Spaces:
Running
Running
File size: 18,427 Bytes
c2ea5ed |
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 |
import logging
from typing import List, Dict, Any, Optional, Tuple
from agentgraph.shared.models.reference_based.entity import Entity
from agentgraph.shared.models.reference_based.relation import Relation
from agentgraph.shared.models.reference_based.content_reference import ContentReference
from agentgraph.input.text_processing.trace_line_processor import TraceLineNumberProcessor
logger = logging.getLogger(__name__)
# Sentinel delimiter used to concatenate multiple prompt snippets when more than one
# reference is resolved. We choose the Unicode "SYMBOL FOR UNIT SEPARATOR" (U+241F)
# which will never legitimately appear inside user-supplied prompt text, eliminating
# delimiter-collision issues seen with the previous "|||" sequence.
MULTI_SNIPPET_DELIMITER = "\u241F"
class ContentReferenceResolver:
"""
Service for resolving ContentReference objects to actual content from original traces.
This enables efficient content retrieval while maintaining position-based references.
"""
def __init__(self):
self.line_processor = TraceLineNumberProcessor()
def resolve_entity_prompts(self,
entities: List[Entity],
original_trace: str,
window_metadata: Dict[str, Any]) -> List[Entity]:
"""
Resolve ContentReference objects in entities to actual prompt content.
Args:
entities: List of Entity objects that may contain ContentReference objects
original_trace: Original trace content (without line numbers)
window_metadata: Metadata about the window including character positions
Returns:
List of Entity objects with resolved prompt content
"""
if not entities or not original_trace:
return entities
# CRITICAL FIX: Use the same character-to-line mapping approach as extraction
# This ensures ContentReferences point to the correct lines
numbered_content = self._create_extraction_compatible_numbering(original_trace)
resolved_entities = []
resolution_stats = {
"total_entities": len(entities),
"entities_with_refs": 0,
"successful_resolutions": 0,
"failed_resolutions": 0
}
for entity in entities:
resolved_entity = entity.model_copy() # Create a copy to avoid modifying original
# Check if entity has a content reference
if entity.raw_prompt_ref:
resolution_stats["entities_with_refs"] += 1
# Resolve the content reference
snippets, is_valid = self.line_processor.extract_content_by_reference(
numbered_content, entity.raw_prompt_ref
)
# Add detailed debug logging to track resolution process
logger.debug(f"Entity {entity.id} resolution debug:")
logger.debug(f" - raw_prompt_ref count: {len(entity.raw_prompt_ref)}")
for idx, ref in enumerate(entity.raw_prompt_ref):
logger.debug(f" - ref[{idx}]: L{ref.line_start}-L{ref.line_end}")
logger.debug(f" - extracted snippets count: {len(snippets) if snippets else 0}")
if snippets:
for idx, snippet in enumerate(snippets):
preview = snippet[:50].replace('\n', '\\n') if snippet else "EMPTY"
logger.debug(f" - snippet[{idx}]: {preview}...")
if snippets:
# Scrub any accidental occurrences of the delimiter inside the snippet
safe_snippets = [
s.replace(MULTI_SNIPPET_DELIMITER, " ") for s in snippets
]
# Concatenate snippets into a single string when multiple references exist
joined_prompt = (
safe_snippets[0]
if len(safe_snippets) == 1
else MULTI_SNIPPET_DELIMITER.join(safe_snippets)
)
resolved_entity.raw_prompt = joined_prompt
resolution_stats["successful_resolutions"] += 1
# Debug logging to check if line numbers are being removed
logger.debug(f"Resolved prompt for entity {entity.id}: {len(joined_prompt)} characters")
if '<L' in joined_prompt and '>' in joined_prompt:
logger.warning(f"Line numbers still present in resolved entity {entity.id}: {joined_prompt[:100]}...")
else:
logger.debug(f"Entity {entity.id} prompt is clean (no line numbers detected)")
if len(safe_snippets) > 1:
logger.debug(f" - joined with delimiter, split count will be: {len(safe_snippets)}")
else:
# Keep original prompt if resolution failed
resolution_stats["failed_resolutions"] += 1
logger.warning(f"Failed to resolve prompt reference for entity {entity.id}")
resolved_entities.append(resolved_entity)
logger.info(f"Entity prompt resolution stats: {resolution_stats}")
return resolved_entities
def resolve_relation_prompts(self,
relations: List[Relation],
original_trace: str,
window_metadata: Dict[str, Any]) -> List[Relation]:
"""
Resolve ContentReference objects in relations to actual interaction prompt content.
