Spaces:
Runtime error
Runtime error
File size: 19,468 Bytes
37a6ee1 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 | import os
from groq import Groq
from typing import Dict, Any
from backend import config
class LLMService:
def __init__(self):
print("[LLMService] Initializing LLMService singleton...")
# Since Groq client requires an API key, we will instantiate it lazily or during init
# If API key is empty/missing, it will raise an error later when invoked.
self.client = None
def _get_client(self) -> Groq:
if self.client is None:
if not config.GROQ_API_KEY:
raise ValueError("Groq API Key is not set in the environment or configuration.")
self.client = Groq(api_key=config.GROQ_API_KEY)
return self.client
def _clean_sql_response(self, response: str) -> str:
"""
Cleans the raw response from the LLM, stripping markdown code blocks,
backticks, and any leading/trailing whitespace.
"""
cleaned = response.strip()
# Strip code blocks starting with ``` (e.g. ```sql or ```)
if cleaned.startswith("```"):
lines = cleaned.split("\n")
sql_lines = [line for line in lines if not line.strip().startswith("```")]
cleaned = "\n".join(sql_lines).strip()
# Strip single backticks
cleaned = cleaned.strip("`").strip()
# Sometimes LLMs add an ending semicolon, which is fine, but let's keep it clean
return cleaned
def generate_sql(self, question: str, schema: dict) -> Dict[str, Any]:
"""
Generates SQLite SELECT query based on the user question and database schema.
"""
print(f"[LLMService] Generating SQL for question: '{question}'")
try:
client = self._get_client()
# Format the columns section of the prompt
columns_prompt = []
for col in schema.get("columns", []):
name = col.get("name")
col_type = col.get("type")
sample_vals = col.get("sample_values", [])
columns_prompt.append(f"- {name} ({col_type}): examples: {sample_vals}")
columns_str = "\n".join(columns_prompt)
system_prompt = (
"You are an expert SQL analyst. Generate SQLite-compatible SELECT queries only.\n"
"Never use INSERT, UPDATE, DELETE, DROP, or any destructive operations.\n"
"Always use the exact table name 'data'.\n"
"For date columns, use SQLite date functions (strftime, date())\n"
"For aggregations, always include ORDER BY to make results meaningful\n"
"Limit results to 20 rows maximum unless the question asks for all data\n"
"Always alias aggregated columns with readable names (e.g. SUM(revenue) AS total_revenue)\n"
"Return ONLY the SQL query, no explanation, no markdown, no backticks."
)
user_prompt = (
f"Table schema:\n"
f"Table name: data\n"
f"Columns:\n"
f"{columns_str}\n\n"
f"Question: {question}\n\n"
f"Write a SQLite SELECT query to answer this question."
)
print("[LLMService] Sending generation request to Groq...")
completion = client.chat.completions.create(
model=config.GROQ_MODEL,
messages=[
{"role": "system", "content": system_prompt},
{"role": "user", "content": user_prompt}
],
temperature=0.0
)
raw_response = completion.choices[0].message.content
sql = self._clean_sql_response(raw_response)
print(f"[LLMService] Generated SQL:\n{sql}")
return {
"sql": sql,
"raw_response": raw_response
}
except Exception as e:
print(f"[LLMService] Error generating SQL: {e}")
raise ValueError(f"Failed to generate SQL: {str(e)}")
def fix_sql(self, original_sql: str, error: str, schema: dict) -> Dict[str, str]:
"""
Takes a failed SQL query, the SQLite execution error, and the database schema,
and requests the LLM to return a corrected query.
"""
print(f"[LLMService] Fixing SQL query: '{original_sql}' due to error: '{error}'")
try:
client = self._get_client()
# Format the columns section of the prompt
columns_prompt = []
for col in schema.get("columns", []):
name = col.get("name")
col_type = col.get("type")
sample_vals = col.get("sample_values", [])
columns_prompt.append(f"- {name} ({col_type}): examples: {sample_vals}")
columns_str = "\n".join(columns_prompt)
system_prompt = (
"You are an expert SQL analyst. Correct the failing SQL query.\n"
"Generate SQLite-compatible SELECT queries only.\n"
"Never use INSERT, UPDATE, DELETE, DROP, or any destructive operations.\n"
"Always use the exact table name 'data'.\n"
"Return ONLY the corrected SQL query, no explanation, no markdown, no backticks."
)
user_prompt = (
f"Table schema:\n"
f"Table name: data\n"
f"Columns:\n"
f"{columns_str}\n\n"
f"This SQL query failed:\n"
f"{original_sql}\n\n"
f"Error details:\n"
f"{error}\n\n"
f"Fix the query and write a correct SQLite SELECT query."
