File size: 14,175 Bytes
ebd834c c65382c ebd834c a55ad28 ebd834c c99694c ebd834c c65382c a55ad28 f2228f6 a55ad28 c65382c 35b8855 1a596a8 35b8855 ebd834c a55ad28 ebd834c 9815e49 ebd834c 9815e49 ebd834c c65382c ebd834c c65382c 8f35c1b c65382c 8f35c1b c65382c ebd834c fbd18c8 ebd834c 9815e49 ebd834c c65382c ebd834c c99694c 1a596a8 c99694c a55ad28 ebd834c a55ad28 ebd834c 1330e5c ebd834c 1330e5c ebd834c 9815e49 ebd834c c65382c ebd834c 9815e49 ebd834c c65382c ebd834c 9815e49 9093d8f 9815e49 9093d8f c65382c a8fd301 c65382c 9093d8f 106354e 9093d8f ebd834c 9815e49 ebd834c c65382c ebd834c | 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 | """Query understanding layer: reformulates user queries for better retrieval."""
import json
import os
import re
from typing import Dict
from src.llm import generate
from src.prompts import (build_query_understanding_prompt,
build_categorizer_prompt, build_init_prompt, build_vc_analyzer_prompt)
def analyze_committment_to_vc(value_chain:str,
cfg:Dict,):
domain = cfg.get("chatbot", {}).get("domain", "research")
prompt = build_vc_analyzer_prompt(value_chain, domain)
try:
raw = generate(
"",
prompt,
cfg,
max_tokens=512,
)
return _parse_vc_analyzer_result(raw)
except Exception:
# LLM failure: fall back to raw query
return {
"value_chain": None,
"financial_commitment": {
"budget_start_year": "",
"budget_start_value": "",
"budget_end_year": "",
"budget_end_value": "",
"currency": "",
"responsible_institution": "",
"evidence": [
{
"quote": "",
"source_document": "",
"page_or_section": ""
}
]
},
"strategic_commitment": {
"documents_examined": 0,
"documents_discussing_value_chain": 0,
"coverage_weight": 0,
"development_plan": {
"value": "NOT_FOUND",
"evidence": []
},
"agricultural_strategy": {
"value": "NOT_FOUND",
"evidence": []
},
"standalone_vc_strategy": {
"value": "NOT_FOUND",
"evidence": []
},
"flagship_project": {
"value": "NOT_FOUND",
"evidence": []
}
},
"institutional_commitment": {
"dedicated_agency": {
"value": "NOT_FOUND",
"evidence": []
},
"dedicated_research_institute": {
"value": "NOT_FOUND",
"evidence": []
},
"dedicated_ministerial_department": {
"value": "NOT_FOUND",
"evidence": []
}
}
}
def _load_sql_schema_summary(sql_db_dir: str) -> str:
"""Load SQL schema summary from sql_schemas.json if it exists."""
schema_path = os.path.join(sql_db_dir, "sql_schemas.json")
if not os.path.exists(schema_path):
return ""
try:
with open(schema_path, "r", encoding="utf-8") as f:
schema = json.load(f)
from src.sql_retriever import build_schema_summary
return build_schema_summary(schema)
except Exception:
return ""
def categorize_query(user_query:str,
cfg:Dict,
conversation_history: list[dict] | None = None,):
domain = cfg.get("chatbot", {}).get("domain", "research")
cat_cfg = cfg.get("query_categorization", {})
max_history = cat_cfg.get("max_history", 6)
# Trim conversation history
history = conversation_history or []
if len(history) > max_history:
history = history[-max_history:]
prompt = build_categorizer_prompt(user_query, domain)
try:
raw = generate(
"You are a query-routing classifier. Return only JSON.",
prompt,
cfg,
max_tokens=512,
)
return _parse_qcat_result(raw)
except Exception:
# LLM failure: fall back to raw query
return {
"category": "complete",
"sub-action": "unresolved",
}
def init_query(user_query:str,
cfg:Dict,
conversation_history: list[dict] | None = None,):
domain = cfg.get("chatbot", {}).get("domain", "research")
cat_cfg = cfg.get("query_categorization", {})
max_history = cat_cfg.get("max_history", 6)
# Trim conversation history
history = conversation_history or []
if len(history) > max_history:
history = history[-max_history:]
prompt = build_init_prompt(user_query, domain)
try:
raw = generate(
"You are a query-routing classifier. Return only JSON.",
prompt,
cfg,
max_tokens=512,
)
return _parse_qinit_result(raw)
except Exception:
# LLM failure: fall back to raw query
return {
"country": None,
"value_chain": None,
"investment": None,
}
def understand_query(
user_query: str,
cfg: Dict,
conversation_history: list[Dict] | None = None,
) -> dict:
"""Analyze and reformulate a user query for better retrieval.
