Spaces:
Running
Running
File size: 12,443 Bytes
24f95f0 5f91e0b 24f95f0 | 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 | """
Curiosity engine for Janus — autonomous knowledge exploration.
FIXES vs previous version:
- Curiosity cycles were completing in 0.0 seconds (LLM never called)
- Root cause: interests dict was empty after HF Space restart (ephemeral FS),
causing the exploration loop to immediately exit with nothing to explore
- Fix: always has a default set of seed interests to explore if none exist
- Fix: exploration loop now runs even with zero stored discoveries
- