Spaces:
Sleeping
Sleeping
File size: 25,421 Bytes
e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 cb3e125 e2d9873 e2fa5a7 e2d9873 e2fa5a7 cb3e125 e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 e2d9873 e2fa5a7 e2d9873 e2fa5a7 ee8289e e2fa5a7 e2d9873 e2fa5a7 e2d9873 e2fa5a7 a6e4432 e2d9873 a6e4432 e2fa5a7 ee8289e e2d9873 e2fa5a7 e2d9873 7f34cf9 e2d9873 7f34cf9 e2d9873 7f34cf9 e2d9873 7f34cf9 e2d9873 7f34cf9 e2d9873 7f34cf9 e2d9873 7f34cf9 e2d9873 e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e cb3e125 e2fa5a7 cb3e125 e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 cb3e125 e2fa5a7 cb3e125 e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e cb3e125 e2fa5a7 cb3e125 e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 cb3e125 e2fa5a7 cb3e125 e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 a6e4432 ee8289e a6e4432 7b3dbf5 a6e4432 e2fa5a7 ee8289e a6e4432 e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 ee8289e e2fa5a7 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 | """
Multi-Stage Reasoning Engine for Agricultural Chatbot
======================================================
Implements the full spec: Claim → Validate → Contradict → Confirm pipeline
with priority-based context selection and evidence tracking.
"""
import json
import logging
from typing import Dict, List, Any, Optional, Tuple, Callable
from dataclasses import dataclass, field
from intent_classifier import IntentClassifier
from priority_mapper import PriorityContextMapper
from prompts import (
SYSTEM_PROMPT, CLAIM_PROMPT, VALIDATE_PROMPT,
CONTRADICT_PROMPT, CONFIRM_PROMPT, RESPONSE_PROMPT,
format_stage_prompt, build_context_prompt, generate_followup_questions,
# Compact prompts for token reduction
COMPACT_CLAIM_PROMPT, COMPACT_VALIDATE_PROMPT, COMPACT_CONTRADICT_PROMPT,
COMPACT_CONFIRM_PROMPT, COMPACT_RESPONSE_PROMPT,
build_compact_context, get_compact_prompt, format_minimal_diagnosis,
# Hybrid Prompts
FAST_LANE_PROMPT, DEEP_DIVE_HYPOTHESIS_PROMPT,
DEEP_DIVE_ADVERSARY_PROMPT, DEEP_DIVE_JUDGE_PROMPT
)
from context_aggregator import ContextAggregator
logger = logging.getLogger("ReasoningEngine")
# Toggle compact prompts to reduce token usage (saves ~50% tokens)
USE_COMPACT_PROMPTS = True
# =============================================================================
# DATA CLASSES
# =============================================================================
@dataclass
class StageResult:
"""Result from a single reasoning stage."""
stage: str
output: Dict[str, Any]
context_used: List[str]
confidence: float
raw_response: str = ""
@dataclass
class ReasoningResult:
"""Complete reasoning chain result."""
claim: StageResult
validation: StageResult
contradiction: StageResult
confirmation: StageResult
final_diagnosis: str
final_confidence: float
causal_chain: str
root_cause: str
symptoms: List[str]
recommendation: str
evidence_summary: Dict[str, List[str]]
# =============================================================================
# REASONING ENGINE
# =============================================================================
class ReasoningEngine:
"""
Multi-stage reasoning engine for agricultural chatbot.
Following the spec:
- Does NOT ingest all context at once
- Uses priority-based context selection per intent
- Actively seeks contradictions in Stage C
- Tracks evidence chain for transparency
"""
def __init__(self, llm_caller: Callable[[str], str]):
"""
Args:
llm_caller: Function that takes (prompt: str) -> str
"""
self.llm = llm_caller
self.intent_classifier = IntentClassifier()
self.priority_mapper = PriorityContextMapper()
self.aggregator = ContextAggregator()
def process_query(
self,
query: str,
context: Optional[Dict[str, Any]] = None
) -> Tuple[str, Dict[str, Any]]:
"""
Process user query through Hybrid Architecture (Fast Lane vs Deep Dive).
"""
logger.info(f"Processing query: {query[:50]}...")
