Spaces:
Sleeping
Sleeping
File size: 11,948 Bytes
cacd4d0 | 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 | """
Candidate and Feedback Collector for Presentation
This module collects all candidates generated during optimization along with
their feedback, scores, and metadata for presentation purposes.
"""
import json
from datetime import datetime
from pathlib import Path
from typing import Dict, List, Any, Optional
from dataclasses import dataclass, asdict, field
@dataclass
class CandidateInfo:
"""Information about a single candidate prompt"""
iteration: int
candidate_id: str
source: str # "GEPA_Reflection", "LLEGO_Crossover", "LLEGO_Mutation", "Seed"
prompt: str
score: Optional[float] = None
feedback: Optional[str] = None
feedback_details: Optional[Dict[str, Any]] = None
timestamp: str = field(default_factory=lambda: datetime.now().isoformat())
@dataclass
class IterationInfo:
"""Information about a single optimization iteration"""
iteration: int
candidates: List[CandidateInfo] = field(default_factory=list)
best_candidate: Optional[CandidateInfo] = None
best_score: Optional[float] = None
timestamp: str = field(default_factory=lambda: datetime.now().isoformat())
class CandidateCollector:
"""
Collects all candidates and feedback during optimization for presentation.
"""
def __init__(self, output_dir: str = "presentation_data"):
"""
Initialize the collector.
Args:
output_dir: Directory to save collected data
"""
self.output_dir = Path(output_dir)
self.output_dir.mkdir(exist_ok=True)
self.iterations: List[IterationInfo] = []
self.current_iteration: Optional[IterationInfo] = None
self.all_candidates: List[CandidateInfo] = []
# Track seed prompt
self.seed_prompt: Optional[str] = None
def set_seed_prompt(self, seed_prompt: str):
"""Set the seed prompt for reference"""
self.seed_prompt = seed_prompt
def start_iteration(self, iteration: int):
"""Start tracking a new iteration"""
self.current_iteration = IterationInfo(iteration=iteration)
self.iterations.append(self.current_iteration)
def add_candidate(
self,
iteration: int,
candidate_id: str,
source: str,
prompt: str,
score: Optional[float] = None,
feedback: Optional[str] = None,
feedback_details: Optional[Dict[str, Any]] = None
):
"""
Add a candidate to the collection.
Args:
iteration: Iteration number
candidate_id: Unique identifier for the candidate
source: Source of the candidate ("GEPA_Reflection", "LLEGO_Crossover", etc.)
prompt: The candidate prompt text
score: Evaluation score (if available)
feedback: Feedback text (if available)
feedback_details: Additional feedback details (if available)
"""
candidate = CandidateInfo(
iteration=iteration,
candidate_id=candidate_id,
source=source,
prompt=prompt,
score=score,
feedback=feedback,
feedback_details=feedback_details
)
# Add to current iteration
if self.current_iteration and self.current_iteration.iteration == iteration:
self.current_iteration.candidates.append(candidate)
# Update best candidate if this is better
if score is not None:
if (self.current_iteration.best_score is None or
score > self.current_iteration.best_score):
self.current_iteration.best_candidate = candidate
self.current_iteration.best_score = score
# Add to all candidates list
self.all_candidates.append(candidate)
def add_feedback(
self,
candidate_id: str,
feedback: str,
feedback_details: Optional[Dict[str, Any]] = None
):
"""
Add feedback to an existing candidate.
Args:
candidate_id: ID of the candidate to update
feedback: Feedback text
feedback_details: Additional feedback details
"""
for candidate in self.all_candidates:
if candidate.candidate_id == candidate_id:
candidate.feedback = feedback
candidate.feedback_details = feedback_details
break
# Also update in iterations
for iteration in self.iterations:
for candidate in iteration.candidates:
if candidate.candidate_id == candidate_id:
candidate.feedback = feedback
candidate.feedback_details = feedback_details
break
def add_score(
self,
candidate_id: str,
score: float
):
"""
Add score to an existing candidate.
Args:
candidate_id: ID of the candidate to update
score: Evaluation score
"""
for candidate in self.all_candidates:
if candidate.candidate_id == candidate_id:
candidate.score = score
break
# Also update in iterations
for iteration in self.iterations:
for candidate in iteration.candidates:
if candidate.candidate_id == candidate_id:
candidate.score = score
# Update best candidate if needed
if (iteration.best_score is None or score > iteration.best_score):
iteration.best_candidate = candidate
iteration.best_score = score
break
def save_to_json(self, filename: Optional[str] = None) -> Path:
"""
Save collected data to JSON file.
