Spaces:
Sleeping
Sleeping
Create demo.py
Browse files
demo.py
ADDED
@@ -0,0 +1,357 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from flask import Flask, render_template, request, redirect, url_for, send_from_directory, flash
|
2 |
+
from flask_socketio import SocketIO
|
3 |
+
import threading
|
4 |
+
import os
|
5 |
+
from dotenv import load_dotenv
|
6 |
+
import sqlite3
|
7 |
+
from werkzeug.utils import secure_filename
|
8 |
+
|
9 |
+
# LangChain and agent imports
|
10 |
+
from langchain_community.chat_models.huggingface import ChatHuggingFace # if needed later
|
11 |
+
from langchain.agents import Tool
|
12 |
+
from langchain.agents.format_scratchpad import format_log_to_str
|
13 |
+
from langchain.agents.output_parsers import ReActJsonSingleInputOutputParser
|
14 |
+
from langchain_core.callbacks import CallbackManager, BaseCallbackHandler
|
15 |
+
from langchain_community.agent_toolkits.load_tools import load_tools
|
16 |
+
from langchain_core.tools import tool
|
17 |
+
from langchain_community.agent_toolkits import PowerBIToolkit
|
18 |
+
from langchain.chains import LLMMathChain
|
19 |
+
from langchain import hub
|
20 |
+
from langchain_community.tools import DuckDuckGoSearchRun
|
21 |
+
|
22 |
+
# Agent requirements and type hints
|
23 |
+
from typing import Annotated, Literal, TypedDict, Any
|
24 |
+
from langchain_core.messages import AIMessage, ToolMessage
|
25 |
+
from pydantic import BaseModel, Field
|
26 |
+
from typing_extensions import TypedDict
|
27 |
+
from langgraph.graph import END, StateGraph, START
|
28 |
+
from langgraph.graph.message import AnyMessage, add_messages
|
29 |
+
from langchain_core.runnables import RunnableLambda, RunnableWithFallbacks
|
30 |
+
from langgraph.prebuilt import ToolNode
|
31 |
+
|
32 |
+
import traceback
|
33 |
+
|
34 |
+
# Load environment variables
|
35 |
+
load_dotenv()
|
36 |
+
|
37 |
+
|
38 |
+
# Global configuration variables
|
39 |
+
UPLOAD_FOLDER = os.path.join(os.getcwd(), "uploads")
|
40 |
+
BASE_DIR = os.path.abspath(os.path.dirname(__file__))
|
41 |
+
DATABASE_URI = f"sqlite:///{os.path.join(BASE_DIR, 'data', 'mydb.db')}"
|
42 |
+
print("DATABASE URI:", DATABASE_URI)
|
43 |
+
|
44 |
+
# API Keys from .env file
|
45 |
+
GROQ_API_KEY = os.getenv("GROQ_API_KEY")
|
46 |
+
MISTRAL_API_KEY = os.getenv("MISTRAL_API_KEY")
|
47 |
+
os.environ["GROQ_API_KEY"] = GROQ_API_KEY
|
48 |
+
os.environ["MISTRAL_API_KEY"] = MISTRAL_API_KEY
|
49 |
+
|
50 |
+
# Global variables for dynamic agent and DB file path; initially None.
|
51 |
+
agent_app = None
|
52 |
+
abs_file_path = None
|
53 |
+
db_path = None
|
54 |
+
|
55 |
+
print(traceback.format_exc())
|
56 |
+
|
57 |
+
# =============================================================================
|
58 |
+
# create_agent_app: Given a database path, initialize the agent workflow.
|
59 |
+
# =============================================================================
|
60 |
+
|
61 |
+
def create_agent_app(db_path: str):
|
62 |
+
# Use ChatGroq as our LLM here; you can swap to ChatMistralAI if preferred.
|
63 |
+
from langchain_groq import ChatGroq
|
64 |
+
llm = ChatGroq(model="llama3-70b-8192")
|
65 |
+
|
66 |
+
# -------------------------------------------------------------------------
|
67 |
+
# Define a tool for executing SQL queries.
|
68 |
+
# -------------------------------------------------------------------------
|
69 |
+
@tool
|
70 |
+
def db_query_tool(query: str) -> str:
|
71 |
+
"""
|
72 |
+
Executes a SQL query on the connected SQLite database.