Args:
relations: List of Relation objects that may contain ContentReference objects
original_trace: Original trace content (without line numbers)
window_metadata: Metadata about the window including character positions
Returns:
List of Relation objects with resolved interaction prompt content
"""
if not relations or not original_trace:
return relations
numbered_content = self._create_extraction_compatible_numbering(original_trace)
resolved_relations = []
resolution_stats = {
"total_relations": len(relations),
"relations_with_refs": 0,
"successful_resolutions": 0,
"failed_resolutions": 0
}
for relation in relations:
resolved_relation = relation.model_copy() # Create a copy to avoid modifying original
# Check if relation has a content reference
if relation.interaction_prompt_ref:
resolution_stats["relations_with_refs"] += 1
# Resolve the content reference
snippets, is_valid = self.line_processor.extract_content_by_reference(
numbered_content, relation.interaction_prompt_ref
)
if snippets:
# Scrub any accidental occurrences of the delimiter inside the snippet
safe_snippets = [
s.replace(MULTI_SNIPPET_DELIMITER, " ") for s in snippets
]
# Concatenate snippets into a single string when multiple references exist
joined_prompt = (
safe_snippets[0]
if len(safe_snippets) == 1
else MULTI_SNIPPET_DELIMITER.join(safe_snippets)
)
resolved_relation.interaction_prompt = joined_prompt
resolution_stats["successful_resolutions"] += 1
# Debug logging to check if line numbers are being removed
logger.debug(f"Resolved interaction prompt for relation {relation.id}: {len(joined_prompt)} characters")
if '<L' in joined_prompt and '>' in joined_prompt:
logger.warning(f"Line numbers still present in resolved relation {relation.id}: {joined_prompt[:100]}...")
else:
logger.debug(f"Relation {relation.id} prompt is clean (no line numbers detected)")
else:
# Keep original prompt if resolution failed
resolution_stats["failed_resolutions"] += 1
logger.warning(f"Failed to resolve interaction prompt reference for relation {relation.id}")
resolved_relations.append(resolved_relation)
logger.info(f"Relation prompt resolution stats: {resolution_stats}")
return resolved_relations
def resolve_knowledge_graph_content(self,
knowledge_graph: Dict[str, Any],
original_trace: str,
window_metadata: Dict[str, Any]) -> Dict[str, Any]:
"""
Resolve all ContentReference objects in a knowledge graph to actual content.
Args:
knowledge_graph: Knowledge graph dictionary containing entities and relations
original_trace: Original trace content (without line numbers)
window_metadata: Metadata about the window including character positions
Returns:
Knowledge graph with resolved content references
"""
if not knowledge_graph or not original_trace:
return knowledge_graph
resolved_kg = knowledge_graph.copy()
# Resolve entity prompts
if "entities" in resolved_kg:
# Convert dict entities to Entity objects if needed
entities = []
for entity_data in resolved_kg["entities"]:
if isinstance(entity_data, dict):
entity = Entity(**entity_data)
else:
entity = entity_data
entities.append(entity)
resolved_entities = self.resolve_entity_prompts(entities, original_trace, window_metadata)
# Convert back to dict format
resolved_kg["entities"] = [entity.model_dump() for entity in resolved_entities]
# Resolve relation prompts
if "relations" in resolved_kg:
# Convert dict relations to Relation objects if needed
relations = []
for relation_data in resolved_kg["relations"]:
if isinstance(relation_data, dict):
relation = Relation(**relation_data)
else:
relation = relation_data
relations.append(relation)
resolved_relations = self.resolve_relation_prompts(relations, original_trace, window_metadata)
# Convert back to dict format
resolved_kg["relations"] = [relation.model_dump() for relation in resolved_relations]
# Add resolution metadata
if "metadata" not in resolved_kg:
resolved_kg["metadata"] = {}
resolved_kg["metadata"]["content_resolution"] = {
"resolved_at": self._get_current_timestamp(),
"original_trace_length": len(original_trace),
"resolution_method": "content_reference_resolver"
}
logger.info(f"Resolved content references for knowledge graph with {len(resolved_kg.get('entities', []))} entities and {len(resolved_kg.get('relations', []))} relations")
return resolved_kg
def validate_content_references(self,
content_refs: List[ContentReference],
original_trace: str) -> Dict[str, Any]:
"""
Validate a list of ContentReference objects against the original trace.
