File size: 18,574 Bytes
4851501 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 |
import os
import asyncio
import json
from google import genai
from google.genai import types
from dotenv import load_dotenv
from backend.core.prompts import (
SYSTEM_INSTRUCTION,
INTENT_DETECTION_PROMPT,
DATA_DISCOVERY_PROMPT,
SQL_GENERATION_PROMPT,
EXPLANATION_PROMPT,
SPATIAL_SQL_PROMPT,
SPATIAL_SQL_PROMPT,
SQL_CORRECTION_PROMPT,
LAYER_NAME_PROMPT
)
class LLMGateway:
def __init__(self, model_name: str = "gemini-3-flash-preview"):
# Load environment variables if not already loaded
load_dotenv()
self.api_key = os.getenv("GEMINI_API_KEY") or os.getenv("GOOGLE_API_KEY")
if not self.api_key:
print("WARNING: GEMINI_API_KEY/GOOGLE_API_KEY not found. LLM features will not work.")
self.client = None
else:
# Explicitly setting the environment variable for the SDK if it's not set
if "GEMINI_API_KEY" not in os.environ and self.api_key:
os.environ["GEMINI_API_KEY"] = self.api_key
# The SDK automatically picks up GEMINI_API_KEY
self.client = genai.Client()
self.model = model_name
def _build_contents_from_history(self, history: list[dict], current_message: str) -> list:
"""
Converts conversation history to the format expected by the Gemini API.
History format: [{"role": "user"|"assistant", "content": "..."}]
"""
contents = []
for msg in history:
# Map 'assistant' to 'model' for Gemini API
role = "model" if msg["role"] == "assistant" else "user"
contents.append(
types.Content(
role=role,
parts=[types.Part.from_text(text=msg["content"])]
)
)
# Add the current message
contents.append(
types.Content(
role="user",
parts=[types.Part.from_text(text=current_message)]
)
)
return contents
async def generate_response_stream(self, user_query: str, history: list[dict] = None):
"""
Generates a streaming response using conversation history for context.
Yields chunks of text and thought summaries.
"""
if not self.client:
yield "I couldn't generate a response because the API key is missing."
return
if history is None:
history = []
try:
contents = self._build_contents_from_history(history, user_query)
# Enable thinking mode for general chat as well
config = types.GenerateContentConfig(
system_instruction=SYSTEM_INSTRUCTION,
thinking_config=types.ThinkingConfig(
include_thoughts=True # Enable thought summaries
)
)
stream = await asyncio.to_thread(
self.client.models.generate_content_stream,
model=self.model,
contents=contents,
config=config,
)
for chunk in stream:
for part in chunk.candidates[0].content.parts:
if part.thought:
yield {"type": "thought", "content": part.text}
elif part.text:
yield {"type": "content", "text": part.text}
except Exception as e:
print(f"Error calling Gemini stream: {e}")
yield f"Error: {str(e)}"
async def generate_response(self, user_query: str, history: list[dict] = None) -> str:
"""
Generates a response using conversation history for context.
"""
if not self.client:
return "I couldn't generate a response because the API key is missing."
if history is None:
history = []
try:
contents = self._build_contents_from_history(history, user_query)
config = types.GenerateContentConfig(
system_instruction=SYSTEM_INSTRUCTION,
)
response = await asyncio.to_thread(
self.client.models.generate_content,
model=self.model,
contents=contents,
config=config,
)
return response.text
except Exception as e:
print(f"Error calling Gemini: {e}")
return f"I encountered an error: {e}"
async def detect_intent(self, user_query: str, history: list[dict] = None) -> str:
"""
Detects the intent of the user's query using Gemini thinking mode.
Returns: GENERAL_CHAT, DATA_QUERY, MAP_REQUEST, SPATIAL_OP, or STAT_QUERY
"""
if not self.client:
return "GENERAL_CHAT"
intent_prompt = INTENT_DETECTION_PROMPT.format(user_query=user_query)
try:
# Use thinking mode for better intent classification
config = types.GenerateContentConfig(
thinking_config=types.ThinkingConfig(
thinking_level="medium" # Balanced thinking for intent detection
)
)
response = await asyncio.to_thread(
self.client.models.generate_content,
model=self.model,
contents=intent_prompt,
config=config,
)
intent = response.text.strip().upper()
# Validate the intent
if intent in ["GENERAL_CHAT", "DATA_QUERY", "MAP_REQUEST", "SPATIAL_OP", "STAT_QUERY"]:
return intent
# Default fallback
return "GENERAL_CHAT"
except Exception as e:
print(f"Error detecting intent: {e}")
return "GENERAL_CHAT"
async def stream_intent(self, user_query: str, history: list[dict] = None):
"""
Streams intent detection, yielding thoughts.