# Stage 1: Classify intent
intent = self.intent_classifier.classify(query)
logger.info(f"Intent: {intent['primary_intent']} ({intent['confidence']})")
# Stage 2: Route Query
mode = self.route_query(query, intent)
logger.info(f"Routing mode: {mode}")
# Stage 3: Execute Logic
if mode == "FAST_LANE":
reasoning_result = self._execute_fast_lane(query, intent, context or {})
else:
reasoning_result = self._execute_deep_dive(query, intent, context or {})
# Stage 4: Generate response (pass full context for persona/weather/zone)
# Note: Fast Lane already generates action/diagnosis, but we standardize output format
response = self._generate_response(query, reasoning_result, context)
# Stage 5: Generate followups
followups = generate_followup_questions(
intent["primary_intent"],
reasoning_result.final_diagnosis
)
# Build complete trace
trace = self._build_trace(intent, reasoning_result, {}, followups)
trace["routing_mode"] = mode
return response, trace
def route_query(self, query: str, intent: Dict) -> str:
"""Decide between Fast Lane and Deep Dive."""
# Intention-based routing
fast_intents = ["vegetation_health", "water_stress", "nutrient_status"]
if intent["primary_intent"] in fast_intents and intent["confidence"] > 0.7:
return "FAST_LANE"
# "Why" questions or Comparisons usually need Deep Dive
if "compare" in query.lower() or "difference" in query.lower():
return "DEEP_DIVE"
return "DEEP_DIVE" # Default to robust mode for safety
def _execute_fast_lane(self, query: str, intent: Dict, context: Dict) -> ReasoningResult:
"""Execute 1-Shot Reasoning."""
logger.info("Executing FAST LANE (1-Call)...")
# Build ultra-compact context
compact_ctx = self.aggregator.build_ultra_compact_context(context)
# Include user query in prompt
prompt = FAST_LANE_PROMPT.format(query=query, context=compact_ctx)
full_prompt = f"{SYSTEM_PROMPT}\n\n{prompt}"
response = self.llm(full_prompt)
output = self._parse_json_safe(response, {
"reasoning_trace": "Analysis failed",
"diagnosis": "Unknown",
"confidence": 0.0,
"action": "Consult expert"
})
# Create Dummy StageResult for compatibility
dummy_stage = StageResult("fast_lane", output, [], output.get("confidence", 0.0))
return ReasoningResult(
claim=dummy_stage, # Fill for struct compatibility
validation=dummy_stage,
contradiction=dummy_stage,
confirmation=dummy_stage,
final_diagnosis=output.get("diagnosis", "Unknown"),
final_confidence=output.get("confidence", 0.0),
causal_chain=output.get("reasoning_trace", ""),
root_cause=output.get("diagnosis", "Unknown"),
symptoms=[],
recommendation=output.get("action", ""),
evidence_summary={"method": ["fast_lane_optimization"]}
)
def _execute_deep_dive(self, query: str, intent: Dict, context: Dict) -> ReasoningResult:
"""Execute 3-Stage Deep Dive."""
logger.info("Executing DEEP DIVE (3-Call)...")
# 1. Hypothesis Generation - Include query
ctx_hyp = self.aggregator.build_deep_dive_context(context, "hypothesis")
resp_hyp = self.llm(f"{SYSTEM_PROMPT}\n\nUSER QUERY: {query}\n\n{DEEP_DIVE_HYPOTHESIS_PROMPT.format(query=query, context=ctx_hyp)}")
out_hyp = self._parse_json_safe(resp_hyp, {"hypotheses": []})
# 2. Adversarial Check - Include query context
ctx_adv = self.aggregator.build_deep_dive_context(context, "adversary")
hyp_str = json.dumps(out_hyp, indent=2)
resp_adv = self.llm(f"{SYSTEM_PROMPT}\n\nUSER QUERY: {query}\n\n{DEEP_DIVE_ADVERSARY_PROMPT.format(query=query, hypotheses=hyp_str, context=ctx_adv)}")
out_adv = self._parse_json_safe(resp_adv, {"surviving_hypothesis": "Unknown"})
# 3. Final Verdict - Include query
ctx_judge = self.aggregator.build_deep_dive_context(context, "judge")
winner = out_adv.get("surviving_hypothesis", "Unknown")
resp_judge = self.llm(f"{SYSTEM_PROMPT}\n\nUSER QUERY: {query}\n\n{DEEP_DIVE_JUDGE_PROMPT.format(query=query, hypothesis=winner, context=ctx_judge)}")
out_judge = self._parse_json_safe(resp_judge, {"final_diagnosis": winner, "action_plan": {}})
# Map to ReasoningResult
# We map stages roughly to Maintain compatibility
result_hyp = StageResult("hypothesis", out_hyp, [], 0.0)
result_adv = StageResult("adversary", out_adv, [], 0.0)
result_judge = StageResult("judge", out_judge, [], 0.0)
return ReasoningResult(
claim=result_hyp,
validation=result_adv,
contradiction=result_adv,
confirmation=result_judge,
final_diagnosis=out_judge.get("final_diagnosis", "Unknown"),
final_confidence=0.9, # Deep dive implies high confidence
causal_chain=out_judge.get("detailed_reasoning", ""),
root_cause=out_judge.get("root_cause", ""),
symptoms=[],
recommendation=str(out_judge.get("action_plan", "")),
evidence_summary={"method": ["deep_dive_3_stage"]}
)
def _reason(
self,
query: str,
intent: Dict,
staged_context: Dict
) -> ReasoningResult:
"""Execute 4-stage reasoning pipeline as per spec."""
