make chatbot
Browse files
app.py
CHANGED
|
@@ -1,44 +1,43 @@
|
|
| 1 |
import os
|
| 2 |
from dotenv import load_dotenv
|
| 3 |
-
import gradio as gr
|
|
|
|
| 4 |
|
| 5 |
-
# βββ 1. Load environment variables βββββββββββββββββββββββββββββββββββββββββ
|
| 6 |
load_dotenv()
|
| 7 |
COHERE_API_KEY = os.getenv("COHERE_API_KEY")
|
| 8 |
GEMINI_API_KEY = os.getenv("GEMINI_API_KEY")
|
| 9 |
|
| 10 |
if not COHERE_API_KEY or not GEMINI_API_KEY:
|
| 11 |
-
raise ValueError("
|
|
|
|
| 12 |
|
| 13 |
-
# βββ 2. Initialize vector store and embedder clients βββββββββββββββββββββββ
|
| 14 |
import cohere
|
| 15 |
import chromadb
|
| 16 |
from google import genai
|
| 17 |
from google.genai import types
|
| 18 |
|
| 19 |
-
|
| 20 |
co = cohere.Client(COHERE_API_KEY)
|
| 21 |
|
| 22 |
-
|
| 23 |
genai_client = genai.Client(api_key=GEMINI_API_KEY)
|
| 24 |
|
| 25 |
-
|
| 26 |
client = chromadb.Client()
|
| 27 |
|
| 28 |
-
|
| 29 |
collection = client.get_or_create_collection(name="inha-well", embedding_function=None)
|
| 30 |
|
| 31 |
-
|
| 32 |
-
# Check if collection is empty to avoid re-ingesting on each run
|
| 33 |
total_docs = collection.count() if hasattr(collection, 'count') else len(collection.get()['documents'])
|
| 34 |
|
| 35 |
if total_docs == 0:
|
| 36 |
content_chunks = []
|
| 37 |
for i in range(1, 4):
|
| 38 |
-
|
| 39 |
folder_path = os.path.join(os.getcwd(), "docs", f"p0000{i}")
|
| 40 |
|
| 41 |
-
|
| 42 |
if not os.path.exists(folder_path):
|
| 43 |
print(f"Warning: Folder {folder_path} not found")
|
| 44 |
continue
|
|
@@ -63,7 +62,6 @@ if total_docs == 0:
|
|
| 63 |
embeddings=embeddings
|
| 64 |
)
|
| 65 |
|
| 66 |
-
# βββ 4. Retrieval & Prompt Utilities ββββββββββββββββββββββββββββββββββββββββ
|
| 67 |
def retrieve_context(question, collection, top_k=2):
|
| 68 |
qr = co.embed(
|
| 69 |
texts=[question],
|
|
@@ -79,7 +77,7 @@ def get_prompt_plain(context: str, question: str) -> str:
|
|
| 79 |
<<START>>
|
| 80 |
You are a responsible person for answering Inha University (South Korea) information. Using the context below, answer within 300 tokens.
|
| 81 |
Provide concise, well-structured, answer-oriented responses using markdown formatting for better readability.
|
| 82 |
-
Use bullet points, bold text, and proper formatting to make the information clear and easy to read.
|
| 83 |
Do not repeat the prompt text in your output.
|
| 84 |
|
| 85 |
Context:
|
|
@@ -101,14 +99,14 @@ def generate_agent_answer(context: str, question: str) -> str:
|
|
| 101 |
stop_sequences=["<<END>>", "<<START>>"]
|
| 102 |
)
|
| 103 |
)
|
| 104 |
-
|
| 105 |
return response.text.strip()
|
| 106 |
|
| 107 |
def rag_answer(question: str, collection) -> str:
|
| 108 |
context = retrieve_context(question, collection, top_k=2)
|
| 109 |
return generate_agent_answer(context, question)
|
| 110 |
|
| 111 |
-
|
| 112 |
def answer_question(question):
|
| 113 |
"""
|
| 114 |
Main function that processes the question and returns the answer
|
|
@@ -122,33 +120,42 @@ def answer_question(question):
|
|
| 122 |
except Exception as e:
|
| 123 |
return f"Sorry, I encountered an error: {str(e)}"
|
| 124 |
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 128 |
fn=answer_question,
|
| 129 |
-
|
|
|
|
| 130 |
label="Ask me anything about Inha University SGCSβ¦",
|
| 131 |
-
placeholder="e.g. How many Major Required credits should I take for graduation?