|
73 |
+
|
74 |
+
Parameters:
|
75 |
+
query (str): A SQL query string to be executed.
|
76 |
+
|
77 |
+
Returns:
|
78 |
+
str: The result from the database if successful, or an error message if not.
|
79 |
+
"""
|
80 |
+
result = db_instance.run_no_throw(query)
|
81 |
+
return result if result else "Error: Query failed. Please rewrite your query and try again."
|
82 |
+
|
83 |
+
# -------------------------------------------------------------------------
|
84 |
+
# Pydantic model for final answer
|
85 |
+
# -------------------------------------------------------------------------
|
86 |
+
class SubmitFinalAnswer(BaseModel):
|
87 |
+
final_answer: str = Field(..., description="The final answer to the user")
|
88 |
+
|
89 |
+
# -------------------------------------------------------------------------
|
90 |
+
# Define state type for our workflow.
|
91 |
+
# -------------------------------------------------------------------------
|
92 |
+
class State(TypedDict):
|
93 |
+
messages: Annotated[list[AnyMessage], add_messages]
|
94 |
+
|
95 |
+
# -------------------------------------------------------------------------
|
96 |
+
# Set up prompt templates (using langchain_core.prompts) for query checking
|
97 |
+
# and query generation.
|
98 |
+
# -------------------------------------------------------------------------
|
99 |
+
from langchain_core.prompts import ChatPromptTemplate
|
100 |
+
|
101 |
+
query_check_system = (
|
102 |
+
"You are a SQL expert with a strong attention to detail.\n"
|
103 |
+
"Double check the SQLite query for common mistakes, including:\n"
|
104 |
+
"- Using NOT IN with NULL values\n"
|
105 |
+
"- Using UNION when UNION ALL should have been used\n"
|
106 |
+
"- Using BETWEEN for exclusive ranges\n"
|
107 |
+
"- Data type mismatch in predicates\n"
|
108 |
+
"- Properly quoting identifiers\n"
|
109 |
+
"- Using the correct number of arguments for functions\n"
|
110 |
+
"- Casting to the correct data type\n"
|
111 |
+
"- Using the proper columns for joins\n\n"
|
112 |
+
"If there are any of the above mistakes, rewrite the query. If there are no mistakes, just reproduce the original query.\n"
|
113 |
+
"You will call the appropriate tool to execute the query after running this check."
|
114 |
+
)
|
115 |
+
query_check_prompt = ChatPromptTemplate.from_messages([
|
116 |
+
("system", query_check_system),
|
117 |
+
("placeholder", "{messages}")
|
118 |
+
])
|
119 |
+
query_check = query_check_prompt | llm.bind_tools([db_query_tool])
|
120 |
+
|
121 |
+
query_gen_system = (
|
122 |
+
"You are a SQL expert with a strong attention to detail.\n\n"
|
123 |
+
"Given an input question, output a syntactically correct SQLite query to run, then look at the results of the query and return the answer.\n\n"
|
124 |
+
"DO NOT call any tool besides SubmitFinalAnswer to submit the final answer.\n\n"
|
125 |
+
"When generating the query:\n"
|
126 |
+
"Output the SQL query that answers the input question without a tool call.\n"
|
127 |
+
"Unless the user specifies a specific number of examples they wish to obtain, always limit your query to at most 5 results.\n"
|
128 |
+
"You can order the results by a relevant column to return the most interesting examples in the database.\n"
|
129 |
+
"Never query for all the columns from a specific table, only ask for the relevant columns given the question.\n\n"
|
130 |
+
"If you get an error while executing a query, rewrite the query and try again.\n"
|
131 |
+
"If you get an empty result set, you should try to rewrite the query to get a non-empty result set.\n"
|
132 |
+
"NEVER make stuff up if you don't have enough information to answer the query... just say you don't have enough information.\n\n"
|
133 |
+
"If you have enough information to answer the input question, simply invoke the appropriate tool to submit the final answer to the user.\n"
|
134 |
+
"DO NOT make any DML statements (INSERT, UPDATE, DELETE, DROP etc.) to the database. Do not return any SQL query except answer."