# Stage A: Initial Claim (Priority 1 context only)
logger.info("Stage A: Making initial claim...")
claim = self._stage_claim(query, staged_context["claim_context"])
# Stage B: Validate (Add Priority 2 context)
logger.info(f"Stage B: Validating hypothesis '{claim.output.get('hypothesis')}'...")
validation = self._stage_validate(
hypothesis=claim.output.get("hypothesis", "unknown"),
confidence=claim.confidence,
context=staged_context["validate_context"]
)
# Stage C: Contradict (Priority 3 - actively seek alternatives)
current_hypothesis = validation.output.get("hypothesis",
claim.output.get("hypothesis", "unknown"))
logger.info(f"Stage C: Seeking contradictions to '{current_hypothesis}'...")
contradiction = self._stage_contradict(
hypothesis=current_hypothesis,
confidence=validation.confidence,
context=staged_context["contradict_context"]
)
# Stage D: Confirm (Priority 4 - final decision)
logger.info("Stage D: Final confirmation...")
confirmation = self._stage_confirm(
hypothesis_1=current_hypothesis,
conf_1=validation.confidence,
hypothesis_2=contradiction.output.get("alternative_hypothesis", "none"),
conf_2=contradiction.output.get("alternative_confidence", 0),
context=staged_context["confirm_context"]
)
return ReasoningResult(
claim=claim,
validation=validation,
contradiction=contradiction,
confirmation=confirmation,
final_diagnosis=confirmation.output.get("final_diagnosis", "Undetermined"),
final_confidence=confirmation.confidence,
causal_chain=confirmation.output.get("causal_chain", ""),
root_cause=confirmation.output.get("root_cause", "unknown"),
symptoms=confirmation.output.get("symptoms", []),
recommendation=confirmation.output.get("recommendation", ""),
evidence_summary={
"primary": claim.context_used,
"supporting": validation.context_used,
"alternative": contradiction.context_used,
"validation": confirmation.context_used,
"supporting_evidence": confirmation.output.get("evidence_summary", {}).get("supporting", []),
"contradicting_evidence": confirmation.output.get("evidence_summary", {}).get("contradicting", []),
"inconclusive_evidence": confirmation.output.get("evidence_summary", {}).get("inconclusive", [])
}
)
def _stage_claim(self, query: str, context: Dict) -> StageResult:
"""Stage 3A: Make initial claim using Priority 1 context only."""
if USE_COMPACT_PROMPTS:
compact_ctx = build_compact_context(context) if isinstance(context, dict) else str(context)
prompt = COMPACT_CLAIM_PROMPT.format(query=query, context=compact_ctx)
else:
prompt = format_stage_prompt(
CLAIM_PROMPT,
query=query,
priority_1_context=context
)
full_prompt = f"{SYSTEM_PROMPT}\n\n{prompt}" if not USE_COMPACT_PROMPTS else prompt
response = self.llm(full_prompt)
output = self._parse_json_safe(response, {
"initial_claim": response[:200] if response else "No analysis available",
"hypothesis": "general_issue",
"evidence_cited": list(context.keys()),
"confidence": 0.5,
"uncertainties": ["Limited data available"]
})
return StageResult(
stage="claim",
output=output,
context_used=list(context.keys()),
confidence=output.get("confidence", 0.5),
raw_response=response
)
def _stage_validate(
self,
hypothesis: str,
confidence: float,
context: Dict
) -> StageResult:
"""Stage 3B: Validate hypothesis using Priority 2 context."""