Args:
content_refs: List of ContentReference objects to validate
original_trace: Original trace content
Returns:
Validation report dictionary
"""
if not content_refs or not original_trace:
return {"valid_references": 0, "invalid_references": 0, "details": []}
# Use extraction-compatible numbering for validation
numbered_content = self._create_extraction_compatible_numbering(original_trace)
total_lines = len(original_trace.split('\n'))
validation_report = {
"total_references": len(content_refs),
"valid_references": 0,
"invalid_references": 0,
"details": []
}
for i, content_ref in enumerate(content_refs):
detail = {
"index": i,
"content_type": content_ref.content_type,
"line_range": f"{content_ref.line_start}-{content_ref.line_end}",
"is_valid": True,
"issues": []
}
# Check line range validity
if content_ref.line_start < 1 or content_ref.line_end < 1:
detail["is_valid"] = False
detail["issues"].append("Line numbers must be >= 1")
if content_ref.line_start > total_lines or content_ref.line_end > total_lines:
detail["is_valid"] = False
detail["issues"].append(f"Line numbers exceed total lines ({total_lines})")
if not content_ref.validate_line_range():
detail["is_valid"] = False
detail["issues"].append("line_end must be >= line_start")
# Try to extract content and validate
try:
extracted_content, content_valid = self.line_processor.extract_content_by_reference(
numbered_content, content_ref
)
if not content_valid:
detail["is_valid"] = False
detail["issues"].append("Content does not match summary")
except Exception as e:
detail["is_valid"] = False
detail["issues"].append(f"Extraction error: {str(e)}")
if detail["is_valid"]:
validation_report["valid_references"] += 1
else:
validation_report["invalid_references"] += 1
validation_report["details"].append(detail)
return validation_report
def _create_extraction_compatible_numbering(self, original_trace: str) -> str:
"""
Create numbered content using the same line numbering scheme as extraction.
This method replicates the character-to-line mapping logic from ChunkingService
to ensure ContentReferences resolve to the correct content.
Args:
original_trace: Original trace content (without line numbers)
Returns:
Content with line numbers that match extraction numbering
"""
# Step 1: Create character-to-line mapping (same as ChunkingService)
original_lines = original_trace.split('\n')
char_to_line_map = {}
char_pos = 0
for line_num, line in enumerate(original_lines, 1):
# Map every character in this line to this line number
for i in range(len(line) + 1): # +1 for newline
if char_pos + i < len(original_trace):
char_to_line_map[char_pos + i] = line_num
char_pos += len(line) + 1 # +1 for newline
# Step 2: Add line numbers to each line using its actual line number
numbered_lines = []
for line_num, line in enumerate(original_lines, 1):
numbered_line = f"<L{line_num}> {line}"
numbered_lines.append(numbered_line)
numbered_content = '\n'.join(numbered_lines)
logger.debug(f"Created extraction-compatible numbering for {len(original_lines)} lines")
return numbered_content
def _get_current_timestamp(self) -> str:
"""Get current timestamp in ISO format."""
from datetime import datetime
return datetime.now().isoformat()
def get_resolution_statistics(self,
knowledge_graph: Dict[str, Any]) -> Dict[str, Any]:
"""
Get statistics about content references in a knowledge graph.
Args:
knowledge_graph: Knowledge graph to analyze
Returns:
Statistics dictionary
"""
stats = {
"entities": {
"total": 0,
"with_references": 0,
"with_resolved_content": 0
},
"relations": {
"total": 0,
"with_references": 0,
"with_resolved_content": 0
}
}
# Analyze entities
if "entities" in knowledge_graph:
stats["entities"]["total"] = len(knowledge_graph["entities"])
for entity_data in knowledge_graph["entities"]:
if "raw_prompt_ref" in entity_data and entity_data["raw_prompt_ref"]:
stats["entities"]["with_references"] += 1
if "raw_prompt" in entity_data and entity_data["raw_prompt"]:
stats["entities"]["with_resolved_content"] += 1
# Analyze relations
if "relations" in knowledge_graph:
stats["relations"]["total"] = len(knowledge_graph["relations"])
for relation_data in knowledge_graph["relations"]:
if "interaction_prompt_ref" in relation_data and relation_data["interaction_prompt_ref"]:
stats["relations"]["with_references"] += 1
if "interaction_prompt" in relation_data and relation_data["interaction_prompt"]:
stats["relations"]["with_resolved_content"] += 1
return stats |