.get('name')}: {e}")
|
| 41 |
-
return latest_time
|
| 42 |
-
|
| 43 |
-
|
| 44 |
-
def get_supabase_vector_store_timestamp() -> Optional[str]:
|
| 45 |
-
"""Get the latest timestamp of vector store files in the Supabase storage."""
|
| 46 |
-
try:
|
| 47 |
-
response = supabase.storage.from_(BUCKET_NAME).list()
|
| 48 |
-
timestamps = []
|
| 49 |
-
for file in response:
|
| 50 |
-
if file["name"].startswith(VECTOR_STORE_PREFIX) and (
|
| 51 |
-
file["name"].endswith(".faiss") or file["name"].endswith(".pkl")
|
| 52 |
-
):
|
| 53 |
-
timestamps.append(file["updated_at"])
|
| 54 |
-
if len(timestamps) >= 2:
|
| 55 |
-
return max(timestamps)
|
| 56 |
-
return None
|
| 57 |
-
except Exception as e:
|
| 58 |
-
print(f"Error getting Supabase timestamp: {e}")
|
| 59 |
-
return None
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
def vector_store_is_outdated() -> bool:
|
| 63 |
-
"""Check if vector store needs to be updated based on files in Supabase storage."""
|
| 64 |
-
supabase_timestamp = get_supabase_vector_store_timestamp()
|
| 65 |
-
if supabase_timestamp is None:
|
| 66 |
-
return True
|
| 67 |
-
supabase_time = datetime.fromisoformat(supabase_timestamp.replace("Z", "+00:00")).timestamp()
|
| 68 |
-
data_time = get_latest_data_timestamp_from_files("pnp-bot-storage")
|
| 69 |
-
|
| 70 |
-
return data_time > supabase_time
|
| 71 |
-
|
| 72 |
|
| 73 |
def reorder_embedding(docs):
|
| 74 |
"""Reorder documents for long context retrieval."""
|
|
@@ -76,9 +24,6 @@ def reorder_embedding(docs):
|
|
| 76 |
return reordering.transform_documents(docs)
|
| 77 |
|
| 78 |
|
| 79 |
-
# ---------------------------------------------------------
|
| 80 |
-
# ⚡️ RAG CHAIN
|
| 81 |
-
# ---------------------------------------------------------
|
| 82 |
@traceable(name="Create RAG Conversational Chain")
|
| 83 |
def create_conversational_chain(vector_store):
|
| 84 |
"""Create a Conversational Retrieval Chain for RAG."""
|
|
@@ -101,10 +46,6 @@ def get_rag_chain(vector_store):
|
|
| 101 |
"""Return a Conversational Retrieval Chain for external use."""
|
| 102 |
return create_conversational_chain(vector_store)
|
| 103 |
|
| 104 |
-
|
| 105 |
-
# ---------------------------------------------------------
|
| 106 |
-
# ⚡️ MAIN FUNCTION
|
| 107 |
-
# ---------------------------------------------------------
|
| 108 |
@traceable(name="Main Chatbot RAG App")
|
| 109 |
def main():
|
| 110 |
initialize_session_state()
|
|
|
|
| 2 |
import os
|
| 3 |
from dotenv import load_dotenv
|
| 4 |
from langsmith import traceable
|
|
|
|
|
|
|
|
|
|
| 5 |
from app.chat import initialize_session_state, display_chat_history
|
| 6 |
+
from app.data_loader import get_data, load_docs
|
| 7 |
+
from app.document_processor import process_documents, save_vector_store_to_supabase, load_vector_store_from_supabase, get_latest_data_timestamp_from_files,get_supabase_vector_store_timestamp, vector_store_is_outdated
|
| 8 |
from app.prompts import sahabat_prompt
|
| 9 |
from app.db import supabase
|
| 10 |
from langchain_community.llms import Replicate
|
|
|
|
| 14 |
|
| 15 |
load_dotenv()
|
| 16 |
|
|
|
|
|
|
|
|
|
|
| 17 |
BUCKET_NAME = "pnp-bot-storage-archive"
|
| 18 |
VECTOR_STORE_PREFIX = "vector_store"
|
| 19 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
|
| 21 |
def reorder_embedding(docs):
|
| 22 |
"""Reorder documents for long context retrieval."""
|
|
|
|
| 24 |
return reordering.transform_documents(docs)
|
| 25 |
|
| 26 |
|
|
|
|
|
|
|
|
|
|
| 27 |
@traceable(name="Create RAG Conversational Chain")
|
| 28 |
def create_conversational_chain(vector_store):
|
| 29 |
"""Create a Conversational Retrieval Chain for RAG."""