Uses the LLM to either:
- SEARCH: Rewrite the query into a search-optimized form
- CLARIFY: Ask the user a clarification question
Args:
user_query: Raw user input.
cfg: App config (provides domain, LLM settings).
conversation_history: Recent messages for pronoun/reference resolution.
Returns:
Dict with keys:
- ``action``: ``"search"`` or ``"clarify"``
- ``search_query``: Keyword-optimized query for retrieval (always present)
- ``display_query``: Clear natural-language question for response generation
- ``original_query``: The raw user input
- ``clarification_question``: Question to ask (only if action is "clarify")
"""
qu_cfg = cfg.get("query_understanding", {})
# Disabled: pass through raw query
if not qu_cfg.get("enabled", True):
return {
"action": "search",
"search_query": user_query,
"display_query": user_query,
"original_query": user_query,
"route": "vector",
"sql_query": None,
}
domain = cfg.get("chatbot", {}).get("domain", "research")
max_history = qu_cfg.get("max_history", 6)
# Trim conversation history
history = conversation_history or []
if len(history) > max_history:
history = history[-max_history:]
# Load SQL schema summary for prompt injection
sql_enabled = cfg.get("sql", {}).get("enabled", True)
schema_summary = ""
if sql_enabled:
sql_db_dir = cfg.get("paths", {}).get("sql_db", "sql_db")
if not os.path.isabs(sql_db_dir):
from pathlib import Path as _Path
project_root = _Path(__file__).resolve().parent.parent
sql_db_dir = os.path.join(str(project_root), sql_db_dir)
schema_summary = _load_sql_schema_summary(sql_db_dir)
# Load KB meta overview for routing awareness
from src.kb_meta import load_kb_meta
kb_overview = load_kb_meta(cfg)
prompt = build_query_understanding_prompt(
user_query, domain, history,
sql_schema_summary=schema_summary,
kb_overview=kb_overview,
)
try:
raw = generate(
"You are a query reformulation assistant. Return only JSON.",
prompt,
cfg,
max_tokens=512,
)
result = _parse_qu_result(raw, user_query)
except Exception:
# LLM failure: fall back to raw query
result = {
"action": "search",
"search_query": user_query,
"display_query": user_query,
"original_query": user_query,
"route": "vector",
"sql_query": None,
}
return result
def _parse_vc_analyzer_result(raw:str) -> Dict:
parsed = None
# Try direct parse
try:
parsed = json.loads(raw)
except (json.JSONDecodeError, TypeError):
pass
# Fallback: scan for the first valid JSON object using raw_decode
if parsed is None:
decoder = json.JSONDecoder()
for i, ch in enumerate(raw):
if ch == '{':
try:
parsed, _ = decoder.raw_decode(raw, i)
break
except json.JSONDecodeError:
continue
# Unparseable: fall back to raw query
if not isinstance(parsed, dict):
return {
"value_chain": None,
"financial_commitment": {
"budget_start_year": "",
"budget_start_value": "",
"budget_end_year": "",
"budget_end_value": "",
"currency": "",
"responsible_institution": "",
"evidence": [
{
"quote": "",
"source_document": "",
"page_or_section": ""
}
]
},
"strategic_commitment": {
"documents_examined": 0,
"documents_discussing_value_chain": 0,
"coverage_weight": 0,
"development_plan": {
"value": "NOT_FOUND",
"evidence": []
},
"agricultural_strategy": {
"value": "NOT_FOUND",
"evidence": []
},
"standalone_vc_strategy": {
"value": "NOT_FOUND",
"evidence": []
},
"flagship_project": {
"value": "NOT_FOUND",
"evidence": []
}
},
