Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -2,10 +2,19 @@ import os
|
|
| 2 |
import re
|
| 3 |
import io
|
| 4 |
import json
|
|
|
|
| 5 |
import sqlite3
|
| 6 |
import tempfile
|
|
|
|
| 7 |
import pandas as pd
|
| 8 |
import urllib.request
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
|
| 10 |
from fastapi import FastAPI, File, UploadFile, HTTPException, Request
|
| 11 |
from fastapi.staticfiles import StaticFiles
|
|
@@ -41,89 +50,85 @@ class QueryRequest(BaseModel):
|
|
| 41 |
question: str
|
| 42 |
|
| 43 |
|
| 44 |
-
# ─────────────────────────────
|
| 45 |
-
# HEURISTIC SQL — disabled, Gemini handles everything
|
| 46 |
-
# ─────────────────────────────
|
| 47 |
-
|
| 48 |
-
def _heuristic_sql(question: str, table: str, cols: list = None):
|
| 49 |
-
return None
|
| 50 |
-
|
| 51 |
-
|
| 52 |
# ─────────────────────────────
|
| 53 |
# GEMINI CALL
|
|
|
|
|
|
|
| 54 |
# ─────────────────────────────
|
| 55 |
|
| 56 |
-
def _call_gemini(question: str, schema: str, columns: list, table: str
|
|
|
|
| 57 |
|
| 58 |
if not GEMINI_API_KEY:
|
| 59 |
return ""
|
| 60 |
|
| 61 |
prompt = f"""
|
| 62 |
You are an expert SQLite data analyst.
|
| 63 |
-
|
| 64 |
Return ONLY a valid SQL query. No explanation, no markdown, no text before or after.
|
| 65 |
-
|
| 66 |
STRICT RULES:
|
| 67 |
- Use only the table "{table}"
|
| 68 |
- Never use DROP or DELETE
|
| 69 |
- For GROUP BY queries, always include the grouped column in SELECT
|
| 70 |
- For filtering, use the exact case as in the data
|
| 71 |
- Return only one SQL statement ending without semicolon
|
| 72 |
-
|
| 73 |
COLUMNS AVAILABLE:
|
| 74 |
{", ".join(columns)}
|
| 75 |
-
|
| 76 |
SCHEMA:
|
| 77 |
{schema}
|
| 78 |
-
|
| 79 |
QUESTION:
|
| 80 |
{question}
|
| 81 |
"""
|
| 82 |
|
| 83 |
url = (
|
| 84 |
-
|
| 85 |
-
|
| 86 |
f":generateContent?key={GEMINI_API_KEY}"
|
| 87 |
)
|
| 88 |
|
| 89 |
payload = json.dumps({
|
| 90 |
-
"contents": [
|
| 91 |
-
{
|
| 92 |
-
"role": "user",
|
| 93 |
-
"parts": [{"text": prompt}]
|
| 94 |
-
}
|
| 95 |
-
]
|
| 96 |
}).encode("utf-8")
|
| 97 |
|
| 98 |
-
|
| 99 |
-
req = urllib.request.Request(
|
| 100 |
-
url,
|
| 101 |
-
data=payload,
|
| 102 |
-
headers={"Content-Type": "application/json"}
|
| 103 |
-
)
|
| 104 |
-
|
| 105 |
-
with urllib.request.urlopen(req, timeout=30) as resp:
|
| 106 |
-
data = json.loads(resp.read().decode())
|
| 107 |
-
|
| 108 |
try:
|
| 109 |
-
|
| 110 |
-
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 114 |
return ""
|
| 115 |
|
| 116 |
-
|
| 117 |
-
sql = sql.split(";")[0].strip()
|
| 118 |
-
|
| 119 |
-
if "drop" in sql.lower() or "delete" in sql.lower():
|
| 120 |
-
return f'SELECT * FROM "{table}" LIMIT 10'
|
| 121 |
-
|
| 122 |
-
return sql
|
| 123 |
-
|
| 124 |
-
except Exception as e:
|
| 125 |
-
print("❌ GEMINI ERROR:", e)
|
| 126 |
-
return ""
|
| 127 |
|
| 128 |
|
| 129 |
# ─────────────────────────────
|
|
@@ -173,7 +178,6 @@ async def upload_csv(file: UploadFile = File(...)):