"""
if not self.client:
yield {"type": "error", "text": "API Key missing"}
return
intent_prompt = INTENT_DETECTION_PROMPT.format(user_query=user_query)
try:
config = types.GenerateContentConfig(
thinking_config=types.ThinkingConfig(
thinking_level="medium",
include_thoughts=True
)
)
stream = await asyncio.to_thread(
self.client.models.generate_content_stream,
model=self.model,
contents=intent_prompt,
config=config,
)
for chunk in stream:
for part in chunk.candidates[0].content.parts:
if part.thought:
yield {"type": "thought", "text": part.text}
elif part.text:
yield {"type": "content", "text": part.text}
except Exception as e:
print(f"Error detecting intent: {e}")
yield {"type": "error", "text": str(e)}
# Legacy generate_sql removed.
async def identify_relevant_tables(self, user_query: str, table_summaries: str) -> list[str]:
"""
Identifies which tables are relevant for the user's query from the catalog summary.
Returns a JSON list of table names.
"""
if not self.client:
return []
prompt = DATA_DISCOVERY_PROMPT.format(user_query=user_query, table_summaries=table_summaries)
try:
config = types.GenerateContentConfig(
response_mime_type="application/json"
)
response = await asyncio.to_thread(
self.client.models.generate_content,
model=self.model,
contents=prompt,
config=config,
)
text = response.text.replace("```json", "").replace("```", "").strip()
tables = json.loads(text)
return tables if isinstance(tables, list) else []
except Exception as e:
print(f"Error identifying tables: {e}")
return []
async def generate_analytical_sql(self, user_query: str, table_schema: str, history: list[dict] = None) -> str:
"""
Generates a DuckDB SQL query for analytical/statistical questions about geographic data.
This is the core of the text-to-SQL system.
"""
if not self.client:
return "-- Error: API Key missing"
prompt = SQL_GENERATION_PROMPT.format(table_schema=table_schema, user_query=user_query)
try:
# Use thinking mode for complex SQL generation
config = types.GenerateContentConfig(
temperature=1,
thinking_config=types.ThinkingConfig(
thinking_level="high" # Maximum reasoning for SQL generation
)
)
response = await asyncio.wait_for(
asyncio.to_thread(
self.client.models.generate_content,
model=self.model,
contents=prompt,
config=config,
),
timeout=120.0
)
sql = response.text.replace("```sql", "").replace("```", "").strip()
# Basic validation: must start with SELECT
if not sql.upper().strip().startswith("SELECT") and "-- ERROR" not in sql:
print(f"Warning: Generated SQL doesn't start with SELECT: {sql[:100]}")
if "SELECT" in sql.upper():
start_idx = sql.upper().find("SELECT")
sql = sql[start_idx:]
return sql
except asyncio.TimeoutError:
print("Gemini API call timed out after 30 seconds")
return "-- Error: API call timed out. Please try again."
except Exception as e:
print(f"Error calling Gemini for analytical SQL: {e}")
return f"-- Error generating SQL: {str(e)}"
async def stream_analytical_sql(self, user_query: str, table_schema: str, history: list[dict] = None):
"""
Streams the generation of DuckDB SQL, yielding thoughts and chunks.
"""
if not self.client:
yield {"type": "error", "text": "API Key missing"}
return
prompt = SQL_GENERATION_PROMPT.format(table_schema=table_schema, user_query=user_query)
try:
config = types.GenerateContentConfig(
temperature=1,
thinking_config=types.ThinkingConfig(
thinking_level="high",
include_thoughts=True
)
)
stream = await asyncio.to_thread(
self.client.models.generate_content_stream,
model=self.model,
contents=prompt,
config=config,
)
for chunk in stream:
for part in chunk.candidates[0].content.parts:
if part.thought:
yield {"type": "thought", "text": part.text}
elif part.text:
yield {"type": "content", "text": part.text}
except Exception as e:
print(f"Error streaming SQL: {e}")
yield {"type": "error", "text": str(e)}
async def stream_explanation(self, user_query: str, sql_query: str, data_summary: str, history: list[dict] = None):
"""
Streams the explanation.
