Spaces:
Running
Running
File size: 24,266 Bytes
0a4529c |
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 |
# DEPENDENCIES
from typing import List
from typing import Dict
from typing import Optional
from config.models import PromptType
from config.settings import get_settings
from config.models import ChunkWithScore
from config.logging_config import get_logger
from utils.error_handler import handle_errors
from generation.token_manager import TokenManager
from utils.error_handler import PromptBuildingError
# Setup Settings and Logging
settings = get_settings()
logger = get_logger(__name__)
class PromptBuilder:
"""
Intelligent prompt building for LLM generation: Constructs optimized prompts for different task types with proper context management and citation handling
"""
def __init__(self, model_name: str = None):
"""
Initialize prompt builder
Arguments:
----------
model_name { str } : Model name for token management
"""
self.logger = logger
self.settings = get_settings()
self.model_name = model_name or self.settings.OLLAMA_MODEL
self.token_manager = TokenManager(model_name)
# Prompt templates for different tasks
self.prompt_templates = {PromptType.QA : {"system": self._get_qa_system_prompt(), "user": self._get_qa_user_template(), "max_context_ratio": 0.6},
PromptType.SUMMARY : {"system": self._get_summary_system_prompt(), "user": self._get_summary_user_template(), "max_context_ratio": 0.8,},
PromptType.ANALYTICAL : {"system": self._get_analytical_system_prompt(), "user": self._get_analytical_user_template(), "max_context_ratio": 0.7},
PromptType.COMPARISON : {"system": self._get_comparison_system_prompt(), "user": self._get_comparison_user_template(), "max_context_ratio": 0.5},
PromptType.EXTRACTION : {"system": self._get_extraction_system_prompt(), "user": self._get_extraction_user_template(), "max_context_ratio": 0.6},
PromptType.CREATIVE : {"system": self._get_creative_system_prompt(), "user": self._get_creative_user_template(), "max_context_ratio": 0.4},
PromptType.CONVERSATIONAL : {"system": self._get_conversational_system_prompt(), "user": self._get_conversational_user_template(), "max_context_ratio": 0.5}
}
def build_prompt(self, query: str, context: str, sources: List[ChunkWithScore], prompt_type: PromptType = PromptType.QA, include_citations: bool = True,
max_completion_tokens: int = 1000) -> Dict[str, str]:
"""
Build complete prompt for LLM generation
Arguments:
----------
query { str } : User query
context { str } : Retrieved context
sources { list } : Source chunks
prompt_type { PromptType } : Type of prompt to build
include_citations { bool } : Whether to include citation instructions
max_completion_tokens { int } : Tokens to reserve for completion
Returns:
--------
{ dict } : Dictionary with 'system' and 'user' prompts
"""
if not query or not context:
raise PromptBuildingError("Query and context cannot be empty")
try:
# Get template for prompt type
template = self.prompt_templates.get(prompt_type, self.prompt_templates[PromptType.QA])
# Optimize context to fit within token limits
optimized_context = self._optimize_context(context = context,
template = template,
max_completion_tokens = max_completion_tokens,
)
# Build system prompt
system_prompt = self._build_system_prompt(template = template,
include_citations = include_citations,
prompt_type = prompt_type,
)
# Build user prompt
user_prompt = self._build_user_prompt(template = template,
query = query,
context = optimized_context,
sources = sources,
prompt_type = prompt_type,
)
# Validate token usage
self._validate_prompt_length(system_prompt = system_prompt,
user_prompt = user_prompt,
max_completion_tokens = max_completion_tokens,
)
self.logger.debug(f"Built {prompt_type.value} prompt: {self.token_manager.count_tokens(system_prompt + user_prompt)} tokens")
return {"system" : system_prompt,
"user" : user_prompt,
"metadata" : {"prompt_type" : prompt_type.value,
"context_tokens" : self.token_manager.count_tokens(optimized_context),
"total_tokens" : self.token_manager.count_tokens(system_prompt + user_prompt),
"sources_count" : len(sources),
}
}
except Exception as e:
self.logger.error(f"Prompt building failed: {repr(e)}")
raise PromptBuildingError(f"Prompt building failed: {repr(e)}")
def _optimize_context(self, context: str, template: Dict, max_completion_tokens: int) -> str:
"""