|
135 |
+
)
|
136 |
+
query_gen_prompt = ChatPromptTemplate.from_messages([
|
137 |
+
("system", query_gen_system),
|
138 |
+
("placeholder", "{messages}")
|
139 |
+
])
|
140 |
+
query_gen = query_gen_prompt | llm.bind_tools([SubmitFinalAnswer])
|
141 |
+
|
142 |
+
# -------------------------------------------------------------------------
|
143 |
+
# Update database URI and file path, create SQLDatabase connection.
|
144 |
+
# -------------------------------------------------------------------------
|
145 |
+
|
146 |
+
abs_db_path_local = os.path.abspath(db_path)
|
147 |
+
global DATABASE_URI
|
148 |
+
DATABASE_URI = abs_db_path_local
|
149 |
+
db_uri = f"sqlite:///{abs_db_path_local}"
|
150 |
+
print("db_uri", db_uri)
|
151 |
+
# Uncomment if flash is needed; ensure you have flask.flash imported if so.
|
152 |
+
# flash(f"db_uri:{db_uri}", "warning")
|
153 |
+
|
154 |
+
from langchain_community.utilities import SQLDatabase
|
155 |
+
db_instance = SQLDatabase.from_uri(db_uri)
|
156 |
+
print("db_instance----->", db_instance)
|
157 |
+
# flash(f"db_instance:{db_instance}", "warning")
|
158 |
+
|
159 |
+
# -------------------------------------------------------------------------
|
160 |
+
# Create SQL toolkit.
|
161 |
+
# -------------------------------------------------------------------------
|
162 |
+
|
163 |
+
from langchain_community.agent_toolkits import SQLDatabaseToolkit
|
164 |
+
toolkit_instance = SQLDatabaseToolkit(db=db_instance, llm=llm)
|
165 |
+
tools_instance = toolkit_instance.get_tools()
|
166 |
+
|
167 |
+
# -------------------------------------------------------------------------
|
168 |
+
# Define workflow nodes and fallback functions.
|
169 |
+
# -------------------------------------------------------------------------
|
170 |
+
|
171 |
+
def first_tool_call(state: State) -> dict[str, list[AIMessage]]:
|
172 |
+
return {"messages": [AIMessage(content="", tool_calls=[{"name": "sql_db_list_tables", "args": {}, "id": "tool_abcd123"}])]}
|
173 |
+
|
174 |
+
def handle_tool_error(state: State) -> dict:
|
175 |
+
error = state.get("error")
|
176 |
+
tool_calls = state["messages"][-1].tool_calls
|
177 |
+
return {"messages": [
|
178 |
+
ToolMessage(content=f"Error: {repr(error)}. Please fix your mistakes.", tool_call_id=tc["id"])
|
179 |
+
for tc in tool_calls
|
180 |
+
]}
|
181 |
+
|
182 |
+
def create_tool_node_with_fallback(tools_list: list) -> RunnableWithFallbacks[Any, dict]:
|
183 |
+
return ToolNode(tools_list).with_fallbacks([RunnableLambda(handle_tool_error)], exception_key="error")
|
184 |
+
|
185 |
+
def query_gen_node(state: State):
|
186 |
+
message = query_gen.invoke(state)
|
187 |
+
tool_messages = []
|
188 |
+
if message.tool_calls:
|
189 |
+
for tc in message.tool_calls:
|
190 |
+
if tc["name"] != "SubmitFinalAnswer":
|
191 |
+
tool_messages.append(ToolMessage(
|
192 |
+
content=f"Error: The wrong tool was called: {tc['name']}. Please fix your mistakes.",
|
193 |
+
tool_call_id=tc["id"]
|
194 |
+
))
|
195 |
+
return {"messages": [message] + tool_messages}
|
196 |
+
|
197 |
+
def should_continue(state: State) -> Literal[END, "correct_query", "query_gen"]:
|
198 |
+
messages = state["messages"]
|
199 |
+
last_message = messages[-1]
|
200 |
+
if getattr(last_message, "tool_calls", None):
|
201 |
+
return END
|
202 |
+
if last_message.content.startswith("Error:"):
|
203 |
+
return "query_gen"
|
204 |
+
return "correct_query"
|
205 |
+
|
206 |
+
def model_check_query(state: State) -> dict[str, list[AIMessage]]:
|
207 |
+
return {"messages": [query_check.invoke({"messages": [state["messages"][-1]]})]}
|
208 |
+
|
209 |
+
# -------------------------------------------------------------------------
|
210 |
+
# Get tools for listing tables and fetching schema.
|
211 |
+
# -------------------------------------------------------------------------
|
212 |
+
|
213 |
+
list_tables_tool = next((tool for tool in tools_instance if tool.name == "sql_db_list_tables"), None)
|
214 |
+
get_schema_tool = next((tool for tool in tools_instance if tool.name == "sql_db_schema"), None)
|
215 |
+
|
216 |
+
workflow = StateGraph(State)
|
217 |
+
workflow.add_node("first_tool_call", first_tool_call)
|
218 |
+
workflow.add_node("list_tables_tool", create_tool_node_with_fallback([list_tables_tool]))
|
219 |
+
workflow.add_node("get_schema_tool", create_tool_node_with_fallback([get_schema_tool]))
|
220 |
+
model_get_schema = llm.bind_tools([get_schema_tool])
|
221 |
+
workflow.add_node("model_get_schema", lambda state: {"messages": [model_get_schema.invoke(state["messages"])],})
|
222 |
+
workflow.add_node("query_gen", query_gen_node)
|
223 |
+
workflow.add_node("correct_query", model_check_query)
|
224 |
+
workflow.add_node("execute_query", create_tool_node_with_fallback([db_query_tool]))
|
225 |
+
|
226 |
+
workflow.add_edge(START, "first_tool_call")
|
227 |
+
workflow.add_edge("first_tool_call", "list_tables_tool")
|
228 |
+
workflow.add_edge("list_tables_tool", "model_get_schema")
|
229 |
+
workflow.add_edge("model_get_schema", "get_schema_tool")
|
230 |
+
workflow.add_edge("get_schema_tool", "query_gen")
|
231 |
+
workflow.add_conditional_edges("query_gen", should_continue)
|
232 |
+
workflow.add_edge("correct_query", "execute_query")
|
233 |
+
workflow.add_edge("execute_query", "query_gen")
|
234 |
+
|
235 |
+
# Return compiled workflow
|
236 |
+
return workflow.compile()
|
237 |
+
|
238 |
+
|
239 |
+
# =============================================================================
|
240 |
+
# create_app: The application factory.
|
241 |
+
# =============================================================================
|
242 |
+
|
243 |
+
def create_app():
|
244 |
+
flask_app = Flask(__name__, static_url_path='/uploads', static_folder='uploads')
|
245 |
+
socketio = SocketIO(flask_app, cors_allowed_origins="*")
|
246 |
+
|
247 |
+
# Ensure uploads folder exists.
|
248 |
+
if not os.path.exists(UPLOAD_FOLDER):
|
249 |
+
os.makedirs(UPLOAD_FOLDER)
|
250 |
+
flask_app.config['UPLOAD_FOLDER'] = UPLOAD_FOLDER
|
251 |
+
|
252 |
+
# -------------------------------------------------------------------------
|
253 |
+
# Serve uploaded files via a custom route.
|
254 |
+
# -------------------------------------------------------------------------
|
255 |
+
|
256 |
+
@flask_app.route("/files/<path:filename>")
|
257 |
+
def uploaded_file(filename):
|
258 |
+
return send_from_directory(flask_app.config['UPLOAD_FOLDER'], filename)
|
259 |
+
|
260 |
+
# -------------------------------------------------------------------------
|
261 |
+
# Helper: run_agent runs the agent with the given prompt.
|
262 |
+
# -------------------------------------------------------------------------
|
263 |
+
|
264 |
+
def run_agent(prompt, socketio):
|
265 |
+
global agent_app, abs_file_path, db_path
|
266 |
+
if not abs_file_path:
|
267 |
+
socketio.emit("log", {"message": "[ERROR]: No DB file uploaded."})
|
268 |
+
socketio.emit("final", {"message": "No database available. Please upload one and try again."})
|
269 |
+
return
|
270 |
+
|
271 |
+
try:
|
272 |
+
# Lazy agent initialization: use the previously uploaded DB.
|
273 |
+
if agent_app is None:
|
274 |
+
print("[INFO]: Initializing agent for the first time...")
|
275 |
+
agent_app = create_agent_app(abs_file_path)
|
276 |
+
socketio.emit("log", {"message": "[INFO]: Agent initialized."})
|
277 |
+
|
278 |
+
query = {"messages": [("user", prompt)]}
|
279 |
+
result = agent_app.invoke(query)
|
280 |
+
try:
|
281 |
+
result = result["messages"][-1].tool_calls[0]["args"]["final_answer"]
|
282 |
+
except Exception:
|
283 |
+
result = "Query failed or no valid answer found."