|
|
|
|
| 46 |
"""Return a Conversational Retrieval Chain for external use."""
|
| 47 |
return create_conversational_chain(vector_store)
|
| 48 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 49 |
@traceable(name="Main Chatbot RAG App")
|
| 50 |
def main():
|
| 51 |
initialize_session_state()
|
app/document_processor.py
CHANGED
|
@@ -4,7 +4,8 @@ from langchain_community.vectorstores import FAISS
|
|
| 4 |
import os
|
| 5 |
import tempfile
|
| 6 |
import streamlit as st
|
| 7 |
-
from
|
|
|
|
| 8 |
|
| 9 |
|
| 10 |
def save_vector_store_to_supabase(vector_store, supabase, bucket_name, file_prefix="vector_store"):
|
|
@@ -112,4 +113,46 @@ def process_documents(docs):
|
|
| 112 |
text_chunks = text_splitter.split_documents(docs)
|
| 113 |
vector_store = FAISS.from_documents(text_chunks, embeddings)
|
| 114 |
|
| 115 |
-
return vector_store
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
import os
|
| 5 |
import tempfile
|
| 6 |
import streamlit as st
|
| 7 |
+
from datetime import datetime
|
| 8 |
+
from data_loader import list_all_files
|
| 9 |
|
| 10 |
|
| 11 |
def save_vector_store_to_supabase(vector_store, supabase, bucket_name, file_prefix="vector_store"):
|
|
|
|
| 113 |
text_chunks = text_splitter.split_documents(docs)
|
| 114 |
vector_store = FAISS.from_documents(text_chunks, embeddings)
|
| 115 |
|
| 116 |
+
return vector_store
|
| 117 |
+
|
| 118 |
+
def get_latest_data_timestamp_from_files(bucket_name: str) -> float:
|
| 119 |
+
"""Get the latest timestamp from files in a Supabase storage bucket."""
|
| 120 |
+
files = list_all_files(bucket_name)
|
| 121 |
+
latest_time = 0.0
|
| 122 |
+
for file in files:
|
| 123 |
+
iso_time = file.get("updated_at") or file.get("created_at")
|
| 124 |
+
if iso_time:
|
| 125 |
+
try:
|
| 126 |
+
timestamp = datetime.fromisoformat(iso_time.replace('Z', '+00:00')).timestamp()
|
| 127 |
+
latest_time = max(latest_time, timestamp)
|
| 128 |
+
except Exception as e:
|
| 129 |
+
print(f"Gagal parsing waktu dari {file.get('name')}: {e}")
|
| 130 |
+
return latest_time
|
| 131 |
+
|
| 132 |
+
|
| 133 |
+
def get_supabase_vector_store_timestamp() -> Optional[str]:
|
| 134 |
+
"""Get the latest timestamp of vector store files in the Supabase storage."""
|
| 135 |
+
try:
|
| 136 |
+
response = supabase.storage.from_(BUCKET_NAME).list()
|
| 137 |
+
timestamps = []
|
| 138 |
+
for file in response:
|
| 139 |
+
if file["name"].startswith(VECTOR_STORE_PREFIX) and (
|
| 140 |
+
file["name"].endswith(".faiss") or file["name"].endswith(".pkl")
|
| 141 |
+
):
|
| 142 |
+
timestamps.append(file["updated_at"])
|
| 143 |
+
if len(timestamps) >= 2:
|
| 144 |
+
return max(timestamps)
|
| 145 |
+
return None
|
| 146 |
+
except Exception as e:
|
| 147 |
+
print(f"Error getting Supabase timestamp: {e}")
|
| 148 |
+
return None
|
| 149 |
+
|
| 150 |
+
def vector_store_is_outdated() -> bool:
|
| 151 |
+
"""Check if vector store needs to be updated based on files in Supabase storage."""
|
| 152 |
+
supabase_timestamp = get_supabase_vector_store_timestamp()
|
| 153 |
+
if supabase_timestamp is None:
|
| 154 |
+
return True
|
| 155 |
+
supabase_time = datetime.fromisoformat(supabase_timestamp.replace("Z", "+00:00")).timestamp()
|
| 156 |
+
data_time = get_latest_data_timestamp_from_files("pnp-bot-storage")
|
| 157 |
+
|
| 158 |
+
return data_time > supabase_time
|