Optimize context to fit within token limits
Arguments:
----------
context { str } : Context text to optimize
template { Dict } : Prompt template
max_completion_tokens { int } : Tokens to reserve for completion
Returns:
--------
{ str } : Optimized context
"""
# Calculate tokens for system prompt
system_tokens = self.token_manager.count_tokens(template["system"])
# Estimate user template tokens by removing placeholders
user_template_clean = template["user"]
# Remove all known placeholders
placeholders_to_remove = ["{query}", "{context}", "{sources_info}", "{focus}"]
for placeholder in placeholders_to_remove:
user_template_clean = user_template_clean.replace(placeholder, "")
user_template_tokens = self.token_manager.count_tokens(user_template_clean)
# Calculate available tokens for context - Reserve: system + user template + completion + buffer
reserved_tokens = system_tokens + user_template_tokens + max_completion_tokens + 100
max_context_tokens = self.token_manager.context_window - reserved_tokens
# Ensure we don't exceed zero
if (max_context_tokens <= 0):
self.logger.warning(f"No tokens available for context. Reserved: {reserved_tokens}, Window: {self.token_manager.context_window}")
return ""
# Apply max context ratio from template config
ratio_limit = int(self.token_manager.context_window * template["max_context_ratio"])
max_context_tokens = min(max_context_tokens, ratio_limit)
# Log optimization details
self.logger.debug(f"Context optimization: max_tokens={max_context_tokens}, ratio={template['max_context_ratio']}")
# Truncate context to fit
optimized_context = self.token_manager.truncate_to_fit(context, max_context_tokens, strategy="end")
# Calculate reduction percentage
original_tokens = self.token_manager.count_tokens(context)
optimized_tokens = self.token_manager.count_tokens(optimized_context)
if (original_tokens > optimized_tokens):
reduction_pct = ((original_tokens - optimized_tokens) / original_tokens) * 100
self.logger.info(f"Context reduced by {reduction_pct:.1f}% ({original_tokens} → {optimized_tokens} tokens)")
return optimized_context
def _build_system_prompt(self, template: Dict, include_citations: bool, prompt_type: PromptType) -> str:
"""
Build system prompt
"""
system_prompt = template["system"]
# Add citation instructions if needed
if (include_citations and (prompt_type != PromptType.CREATIVE)):
citation_instructions = self._get_citation_instructions()
system_prompt += "\n\n" + citation_instructions
return system_prompt
def _build_user_prompt(self, template: Dict, query: str, context: str, sources: List[ChunkWithScore], prompt_type: PromptType) -> str:
"""
Build user prompt
"""
# Format sources information
sources_info = self._format_sources_info(sources) if sources else ""
# Build user prompt using template
user_prompt = template["user"].format(query = query,
context = context,
sources_info = sources_info,
)
# Add task-specific formatting
if (prompt_type == PromptType.COMPARISON):
user_prompt = self._enhance_comparison_prompt(user_prompt = user_prompt,
sources = sources,
)
elif (prompt_type == PromptType.ANALYTICAL):
user_prompt = self._enhance_analytical_prompt(user_prompt = user_prompt,
sources = sources,
)
return user_prompt
def _format_sources_info(self, sources: List[ChunkWithScore]) -> str:
"""
Format sources information for the prompt
"""
if not sources:
return ""
sources_list = list()
for i, source in enumerate(sources, 1):
chunk = source.chunk
source_info = f"Source [{i}]:"
if hasattr(chunk, 'metadata') and 'filename' in chunk.metadata:
source_info += f" {chunk.metadata['filename']}"
if chunk.page_number:
source_info += f" (page {chunk.page_number})"
if chunk.section_title:
source_info += f" - {chunk.section_title}"
sources_list.append(source_info)
return "Available sources:\n" + "\n".join(sources_list)
def _enhance_comparison_prompt(self, user_prompt: str, sources: List[ChunkWithScore]) -> str:
"""
Enhance prompt for comparison tasks
Arguments:
----------
user_prompt { str } : Base user prompt
sources { list } : Source chunks
Returns:
--------
{ str } : Enhanced prompt
"""
if (len(sources) < 2):
return user_prompt
enhancement = "\n\nPlease compare and contrast the information from different sources. "
enhancement += "Highlight agreements, disagreements, and complementary information. "
enhancement += "If sources conflict, present both perspectives clearly."