|
|
| 173 |
table_name = re.sub(r"[^a-zA-Z0-9_]", "_", file.filename)
|
| 174 |
if table_name and table_name[0].isdigit():
|
| 175 |
table_name = "t_" + table_name
|
| 176 |
-
|
| 177 |
table_name = table_name[:40]
|
| 178 |
|
| 179 |
with tempfile.NamedTemporaryFile(delete=False) as tf:
|
|
@@ -185,7 +189,6 @@ async def upload_csv(file: UploadFile = File(...)):
|
|
| 185 |
).fetchone()[0]
|
| 186 |
|
| 187 |
conn.close()
|
| 188 |
-
|
| 189 |
db_bytes = open(tf.name, "rb").read()
|
| 190 |
|
| 191 |
os.remove(tf.name)
|
|
@@ -195,7 +198,6 @@ async def upload_csv(file: UploadFile = File(...)):
|
|
| 195 |
"table": table_name,
|
| 196 |
"cols": list(df.columns)
|
| 197 |
}
|
| 198 |
-
|
| 199 |
_schema_store[session_id] = schema
|
| 200 |
|
| 201 |
return {
|
|
@@ -217,14 +219,12 @@ async def upload_csv(file: UploadFile = File(...)):
|
|
| 217 |
async def query(req: QueryRequest):
|
| 218 |
|
| 219 |
data = _db_store.get(req.session_id)
|
| 220 |
-
|
| 221 |
if not data and _db_store:
|
| 222 |
data = list(_db_store.values())[-1]
|
| 223 |
-
|
| 224 |
if not data:
|
| 225 |
raise HTTPException(status_code=404, detail="No dataset loaded")
|
| 226 |
|
| 227 |
-
table
|
| 228 |
schema = (
|
| 229 |
_schema_store.get(req.session_id)
|
| 230 |
or list(_schema_store.values())[-1]
|
|
@@ -232,15 +232,17 @@ async def query(req: QueryRequest):
|
|
| 232 |
|
| 233 |
sql = _call_gemini(req.question, schema, data["cols"], table)
|
| 234 |
|
| 235 |
-
|
| 236 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 237 |
|
| 238 |
results = execute_sql(sql, data["bytes"])
|
| 239 |
|
| 240 |
-
return {
|
| 241 |
-
"sql": sql,
|
| 242 |
-
"results": results
|
| 243 |
-
}
|
| 244 |
|
| 245 |
|
| 246 |
# ─────────────────────────────
|
|
@@ -257,32 +259,23 @@ def health():
|
|
| 257 |
|
| 258 |
|
| 259 |
# ─────────────────────────────
|
| 260 |
-
# TELEGRAM WEBHOOK
|
|
|
|
|
|
|
| 261 |
# ─────────────────────────────
|
| 262 |
|
| 263 |
-
@app.on_event("startup")
|
| 264 |
-
async def on_startup():
|
| 265 |
-
|
| 266 |
-
|
| 267 |
-
setup_webhook()
|
| 268 |
-
except Exception as e:
|
| 269 |
-
print(f"⚠️ Webhook setup skipped: {e}")
|
| 270 |
|
|
|
|
|
|
|
|
|
|
| 271 |
|
| 272 |
-
@app.post("/webhook/{token}")
|
| 273 |
-
async def telegram_webhook(token: str, request: Request):
|
| 274 |
-
|
| 275 |
-
if token != expected:
|
| 276 |
-
raise HTTPException(status_code=403, detail="Invalid token")
|
| 277 |
-
try:
|
| 278 |
-
import telebot
|
| 279 |
-
from bot import bot
|
| 280 |
-
json_data = await request.json()
|
| 281 |
-
update = telebot.types.Update.de_json(json_data)
|
| 282 |
-
bot.process_new_updates([update])
|
| 283 |
-
except Exception as e:
|
| 284 |
-
print(f"❌ Webhook error: {e}")
|
| 285 |
-
return {"ok": True}
|
| 286 |
|
| 287 |
|
| 288 |
# ─────────────────────────────
|
|
|
|
| 2 |
import re
|
| 3 |
import io
|
| 4 |
import json
|
| 5 |
+
import time
|
| 6 |
import sqlite3
|
| 7 |
import tempfile