)
print("[LLMService] Sending fix request to Groq...")
completion = client.chat.completions.create(
model=config.GROQ_MODEL,
messages=[
{"role": "system", "content": system_prompt},
{"role": "user", "content": user_prompt}
],
temperature=0.0
)
raw_response = completion.choices[0].message.content
fixed_sql = self._clean_sql_response(raw_response)
print(f"[LLMService] Fixed SQL:\n{fixed_sql}")
return {
"sql": fixed_sql
}
except Exception as e:
print(f"[LLMService] Error fixing SQL: {e}")
raise ValueError(f"Failed to fix SQL: {str(e)}")
def generate_question_suggestions(self, schema: dict) -> list:
"""
Generates 6 analytical questions that a business user might ask about the table dataset.
"""
print("[LLMService] Generating question suggestions from schema...")
try:
client = self._get_client()
# Format the columns section
columns_prompt = []
for col in schema.get("columns", []):
name = col.get("name")
col_type = col.get("type")
sample_vals = col.get("sample_values", [])
columns_prompt.append(f"- {name} ({col_type}): examples: {sample_vals}")
columns_str = "\n".join(columns_prompt)
system_prompt = (
"You are an expert data analyst assistant. Generate interesting question suggestions for a database schema.\n"
"Return ONLY a JSON array of strings, nothing else. No explanation, no markdown code block, no backticks."
)
user_prompt = (
f"Given a table with these columns:\n"
f"{columns_str}\n\n"
f"generate 6 interesting analytical questions a business user might ask.\n"
f"Return ONLY a JSON array of strings, nothing else."
)
completion = client.chat.completions.create(
model=config.GROQ_MODEL,
messages=[
{"role": "system", "content": system_prompt},
{"role": "user", "content": user_prompt}
],
temperature=0.5
)
raw_response = completion.choices[0].message.content.strip()
print(f"[LLMService] Raw suggestions response: {raw_response}")
# Strip markdown block wraps if present
cleaned = raw_response
if cleaned.startswith("```"):
lines = cleaned.split("\n")
content_lines = [line for line in lines if not line.strip().startswith("```")]
cleaned = "".join(content_lines).strip()
cleaned = cleaned.strip("`").strip()
import json
suggestions = json.loads(cleaned)
if isinstance(suggestions, list) and len(suggestions) >= 6:
return suggestions[:6]
else:
raise ValueError("Response is not a valid list of 6 suggestions.")
except Exception as e:
print(f"[LLMService] Error generating suggestions: {e}. Falling back to default list.")
# safe fallback suggestions
return [
"What is the total number of records?",
"Show the first 10 rows of data.",
"Summary statistics of numeric columns.",
"How are the categories distributed?",
"Which column values have the highest values?",
"Filter and search records."
]
def generate_sql_schema(self, question: str, schema_info: dict) -> dict:
"""
Generates SQLite SELECT query using JOINs based on the user question and database relational schema.
"""
print(f"[LLMService] Generating SQL (schema mode) for question: '{question}'")
try:
client = self._get_client()
# Format database schema tables
tables_prompt = []
for tbl in schema_info.get("tables", []):
tbl_name = tbl.get("name")
tbl_rows = tbl.get("row_count", 0)
columns_list = []
for col in tbl.get("columns", []):
name = col.get("name")
col_type = col.get("type")
is_pk = col.get("is_primary_key", False)
is_fk = col.get("is_foreign_key", False)
ref_tbl = col.get("references_table")
ref_col = col.get("references_column")
pk_str = " [PRIMARY KEY]" if is_pk else ""
fk_str = f" [FK → {ref_tbl}.{ref_col}]" if is_fk and ref_tbl and ref_col else ""
columns_list.append(f" - {name} ({col_type}){pk_str}{fk_str}")
columns_str = "\n".join(columns_list)
# Fetch sample rows and format as clean text
sample_vals = []
col_samples_len = max([len(c.get("sample_values", [])) for c in tbl.get("columns", [])]) if tbl.get("columns") else 0
for r_idx in range(min(col_samples_len, 3)):
row_val = {}
for col in tbl.get("columns", []):
samples = col.get("sample_values", [])
if r_idx < len(samples):
row_val[col["name"]] = samples[r_idx]
sample_vals.append(row_val)
tables_prompt.append(
f"Table: {tbl_name} ({tbl_rows} rows)\n"
f"Columns:\n"
f"{columns_str}\n"
f"Sample data (first 3 rows): {str(sample_vals)}"
)
tables_str = "\n\n".join(tables_prompt)
# Format relationships
rel_prompt = []
for rel in schema_info.get("relationships", []):
from_table = rel.get("from_table")
from_column = rel.get("from_column")
to_table = rel.get("to_table")
to_column = rel.get("to_column")
cardinality = rel.get("cardinality", "1:N")
from_participation = rel.get("from_participation", "total")
to_participation = rel.get("to_participation", "partial")
rel_prompt.append(
f" {from_table}.{from_column} → {to_table}.{to_column}\n"
f" Type: {cardinality} | {from_table} participation: {from_participation} | {to_table} participation: {to_participation}"
)
relationships_str = "\n".join(rel_prompt)
system_prompt = (
"You are an expert SQL analyst working with a relational SQLite database.\n"
"Generate SELECT queries only. Never use INSERT, UPDATE, DELETE, DROP, or ALTER.\n"
"Use JOINs when the question requires data from multiple tables.\n"
"Use meaningful table aliases: c for customers, o for orders, "
"p for products, oi for order_items, r for reviews.\n"
"Always alias aggregated columns with readable names.\n"
"Limit results to 20 rows unless the question specifically asks for all.\n"
"Return ONLY the raw SQL query — no explanation, no markdown, no backticks."