return user_prompt + enhancement
def _enhance_analytical_prompt(self, user_prompt: str, sources: List[ChunkWithScore]) -> str:
"""
Enhance prompt for analytical tasks - FIXED SIGNATURE
Arguments:
----------
user_prompt { str } : Base user prompt
sources { list } : Source chunks
Returns:
--------
{ str } : Enhanced prompt
"""
enhancement = "\n\nProvide analytical insights by:"
enhancement += "\n1. Identifying patterns and relationships in the information"
enhancement += "\n2. Analyzing implications and consequences"
enhancement += "\n3. Evaluating the strength of evidence from different sources"
enhancement += "\n4. Drawing well-reasoned conclusions"
return user_prompt + enhancement
def _validate_prompt_length(self, system_prompt: str, user_prompt: str, max_completion_tokens: int):
"""
Validate that prompt fits within context window
Arguments:
----------
system_prompt { str } : System prompt
user_prompt { str } : User prompt
max_completion_tokens { int } : Tokens needed for completion
Raises:
-------
PromptBuildingError : If prompt exceeds context window
"""
system_tokens = self.token_manager.count_tokens(system_prompt)
user_tokens = self.token_manager.count_tokens(user_prompt)
total_tokens = system_tokens + user_tokens
total_required = total_tokens + max_completion_tokens
if (total_required > self.token_manager.context_window):
error_msg = (f"Prompt exceeds context window:\n"
f"- System prompt: {system_tokens} tokens\n"
f"- User prompt: {user_tokens} tokens\n"
f"- Completion reserve: {max_completion_tokens} tokens\n"
f"- Total required: {total_required} tokens\n"
f"- Context window: {self.token_manager.context_window} tokens\n"
f"- Overflow: {total_required - self.token_manager.context_window} tokens"
)
self.logger.error(error_msg)
raise PromptBuildingError(error_msg)
# Log successful validation
utilization = (total_required / self.token_manager.context_window) * 100
self.logger.debug(f"Prompt validation passed: {total_required}/{self.token_manager.context_window} tokens ({utilization:.1f}% utilization)")
def _get_citation_instructions(self) -> str:
"""
Get citation instructions for system prompt
"""
return ("CITATION INSTRUCTIONS:\n"
"1. Always cite your sources using [number] notation\n"
"2. Use the citation number that corresponds to the source in the context\n"
"3. Cite sources for all factual claims and specific information\n"
"4. If information appears in multiple sources, cite the most relevant one\n"
"5. Do not make up information not present in the provided context\n"
"6. If the context doesn't contain the answer, explicitly state this"
)
# Template methods for different prompt types
def _get_qa_system_prompt(self) -> str:
return ("You are a precise and helpful AI assistant that answers questions based solely on the provided context.\n\n"
"Core Principles:\n"
"1. ONLY use information from the provided context\n"
"2. Be concise but complete - don't omit important details\n"
"3. Structure complex answers clearly\n"
"4. If information is ambiguous or conflicting, acknowledge this\n"
"5. Never make up or infer information not present in the context"
)
def _get_qa_user_template(self) -> str:
return ("Context Information:\n{context}\n\n"
"{sources_info}\n\n"
"Question: {query}\n\n"
"Instructions: Answer the question using ONLY the information provided in the context above. "
"Cite your sources using [number] notation. If the context doesn't contain enough information "
"to answer fully, state this clearly.\n\n"
"Answer:"
)
def _get_summary_system_prompt(self) -> str:
return ("You are a thorough AI assistant that provides comprehensive summaries based on the provided context.\n\n"
"Summary Guidelines:\n"
"1. Capture all key points and main ideas\n"
"2. Maintain the original meaning and intent\n"
"3. Organize information logically\n"
"4. Be comprehensive but concise\n"
"5. Highlight important findings and conclusions"
)
def _get_summary_user_template(self) -> str:
return ("Content to summarize:\n{context}\n\n"
"Please provide a comprehensive summary that captures all key points and main ideas. "
"Organize the summary logically and ensure it reflects the original content accurately.\n\n"
"Summary:"
)
def _get_analytical_system_prompt(self) -> str:
return ("You are an analytical AI assistant that provides insights based on the provided context.\n\n"
"Analytical Guidelines:\n"
"1. Analyze patterns and connections in the information\n"
"2. Compare different perspectives if multiple sources exist\n"
"3. Highlight key findings and implications\n"