|
| 8 |
+
import threading
|
| 9 |
import pandas as pd
|
| 10 |
import urllib.request
|
| 11 |
+
import urllib.error
|
| 12 |
+
|
| 13 |
+
try:
|
| 14 |
+
import requests as _requests
|
| 15 |
+
HAS_REQUESTS = True
|
| 16 |
+
except ImportError:
|
| 17 |
+
HAS_REQUESTS = False
|
| 18 |
|
| 19 |
from fastapi import FastAPI, File, UploadFile, HTTPException, Request
|
| 20 |
from fastapi.staticfiles import StaticFiles
|
|
|
|
| 50 |
question: str
|
| 51 |
|
| 52 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 53 |
# ─────────────────────────────
|
| 54 |
# GEMINI CALL
|
| 55 |
+
# retries=1 → fail fast, tell user immediately
|
| 56 |
+
# do NOT change retries back to 4
|
| 57 |
# ─────────────────────────────
|
| 58 |
|
| 59 |
+
def _call_gemini(question: str, schema: str, columns: list, table: str,
|
| 60 |
+
retries: int = 1):
|
| 61 |
|
| 62 |
if not GEMINI_API_KEY:
|
| 63 |
return ""
|
| 64 |
|
| 65 |
prompt = f"""
|
| 66 |
You are an expert SQLite data analyst.
|
|
|
|
| 67 |
Return ONLY a valid SQL query. No explanation, no markdown, no text before or after.
|
|
|
|
| 68 |
STRICT RULES:
|
| 69 |
- Use only the table "{table}"
|
| 70 |
- Never use DROP or DELETE
|
| 71 |
- For GROUP BY queries, always include the grouped column in SELECT
|
| 72 |
- For filtering, use the exact case as in the data
|
| 73 |
- Return only one SQL statement ending without semicolon
|
|
|
|
| 74 |
COLUMNS AVAILABLE:
|
| 75 |
{", ".join(columns)}
|
|
|
|
| 76 |
SCHEMA:
|
| 77 |
{schema}
|
|
|
|
| 78 |
QUESTION:
|
| 79 |
{question}
|
| 80 |
"""
|
| 81 |
|
| 82 |
url = (
|
| 83 |
+
"https://generativelanguage.googleapis.com"
|
| 84 |
+
"/v1beta/models/gemini-2.5-flash"
|
| 85 |
f":generateContent?key={GEMINI_API_KEY}"
|
| 86 |
)
|
| 87 |
|
| 88 |
payload = json.dumps({
|
| 89 |
+
"contents": [{"role": "user", "parts": [{"text": prompt}]}]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 90 |
}).encode("utf-8")
|
| 91 |
|
| 92 |
+
for attempt in range(1, retries + 1):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 93 |
try:
|
| 94 |
+
req = urllib.request.Request(
|
| 95 |
+
url,
|
| 96 |
+
data=payload,
|
| 97 |
+
headers={"Content-Type": "application/json"}
|
| 98 |
+
)
|
| 99 |
+
with urllib.request.urlopen(req, timeout=30) as resp:
|
| 100 |
+
data = json.loads(resp.read().decode())
|
| 101 |
+
|
| 102 |
+
try:
|
| 103 |
+
sql = data["candidates"][0]["content"]["parts"][0]["text"]
|
| 104 |
+
except Exception:
|
| 105 |
+
return ""
|
| 106 |
+
|
| 107 |
+
if not sql:
|
| 108 |
+
return ""
|
| 109 |
+
|
| 110 |
+
sql = sql.replace("```sql", "").replace("```", "").strip()
|
| 111 |
+
sql = sql.split(";")[0].strip()
|
| 112 |
+
|
| 113 |
+
if "drop" in sql.lower() or "delete" in sql.lower():
|
| 114 |
+
return f'SELECT * FROM "{table}" LIMIT 10'
|
| 115 |
+
|
| 116 |
+
return sql
|
| 117 |
+
|
| 118 |
+
except urllib.error.HTTPError as e:
|
| 119 |
+
if e.code == 429:
|
| 120 |
+
print(f"⚠️ Gemini 429 — rate limited (attempt {attempt}/{retries})")