"""
if not self.client:
yield {"type": "error", "text": "API Key missing"}
return
# Build context from history if available
context_str = ""
if history:
context_str = "Previous conversation context:\n"
for msg in history[-4:]: # Last 4 messages for context
context_str += f"- {msg['role']}: {msg['content'][:100]}...\n"
prompt = EXPLANATION_PROMPT.format(context_str=context_str, user_query=user_query, sql_query=sql_query, data_summary=data_summary)
try:
config = types.GenerateContentConfig(
system_instruction=SYSTEM_INSTRUCTION,
thinking_config=types.ThinkingConfig(
thinking_level="low",
include_thoughts=True
)
)
stream = await asyncio.to_thread(
self.client.models.generate_content_stream,
model=self.model,
contents=prompt,
config=config,
)
for chunk in stream:
for part in chunk.candidates[0].content.parts:
if part.thought:
yield {"type": "thought", "text": part.text}
elif part.text:
yield {"type": "content", "text": part.text}
except Exception as e:
print(f"Error generating explanation: {e}")
yield {"type": "error", "text": str(e)}
async def generate_explanation(self, user_query: str, sql_query: str, data_summary: str, history: list[dict] = None) -> str:
"""
Explains the results of the query to the user, maintaining conversation context.
"""
if not self.client:
return "I couldn't generate an explanation because the API key is missing."
# Build context from history if available
context_str = ""
if history:
context_str = "Previous conversation context:\n"
for msg in history[-4:]: # Last 4 messages for context
context_str += f"- {msg['role']}: {msg['content'][:100]}...\n"
prompt = EXPLANATION_PROMPT.format(context_str=context_str, user_query=user_query, sql_query=sql_query, data_summary=data_summary)
try:
config = types.GenerateContentConfig(
system_instruction=SYSTEM_INSTRUCTION,
thinking_config=types.ThinkingConfig(
thinking_level="low" # Fast response for explanations
)
)
response = await asyncio.to_thread(
self.client.models.generate_content,
model=self.model,
contents=prompt,
config=config,
)
return response.text
except Exception as e:
print(f"Error generating explanation: {e}")
return "Here are the results from the query."
async def generate_spatial_sql(self, user_query: str, layer_context: str, history: list[dict] = None) -> str:
"""
Generates a DuckDB Spatial SQL query for geometric operations on layers.
"""
if not self.client:
return "-- Error: API Key missing"
prompt = SPATIAL_SQL_PROMPT.format(layer_context=layer_context, user_query=user_query)
try:
config = types.GenerateContentConfig(
temperature=1,
)
# Add timeout to prevent indefinite hangs
response = await asyncio.wait_for(
asyncio.to_thread(
self.client.models.generate_content,
model=self.model,
contents=prompt,
config=config,
),
timeout=120.0
)
sql = response.text.replace("```sql", "").replace("```", "").strip()
return sql
except asyncio.TimeoutError:
print("Gemini API call timed out after 30 seconds")
return "-- Error: API call timed out. Please try again."
except Exception as e:
print(f"Error calling Gemini: {e}")
return f"-- Error generating SQL: {str(e)}"
async def correct_sql(self, user_query: str, incorrect_sql: str, error_message: str, schema_context: str) -> str:
"""
Corrects a failed SQL query based on the error message.
"""
if not self.client:
return "-- Error: API Key missing"
prompt = SQL_CORRECTION_PROMPT.format(
error_message=error_message,
incorrect_sql=incorrect_sql,
user_query=user_query,
schema_context=schema_context
)
try:
config = types.GenerateContentConfig(
temperature=1,
)
response = await asyncio.to_thread(
self.client.models.generate_content,
model=self.model,
contents=prompt,
config=config,
)
sql = response.text.replace("```sql", "").replace("```", "").strip()
return sql
except Exception as e:
print(f"Error correcting SQL: {e}")
return incorrect_sql
async def generate_layer_name(self, user_query: str, sql_query: str) -> dict:
"""
Generates a short, descriptive name, emoji, and point style for a map layer.
Returns: {"name": str, "emoji": str, "pointStyle": str | None}
"""
if not self.client:
return {"name": "New Layer", "emoji": "📍", "pointStyle": None}
prompt = LAYER_NAME_PROMPT.format(user_query=user_query, sql_query=sql_query)
try:
config = types.GenerateContentConfig(
temperature=1,
response_mime_type="application/json"
)
# Use simple generate content (not streaming)
response = await asyncio.to_thread(
self.client.models.generate_content,
model=self.model,
contents=prompt,
config=config,
)
result = json.loads(response.text)
return {
"name": result.get("name", "Map Layer"),
"emoji": result.get("emoji", "📍"),
"pointStyle": result.get("pointStyle", None)
}
except Exception as e:
print(f"Error generating layer name: {e}")
return {"name": "Map Layer", "emoji": "📍", "pointStyle": None}
|