"4. Identify gaps or limitations in the available information\n"
"5. Provide well-reasoned analysis and conclusions"
)
def _get_analytical_user_template(self) -> str:
return ("Context for analysis:\n{context}\n\n"
"{sources_info}\n\n"
"Analytical task: {query}\n\n"
"Please provide a detailed analysis based on the context above. "
"Identify patterns, relationships, and implications. "
"Cite sources for all analytical claims.\n\n"
"Analysis:"
)
def _get_comparison_system_prompt(self) -> str:
return ("You are an AI assistant that compares information across multiple sources.\n\n"
"Comparison Guidelines:\n"
"1. Identify similarities and differences between sources\n"
"2. Note if sources agree or disagree on specific points\n"
"3. Highlight complementary information\n"
"4. Present conflicting perspectives fairly\n"
"5. Draw conclusions about the overall consensus or disagreement"
)
def _get_comparison_user_template(self) -> str:
return ("Context from multiple sources:\n{context}\n\n"
"{sources_info}\n\n"
"Comparison task: {query}\n\n"
"Please compare how different sources address this topic. "
"Identify agreements, disagreements, and complementary information. "
"Cite specific sources for each point of comparison.\n\n"
"Comparison:"
)
def _get_extraction_system_prompt(self) -> str:
return ("You are a precise AI assistant that extracts specific information from context.\n\n"
"Extraction Guidelines:\n"
"1. Extract only the requested information\n"
"2. Be thorough and complete\n"
"3. Maintain accuracy and precision\n"
"4. Organize extracted information clearly\n"
"5. Cite sources for all extracted information"
)
def _get_extraction_user_template(self) -> str:
return ("Context:\n{context}\n\n"
"{sources_info}\n\n"
"Extraction task: {query}\n\n"
"Please extract the requested information from the context above. "
"Be thorough and precise. Cite sources for all extracted information.\n\n"
"Extracted Information:"
)
def _get_creative_system_prompt(self) -> str:
return ("You are a creative AI assistant that generates content based on provided context.\n\n"
"Creative Guidelines:\n"
"1. Use the context as inspiration and foundation\n"
"2. Be creative and engaging\n"
"3. Maintain coherence with the source material\n"
"4. You may extrapolate and build upon the provided information\n"
"5. Clearly distinguish between source-based content and creative additions"
)
def _get_creative_user_template(self) -> str:
return ("Context and inspiration:\n{context}\n\n"
"Creative task: {query}\n\n"
"Please create content based on the context above. "
"You may use the context as inspiration and build upon it creatively. "
"If you add information beyond what's in the context, make this clear.\n\n"
"Creative Response:"
)
def _get_conversational_system_prompt(self) -> str:
return ("You are a helpful and engaging conversational AI assistant.\n\n"
"Conversational Guidelines:\n"
"1. Be natural and engaging in conversation\n"
"2. Use the provided context to inform your responses\n"
"3. Maintain a friendly and helpful tone\n"
"4. Ask clarifying questions when needed\n"
"5. Cite sources when providing specific information"
)
def _get_conversational_user_template(self) -> str:
return ("Context for our conversation:\n{context}\n\n"
"Current message: {query}\n\n"
"Please respond naturally and helpfully based on the context above. "
"If providing specific information, cite your sources.\n\n"
"Response:"
)
# Global prompt builder instance
_prompt_builder = None
def get_prompt_builder(model_name: str = None) -> PromptBuilder:
"""
Get global prompt builder instance (singleton)
Arguments:
----------
model_name { str } : Model name for token management
Returns:
--------
{ PromptBuilder } : PromptBuilder instance
"""
global _prompt_builder
if _prompt_builder is None or (model_name and _prompt_builder.model_name != model_name):
_prompt_builder = PromptBuilder(model_name)
return _prompt_builder
@handle_errors(error_type = PromptBuildingError, log_error = True, reraise = False)
def build_qa_prompt(query: str, context: str, sources: List[ChunkWithScore], **kwargs) -> Dict[str, str]:
"""
Convenience function for building QA prompts
Arguments:
----------
query { str } : User query
context { str } : Retrieved context
sources { list } : Source chunks
**kwargs : Additional prompt building arguments
Returns:
--------
{ dict } : Dictionary with system and user prompts
"""
builder = get_prompt_builder()
return builder.build_prompt(query, context, sources, PromptType.QA, **kwargs) |