|
| 121 |
+
# fail fast — do not retry, tell user immediately
|
| 122 |
+
return ""
|
| 123 |
+
else:
|
| 124 |
+
print(f"❌ GEMINI HTTP ERROR {e.code}: {e}")
|
| 125 |
+
return ""
|
| 126 |
+
|
| 127 |
+
except Exception as e:
|
| 128 |
+
print(f"❌ GEMINI ERROR: {e}")
|
| 129 |
return ""
|
| 130 |
|
| 131 |
+
return ""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 132 |
|
| 133 |
|
| 134 |
# ─────────────────────────────
|
|
|
|
| 178 |
table_name = re.sub(r"[^a-zA-Z0-9_]", "_", file.filename)
|
| 179 |
if table_name and table_name[0].isdigit():
|
| 180 |
table_name = "t_" + table_name
|
|
|
|
| 181 |
table_name = table_name[:40]
|
| 182 |
|
| 183 |
with tempfile.NamedTemporaryFile(delete=False) as tf:
|
|
|
|
| 189 |
).fetchone()[0]
|
| 190 |
|
| 191 |
conn.close()
|
|
|
|
| 192 |
db_bytes = open(tf.name, "rb").read()
|
| 193 |
|
| 194 |
os.remove(tf.name)
|
|
|
|
| 198 |
"table": table_name,
|
| 199 |
"cols": list(df.columns)
|
| 200 |
}
|
|
|
|
| 201 |
_schema_store[session_id] = schema
|
| 202 |
|
| 203 |
return {
|
|
|
|
| 219 |
async def query(req: QueryRequest):
|
| 220 |
|
| 221 |
data = _db_store.get(req.session_id)
|
|
|
|
| 222 |
if not data and _db_store:
|
| 223 |
data = list(_db_store.values())[-1]
|
|
|
|
| 224 |
if not data:
|
| 225 |
raise HTTPException(status_code=404, detail="No dataset loaded")
|
| 226 |
|
| 227 |
+
table = data["table"]
|
| 228 |
schema = (
|
| 229 |
_schema_store.get(req.session_id)
|
| 230 |
or list(_schema_store.values())[-1]
|
|
|
|
| 232 |
|
| 233 |
sql = _call_gemini(req.question, schema, data["cols"], table)
|
| 234 |
|
| 235 |
+
# If Gemini failed (429 or error), return clear error to Render bot
|
| 236 |
+
if not sql:
|
| 237 |
+
return {
|
| 238 |
+
"sql": "",
|
| 239 |
+
"results": [],
|
| 240 |
+
"error": "⚠️ AI busy (rate limit). Wait 60s and retry."
|
| 241 |
+
}
|
| 242 |
|
| 243 |
results = execute_sql(sql, data["bytes"])
|
| 244 |
|
| 245 |
+
return {"sql": sql, "results": results}
|
|
|
|
|
|
|
|
|
|
| 246 |
|
| 247 |
|
| 248 |
# ─────────────────────────────
|
|
|
|
| 259 |
|
| 260 |
|
| 261 |
# ─────────────────────────────
|
| 262 |
+
# TELEGRAM WEBHOOK — DISABLED
|
| 263 |
+
# Telegram is now handled by Render bot (polling)
|
| 264 |
+
# Do not re-enable — causes conflict with Render polling
|
| 265 |
# ─────────────────────────────
|
| 266 |
|
| 267 |
+
# @app.on_event("startup")
|
| 268 |
+
# async def on_startup():
|
| 269 |
+
# thread = threading.Thread(target=_background_webhook_setup, daemon=True)
|
| 270 |
+
# thread.start()
|
|
|
|
|
|
|
|
|
|
| 271 |
|
| 272 |
+
# @app.get("/set-webhook")
|
| 273 |
+
# async def set_webhook_endpoint():
|
| 274 |
+
# ...
|
| 275 |
|
| 276 |
+
# @app.post("/webhook/{token}")
|
| 277 |
+
# async def telegram_webhook(token: str, request: Request):
|
| 278 |
+
# ...
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 279 |
|
| 280 |
|
| 281 |
# ─────────────────────────────
|