|
284 |
+
|
285 |
+
print("final_answer------>", result)
|
286 |
+
socketio.emit("final", {"message": result})
|
287 |
+
except Exception as e:
|
288 |
+
print(f"[ERROR]: {str(e)}")
|
289 |
+
socketio.emit("log", {"message": f"[ERROR]: {str(e)}"})
|
290 |
+
socketio.emit("final", {"message": "Generation failed."})
|
291 |
+
|
292 |
+
# -------------------------------------------------------------------------
|
293 |
+
# Route: index page.
|
294 |
+
# -------------------------------------------------------------------------
|
295 |
+
|
296 |
+
@flask_app.route("/")
|
297 |
+
def index():
|
298 |
+
return render_template("index.html")
|
299 |
+
|
300 |
+
# -------------------------------------------------------------------------
|
301 |
+
# Route: generate (POST) – receives a prompt and runs the agent.
|
302 |
+
# -------------------------------------------------------------------------
|
303 |
+
|
304 |
+
@flask_app.route("/generate", methods=["POST"])
|
305 |
+
def generate():
|
306 |
+
try:
|
307 |
+
socketio.emit("log", {"message": "[STEP]: Entering query_gen..."})
|
308 |
+
data = request.json
|
309 |
+
prompt = data.get("prompt", "")
|
310 |
+
socketio.emit("log", {"message": f"[INFO]: Received prompt: {prompt}"})
|
311 |
+
thread = threading.Thread(target=run_agent, args=(prompt, socketio))
|
312 |
+
socketio.emit("log", {"message": f"[INFO]: Starting thread: {thread}"})
|
313 |
+
thread.start()
|
314 |
+
return "OK", 200
|
315 |
+
except Exception as e:
|
316 |
+
print(f"[ERROR]: {str(e)}")
|
317 |
+
socketio.emit("log", {"message": f"[ERROR]: {str(e)}"})
|
318 |
+
return "ERROR", 500
|
319 |
+
|
320 |
+
# -------------------------------------------------------------------------
|
321 |
+
# Route: upload (GET/POST) – handles uploading the SQLite DB file.
|
322 |
+
# -------------------------------------------------------------------------
|
323 |
+
|
324 |
+
@flask_app.route("/upload", methods=["GET", "POST"])
|
325 |
+
def upload():
|
326 |
+
global abs_file_path, agent_app, db_path
|
327 |
+
try:
|
328 |
+
if request.method == "POST":
|
329 |
+
file = request.files.get("file")
|
330 |
+
if not file:
|
331 |
+
print("No file uploaded")
|
332 |
+
return "No file uploaded", 400
|
333 |
+
filename = secure_filename(file.filename)
|
334 |
+
if filename.endswith('.db'):
|
335 |
+
db_path = os.path.join(flask_app.config['UPLOAD_FOLDER'], "uploaded.db")
|
336 |
+
print("Saving file to:", db_path)
|
337 |
+
file.save(db_path)
|
338 |
+
abs_file_path = os.path.abspath(db_path) # Save it here; agent init will occur on first query.
|
339 |
+
print(f"[INFO]: File '{filename}' uploaded. Agent will be initialized on first query.")
|
340 |
+
socketio.emit("log", {"message": f"[INFO]: Database file '{filename}' uploaded."})
|
341 |
+
return redirect(url_for("index"))
|
342 |
+
return render_template("upload.html")
|
343 |
+
except Exception as e:
|
344 |
+
print(f"[ERROR]: {str(e)}")
|
345 |
+
socketio.emit("log", {"message": f"[ERROR]: {str(e)}"})
|
346 |
+
return render_template("upload.html")
|
347 |
+
|
348 |
+
return flask_app, socketio
|
349 |
+
|
350 |
+
# =============================================================================
|
351 |
+
# Create the app for Gunicorn compatibility.
|
352 |
+
# =============================================================================
|
353 |
+
|
354 |
+
app, socketio_instance = create_app()
|
355 |
+
|
356 |
+
if __name__ == "__main__":
|
357 |
+
socketio_instance.run(app, debug=True)
|