)
user_prompt = (
f"Database schema:\n\n"
f"{tables_str}\n\n"
f"Relationships:\n"
f"{relationships_str}\n\n"
f"Question: {question}\n\n"
f"Write a SQLite SELECT query using JOINs as needed."
)
print("[LLMService] Sending generation request to Groq...")
completion = client.chat.completions.create(
model=config.GROQ_MODEL,
messages=[
{"role": "system", "content": system_prompt},
{"role": "user", "content": user_prompt}
],
temperature=0.0
)
raw_response = completion.choices[0].message.content
sql = self._clean_sql_response(raw_response)
print(f"[LLMService] Generated Schema SQL:\n{sql}")
return {
"sql": sql,
"raw_response": raw_response
}
except Exception as e:
print(f"[LLMService] Error generating Schema SQL: {e}")
raise ValueError(f"Failed to generate SQL for schema database: {str(e)}")
def fix_sql_schema(self, original_sql: str, error: str, schema_info: dict) -> dict:
"""
Takes a failing schema-mode SQL query, the execution error, and the schema info,
and requests the LLM to correct the query.
"""
print(f"[LLMService] Fixing SQL (schema mode): '{original_sql}' due to error: '{error}'")
try:
client = self._get_client()
# Format database schema tables
tables_prompt = []
for tbl in schema_info.get("tables", []):
tbl_name = tbl.get("name")
tbl_rows = tbl.get("row_count", 0)
columns_list = []
for col in tbl.get("columns", []):
name = col.get("name")
col_type = col.get("type")
is_pk = col.get("is_primary_key", False)
is_fk = col.get("is_foreign_key", False)
ref_tbl = col.get("references_table")
ref_col = col.get("references_column")
pk_str = " [PRIMARY KEY]" if is_pk else ""
fk_str = f" [FK → {ref_tbl}.{ref_col}]" if is_fk and ref_tbl and ref_col else ""
columns_list.append(f" - {name} ({col_type}){pk_str}{fk_str}")
columns_str = "\n".join(columns_list)
col_samples_len = max([len(c.get("sample_values", [])) for c in tbl.get("columns", [])]) if tbl.get("columns") else 0
sample_vals = []
for r_idx in range(min(col_samples_len, 3)):
row_val = {}
for col in tbl.get("columns", []):
samples = col.get("sample_values", [])
if r_idx < len(samples):
row_val[col["name"]] = samples[r_idx]
sample_vals.append(row_val)
tables_prompt.append(
f"Table: {tbl_name} ({tbl_rows} rows)\n"
f"Columns:\n"
f"{columns_str}\n"
f"Sample data (first 3 rows): {str(sample_vals)}"
)
tables_str = "\n\n".join(tables_prompt)
# Format relationships
rel_prompt = []
for rel in schema_info.get("relationships", []):
from_table = rel.get("from_table")
from_column = rel.get("from_column")
to_table = rel.get("to_table")
to_column = rel.get("to_column")
cardinality = rel.get("cardinality", "1:N")
from_participation = rel.get("from_participation", "total")
to_participation = rel.get("to_participation", "partial")
rel_prompt.append(
f" {from_table}.{from_column} → {to_table}.{to_column}\n"
f" Type: {cardinality} | {from_table} participation: {from_participation} | {to_table} participation: {to_participation}"
)
relationships_str = "\n".join(rel_prompt)
system_prompt = (
"You are an expert SQL analyst. Correct the failing SQLite SQL query.\n"
"Generate SELECT queries only. Never use INSERT, UPDATE, DELETE, DROP, or ALTER.\n"
"Return ONLY the corrected SQL query, no explanation, no markdown, no backticks."
)
user_prompt = (
f"Database schema:\n\n"
f"{tables_str}\n\n"
f"Relationships:\n"
f"{relationships_str}\n\n"
f"This SQL query failed:\n"
f"{original_sql}\n\n"
f"Error details:\n"
f"{error}\n\n"
f"Fix the query and write a correct SQLite SELECT query."
)
print("[LLMService] Sending fix request to Groq...")
completion = client.chat.completions.create(
model=config.GROQ_MODEL,
messages=[
{"role": "system", "content": system_prompt},
{"role": "user", "content": user_prompt}
],
temperature=0.0
)
raw_response = completion.choices[0].message.content
fixed_sql = self._clean_sql_response(raw_response)
print(f"[LLMService] Fixed Schema SQL:\n{fixed_sql}")
return {
"sql": fixed_sql
}
except Exception as e:
print(f"[LLMService] Error fixing Schema SQL: {e}")
raise ValueError(f"Failed to fix SQL: {str(e)}")
# Singleton instance
llm_service = LLMService()
|