|
| 132 |
lines=2
|
| 133 |
),
|
| 134 |
-
outputs=gr.Markdown(
|
| 135 |
-
label="π Answer",
|
| 136 |
-
show_copy_button=True
|
| 137 |
-
),
|
| 138 |
title="π Inha University SGCS Info Assistant",
|
| 139 |
-
description="Get answers to your questions about Inha University SGCS
|
| 140 |
theme=gr.themes.Soft(),
|
| 141 |
examples=[
|
| 142 |
-
|
| 143 |
-
|
| 144 |
-
|
| 145 |
]
|
| 146 |
)
|
| 147 |
|
| 148 |
-
|
|
|
|
|
|
|
| 149 |
if __name__ == "__main__":
|
| 150 |
demo.launch(
|
| 151 |
-
share=True,
|
| 152 |
-
server_name="0.0.0.0",
|
| 153 |
|
| 154 |
)
|
|
|
|
| 1 |
import os
|
| 2 |
from dotenv import load_dotenv
|
| 3 |
+
import gradio as gr
|
| 4 |
+
|
| 5 |
|
|
|
|
| 6 |
load_dotenv()
|
| 7 |
COHERE_API_KEY = os.getenv("COHERE_API_KEY")
|
| 8 |
GEMINI_API_KEY = os.getenv("GEMINI_API_KEY")
|
| 9 |
|
| 10 |
if not COHERE_API_KEY or not GEMINI_API_KEY:
|
| 11 |
+
raise ValueError("COHERE_API_KEY or GEMINI_API_KEY is missing")
|
| 12 |
+
|
| 13 |
|
|
|
|
| 14 |
import cohere
|
| 15 |
import chromadb
|
| 16 |
from google import genai
|
| 17 |
from google.genai import types
|
| 18 |
|
| 19 |
+
|
| 20 |
co = cohere.Client(COHERE_API_KEY)
|
| 21 |
|
| 22 |
+
|
| 23 |
genai_client = genai.Client(api_key=GEMINI_API_KEY)
|
| 24 |
|
| 25 |
+
|
| 26 |
client = chromadb.Client()
|
| 27 |
|
| 28 |
+
|
| 29 |
collection = client.get_or_create_collection(name="inha-well", embedding_function=None)
|
| 30 |
|
| 31 |
+
|
|
|
|
| 32 |
total_docs = collection.count() if hasattr(collection, 'count') else len(collection.get()['documents'])
|
| 33 |
|
| 34 |
if total_docs == 0:
|
| 35 |
content_chunks = []
|
| 36 |
for i in range(1, 4):
|
| 37 |
+
|
| 38 |
folder_path = os.path.join(os.getcwd(), "docs", f"p0000{i}")
|
| 39 |
|
| 40 |
+
|
| 41 |
if not os.path.exists(folder_path):
|
| 42 |
print(f"Warning: Folder {folder_path} not found")
|
| 43 |
continue
|
|
|
|
| 62 |
embeddings=embeddings
|
| 63 |
)
|
| 64 |
|
|
|
|
| 65 |
def retrieve_context(question, collection, top_k=2):
|
| 66 |
qr = co.embed(
|
| 67 |
texts=[question],
|
|
|
|
| 77 |
<<START>>
|
| 78 |
You are a responsible person for answering Inha University (South Korea) information. Using the context below, answer within 300 tokens.
|
| 79 |
Provide concise, well-structured, answer-oriented responses using markdown formatting for better readability.
|
| 80 |
+
Use bullet points, bold text, and proper formatting to make the information interactive, clear and easy to read.
|
| 81 |
Do not repeat the prompt text in your output.
|
| 82 |
|
| 83 |
Context:
|
|
|
|
| 99 |
stop_sequences=["<<END>>", "<<START>>"]
|
| 100 |
)
|
| 101 |
)
|
| 102 |
+
|
| 103 |
return response.text.strip()
|
| 104 |
|
| 105 |
def rag_answer(question: str, collection) -> str:
|
| 106 |
context = retrieve_context(question, collection, top_k=2)
|
| 107 |
return generate_agent_answer(context, question)
|
| 108 |
|
| 109 |
+
|
| 110 |
def answer_question(question):
|
| 111 |
"""
|
| 112 |
Main function that processes the question and returns the answer
|
|
|
|
| 120 |
except Exception as e:
|
| 121 |
return f"Sorry, I encountered an error: {str(e)}"
|
| 122 |
|
| 123 |
+
|
| 124 |
+
import gradio as gr
|
| 125 |
+
|
| 126 |
+
def answer_question(message, history):
|
| 127 |
+
history = history + [[message, ""]]
|
| 128 |
+
|
| 129 |
+
# Replace this with your real logic / LLM call
|
| 130 |
+
response = "Here is the answer about SGCS..." # β Dummy response
|
| 131 |
+
history[-1][1] = response
|
| 132 |
+
return history, history
|
| 133 |
+
|
| 134 |
+
# Define interface as chatbot-style inside Interface-like wrapper
|
| 135 |
+
chat_interface = gr.ChatInterface(
|
| 136 |
fn=answer_question,
|
| 137 |
+
chatbot=gr.Chatbot(label="π Inha University SGCS Info Assistant"),
|
| 138 |
+
textbox=gr.Textbox(
|
| 139 |
label="Ask me anything about Inha University SGCSβ¦",
|
| 140 |
+
placeholder="e.g. How many Major Required credits should I take for graduation?",
|
| 141 |
lines=2
|
| 142 |
),
|
|
|
|
|
|
|
|
|
|
|
|
|
| 143 |
title="π Inha University SGCS Info Assistant",
|
| 144 |
+
description="Get answers to your questions about Inha University SGCS.",
|
| 145 |
theme=gr.themes.Soft(),
|
| 146 |
examples=[
|
| 147 |
+
"What classes should I normally take as 3rd semester ISE student?",
|
| 148 |
+
"Tell me about student organizations and activities",
|
| 149 |
+
"What percentage scholarship could I receive with IELTS 7.0"
|
| 150 |
]
|
| 151 |
)
|
| 152 |
|
| 153 |
+
demo = chat_interface
|
| 154 |
+
|
| 155 |
+
|
| 156 |
if __name__ == "__main__":
|
| 157 |
demo.launch(
|
| 158 |
+
share=True,
|
| 159 |
+
server_name="0.0.0.0",
|
| 160 |
|
| 161 |
)
|