if USE_COMPACT_PROMPTS:
compact_ctx = build_compact_context(context) if isinstance(context, dict) else str(context)
prompt = COMPACT_VALIDATE_PROMPT.format(
previous_hypothesis=hypothesis,
previous_confidence=confidence,
priority_2_context=compact_ctx
)
else:
prompt = format_stage_prompt(
VALIDATE_PROMPT,
previous_hypothesis=hypothesis,
previous_confidence=confidence,
priority_2_context=context
)
full_prompt = f"{SYSTEM_PROMPT}\n\n{prompt}" if not USE_COMPACT_PROMPTS else prompt
response = self.llm(full_prompt)
output = self._parse_json_safe(response, {
"validation_result": "neutral",
"confidence_updated": confidence,
"spatial_notes": "Unable to determine spatial distribution",
"new_evidence_summary": ""
})
# Carry forward hypothesis
output["hypothesis"] = hypothesis
return StageResult(
stage="validate",
output=output,
context_used=list(context.keys()),
confidence=output.get("confidence_updated", confidence),
raw_response=response
)
def _stage_contradict(
self,
hypothesis: str,
confidence: float,
context: Dict
) -> StageResult:
"""Stage 3C: Actively seek contradictions using Priority 3 context."""
if USE_COMPACT_PROMPTS:
compact_ctx = build_compact_context(context) if isinstance(context, dict) else str(context)
prompt = COMPACT_CONTRADICT_PROMPT.format(
hypothesis=hypothesis,
confidence=confidence,
priority_3_context=compact_ctx
)
else:
prompt = format_stage_prompt(
CONTRADICT_PROMPT,
hypothesis=hypothesis,
confidence=confidence,
priority_3_context=context
)
full_prompt = f"{SYSTEM_PROMPT}\n\n{prompt}" if not USE_COMPACT_PROMPTS else prompt
response = self.llm(full_prompt)
output = self._parse_json_safe(response, {
"contradiction_found": False,
"contradicting_evidence": [],
"alternative_hypothesis": "none",
"alternative_confidence": 0.0,
"reasoning": "No strong contradicting evidence found"
})
return StageResult(
stage="contradict",
output=output,
context_used=list(context.keys()),
confidence=output.get("alternative_confidence", 0.0),
raw_response=response
)
def _stage_confirm(
self,
hypothesis_1: str,
conf_1: float,
hypothesis_2: str,
conf_2: float,
context: Dict
) -> StageResult:
"""Stage 3D: Final confirmation using Priority 4 context."""
if USE_COMPACT_PROMPTS:
compact_ctx = build_compact_context(context) if isinstance(context, dict) else str(context)
prompt = COMPACT_CONFIRM_PROMPT.format(
hypothesis_1=hypothesis_1,
conf_1=conf_1,
hypothesis_2=hypothesis_2 if hypothesis_2 != "none" else "no_alt",
conf_2=conf_2,
priority_4_context=compact_ctx
)
else:
prompt = format_stage_prompt(
CONFIRM_PROMPT,
hypothesis_1=hypothesis_1,
conf_1=conf_1,
hypothesis_2=hypothesis_2 if hypothesis_2 != "none" else "no_alternative",
conf_2=conf_2,
priority_4_context=context
)
full_prompt = f"{SYSTEM_PROMPT}\n\n{prompt}" if not USE_COMPACT_PROMPTS else prompt
response = self.llm(full_prompt)
# Pick the more confident hypothesis as default
default_diagnosis = hypothesis_1 if conf_1 >= conf_2 else hypothesis_2
default_conf = max(conf_1, conf_2)
output = self._parse_json_safe(response, {
"final_diagnosis": default_diagnosis,
"confidence": default_conf,
"causal_chain": f"{default_diagnosis} leads to observed symptoms",
"root_cause": default_diagnosis,
"symptoms": [],
"evidence_summary": {
"supporting": list(context.keys()),
"contradicting": [],
"inconclusive": []
},
"recommendation": "Further investigation recommended based on available data"
})
return StageResult(
stage="confirm",
output=output,
context_used=list(context.keys()),
confidence=output.get("confidence", default_conf),
raw_response=response
)
def _generate_response(self, query: str, result: ReasoningResult,
context: Dict = None) -> str:
"""Generate final user-facing response with persona and context."""
diagnosis_data = {
"diagnosis": result.final_diagnosis,
"confidence": result.final_confidence,
"causal_chain": result.causal_chain,
"root_cause": result.root_cause,
"symptoms": result.symptoms,
"recommendation": result.recommendation
}
# Extract persona instructions
persona = context.get("persona", {}) if context else {}
persona_instructions = persona.get("instructions", "Provide clear, helpful farming advice.")