"institutional_commitment": {
"dedicated_agency": {
"value": "NOT_FOUND",
"evidence": []
},
"dedicated_research_institute": {
"value": "NOT_FOUND",
"evidence": []
},
"dedicated_ministerial_department": {
"value": "NOT_FOUND",
"evidence": []
}
}
}
return parsed
def _parse_qinit_result(raw:str) -> Dict:
parsed = None
# Try direct parse
try:
parsed = json.loads(raw)
except (json.JSONDecodeError, TypeError):
pass
# Fallback: scan for the first valid JSON object using raw_decode
if parsed is None:
decoder = json.JSONDecoder()
for i, ch in enumerate(raw):
if ch == '{':
try:
parsed, _ = decoder.raw_decode(raw, i)
break
except json.JSONDecodeError:
continue
# Unparseable: fall back to raw query
if not isinstance(parsed, dict) or "country" not in parsed:
return {
"country": None,
"value_chain": None,
"investment": None,
}
return parsed
def _parse_qcat_result(raw:str) -> Dict:
parsed = None
# Try direct parse
try:
parsed = json.loads(raw)
except (json.JSONDecodeError, TypeError):
pass
# Fallback: scan for the first valid JSON object using raw_decode
if parsed is None:
decoder = json.JSONDecoder()
for i, ch in enumerate(raw):
if ch == '{':
try:
parsed, _ = decoder.raw_decode(raw, i)
break
except json.JSONDecodeError:
continue
# Unparseable: fall back to raw query
if not isinstance(parsed, dict) or "category" not in parsed:
return {
"category": "general",
"action": "continue",
}
return parsed
def _parse_qu_result(raw: str, original_query: str) -> Dict:
"""Parse the LLM's JSON response into a structured result.
Falls back to raw query passthrough if parsing fails.
"""
parsed = None
# Try direct parse
try:
parsed = json.loads(raw)
except (json.JSONDecodeError, TypeError):
pass
# Fallback: scan for the first valid JSON object using raw_decode
if parsed is None:
decoder = json.JSONDecoder()
for i, ch in enumerate(raw):
if ch == '{':
try:
parsed, _ = decoder.raw_decode(raw, i)
break
except json.JSONDecodeError:
continue
# Unparseable: fall back to raw query
if not isinstance(parsed, dict) or "action" not in parsed:
return {
"action": "search",
"search_query": original_query,
"display_query": original_query,
"original_query": original_query,
"route": "vector",
"sql_query": None,
}
action = parsed.get("action", "search")
if action == "clarify":
return {
"action": "clarify",
"search_query": original_query, # fallback if clarification is skipped
"display_query": original_query, # fallback if clarification is skipped
"original_query": original_query,
"clarification_question": parsed.get(
"clarification_question", "Could you be more specific?"
),
"route": "vector",
"sql_query": None,
}
# action == "search"
search_query = parsed.get("search_query", original_query)
if not isinstance(search_query, str):
search_query = original_query
display_query = parsed.get("display_query", original_query)
if not isinstance(display_query, str):
display_query = original_query
route = parsed.get("route", "vector")
if isinstance(route, str):
route = route.strip().lower()
if route not in ("sql", "vector", "both"):
route = "vector"
sql_query = parsed.get("sql_query") or None
if sql_query is not None and not isinstance(sql_query, str):
sql_query = None
return {
"action": "search",
"search_query": search_query,
"display_query": display_query,
"original_query": original_query,
"route": route,
"sql_query": sql_query,
}
|