Args:
filename: Optional filename (auto-generated if not provided)
Returns:
Path to saved file
"""
if filename is None:
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
filename = f"candidates_and_feedback_{timestamp}.json"
filepath = self.output_dir / filename
data = {
"seed_prompt": self.seed_prompt,
"total_iterations": len(self.iterations),
"total_candidates": len(self.all_candidates),
"iterations": [asdict(iter_info) for iter_info in self.iterations],
"all_candidates": [asdict(candidate) for candidate in self.all_candidates],
"timestamp": datetime.now().isoformat()
}
with open(filepath, 'w', encoding='utf-8') as f:
json.dump(data, f, indent=2, ensure_ascii=False)
return filepath
def save_to_markdown(self, filename: Optional[str] = None) -> Path:
"""
Save collected data to Markdown file (presentation-ready format).
Args:
filename: Optional filename (auto-generated if not provided)
Returns:
Path to saved file
"""
if filename is None:
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
filename = f"candidates_and_feedback_{timestamp}.md"
filepath = self.output_dir / filename
with open(filepath, 'w', encoding='utf-8') as f:
# Header
f.write("# Optimization Candidates and Feedback\n\n")
f.write(f"**Generated:** {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}\n\n")
f.write(f"**Total Iterations:** {len(self.iterations)}\n")
f.write(f"**Total Candidates:** {len(self.all_candidates)}\n\n")
# Seed Prompt
if self.seed_prompt:
f.write("---\n\n")
f.write("## 🌱 Seed Prompt\n\n")
f.write("```\n")
f.write(self.seed_prompt)
f.write("\n```\n\n")
# Iterations
for iter_info in self.iterations:
f.write("---\n\n")
f.write(f"## 🔄 Iteration {iter_info.iteration}\n\n")
# Best candidate for this iteration
if iter_info.best_candidate:
f.write(f"### 🏆 Best Candidate (Score: {iter_info.best_score:.4f})\n\n")
f.write(f"**Source:** {iter_info.best_candidate.source}\n\n")
f.write(f"**Prompt:**\n```\n")
f.write(iter_info.best_candidate.prompt)
f.write("\n```\n\n")
if iter_info.best_candidate.feedback:
f.write(f"**Feedback:**\n\n")
f.write(f"{iter_info.best_candidate.feedback}\n\n")
# All candidates in this iteration
f.write(f"### 📝 All Candidates ({len(iter_info.candidates)})\n\n")
for idx, candidate in enumerate(iter_info.candidates, 1):
f.write(f"#### Candidate {idx}: {candidate.source}\n\n")
f.write(f"**ID:** `{candidate.candidate_id}`\n\n")
if candidate.score is not None:
f.write(f"**Score:** `{candidate.score:.4f}`\n\n")
f.write(f"**Prompt:**\n```\n")
f.write(candidate.prompt)
f.write("\n```\n\n")
if candidate.feedback:
f.write(f"**Feedback:**\n\n")
f.write(f"{candidate.feedback}\n\n")
if candidate.feedback_details:
f.write(f"**Feedback Details:**\n\n")
f.write("```json\n")
f.write(json.dumps(candidate.feedback_details, indent=2))
f.write("\n```\n\n")
f.write("---\n\n")
# Summary by source
f.write("---\n\n")
f.write("## 📊 Summary by Source\n\n")
sources = {}
for candidate in self.all_candidates:
if candidate.source not in sources:
sources[candidate.source] = []
sources[candidate.source].append(candidate)
for source, candidates in sources.items():
f.write(f"### {source} ({len(candidates)} candidates)\n\n")
for candidate in candidates:
score_str = f"Score: {candidate.score:.4f}" if candidate.score else "No score"
f.write(f"- **{candidate.candidate_id}** (Iteration {candidate.iteration}, {score_str})\n")
f.write("\n")
return filepath
def get_summary(self) -> Dict[str, Any]:
"""Get a summary of collected data"""
sources = {}
for candidate in self.all_candidates:
if candidate.source not in sources:
sources[candidate.source] = 0
sources[candidate.source] += 1
scored_candidates = [c for c in self.all_candidates if c.score is not None]
avg_score = sum(c.score for c in scored_candidates) / len(scored_candidates) if scored_candidates else None
return {
"total_iterations": len(self.iterations),
"total_candidates": len(self.all_candidates),
"candidates_by_source": sources,
"candidates_with_scores": len(scored_candidates),
"average_score": avg_score,
"candidates_with_feedback": len([c for c in self.all_candidates if c.feedback])
}
|