# Conversation history disabled to reduce token usage
history_text = ""
# Format zone context
zone_data = context.get("zone_analysis", {}) if context else {}
if zone_data and zone_data.get("priority_zones"):
zones = zone_data["priority_zones"]
zone_text = "PRIORITY ZONES:\n"
for i, z in enumerate(zones[:3], 1):
zone_text += f" {i}. {z.get('location', 'Zone')} - Stress: {z.get('stress_score', 0):.0%}\n"
else:
zone_text = "No zone-specific data available."
# Format trend context
trend_data = context.get("historical_trends", {}) if context else {}
if trend_data.get("summary"):
trend_text = trend_data["summary"]
else:
trend_text = "No historical trend data available."
# Format weather context
weather = context.get("weather", {}) if context else {}
if weather:
from prompts import format_weather_context
weather_text = format_weather_context(weather)
else:
weather_text = "No weather data available."
prompt = format_stage_prompt(
RESPONSE_PROMPT,
query=query,
diagnosis=json.dumps(diagnosis_data, indent=2),
evidence=json.dumps(result.evidence_summary, indent=2),
persona_instructions=persona_instructions,
conversation_history=history_text,
zone_context=zone_text,
trend_context=trend_text,
weather_context=weather_text
)
full_prompt = f"{SYSTEM_PROMPT}\n\n{prompt}"
response = self.llm(full_prompt)
return response
def _build_trace(
self,
intent: Dict,
result: ReasoningResult,
staged_context: Dict,
followups: List[str]
) -> Dict[str, Any]:
"""Build reasoning trace for transparency as per spec."""
return {
"intent_detected": intent["primary_intent"],
"intent_confidence": intent["confidence"],
"sub_intents": intent["sub_intents"],
"stages": {
"claim": {
"hypothesis": result.claim.output.get("hypothesis"),
"initial_claim": result.claim.output.get("initial_claim"),
"confidence": result.claim.confidence,
"context_used": result.claim.context_used,
"evidence_cited": result.claim.output.get("evidence_cited", [])
},
"validation": {
"result": result.validation.output.get("validation_result"),
"confidence": result.validation.confidence,
"spatial_notes": result.validation.output.get("spatial_notes"),
"context_used": result.validation.context_used
},
"contradiction": {
"found": result.contradiction.output.get("contradiction_found"),
"alternative": result.contradiction.output.get("alternative_hypothesis"),
"alternative_confidence": result.contradiction.output.get("alternative_confidence"),
"reasoning": result.contradiction.output.get("reasoning"),
"context_used": result.contradiction.context_used
},
"confirmation": {
"final": result.final_diagnosis,
"confidence": result.final_confidence,
"root_cause": result.root_cause,
"causal_chain": result.causal_chain,
"context_used": result.confirmation.context_used
}
},
"causal_chain": result.causal_chain,
"root_cause": result.root_cause,
"symptoms": result.symptoms,
"evidence_summary": result.evidence_summary,
"context_priority_used": {
"priority_1": list(staged_context.get("claim_context", {}).keys()),
"priority_2": list(staged_context.get("validate_context", {}).keys()),
"priority_3": list(staged_context.get("contradict_context", {}).keys()),
"priority_4": list(staged_context.get("confirm_context", {}).keys())
},
"suggested_followups": followups
}
def _parse_json_safe(self, text: str, default: Dict) -> Dict:
"""Safely extract and parse JSON from LLM response."""
if not text:
return default
text = text.strip()
# Look for JSON code block
if "```json" in text:
start = text.find("```json") + 7
end = text.find("```", start)
if end > start:
text = text[start:end].strip()
elif "```" in text:
start = text.find("```") + 3
end = text.find("```", start)
if end > start:
text = text[start:end].strip()
# Find JSON object
start = text.find("{")
end = text.rfind("}") + 1
if start >= 0 and end > start:
try:
return json.loads(text[start:end])
except json.JSONDecodeError as e:
logger.warning(f"JSON parse error: {e}")
return default
# =============================================================================
# SIMPLE REASONING (Fallback)
# =============================================================================
def simple_reason(query: str, context: Dict, llm_caller: Callable) -> str:
"""Simplified single-stage reasoning for when full pipeline isn't needed."""
context_str = build_context_prompt(context) if context else "No specific field data available."
prompt = f"""{SYSTEM_PROMPT}
User Query: {query}
Available Context:
{context_str}
Provide a helpful, actionable response. If specific data is available, cite it.
If not, provide general guidance based on the query.
Use this format:
1. Direct answer to the question
2. Key observations from data (if available)
3. Practical recommendations
4. What to monitor or check next"""
return llm_caller(prompt)
|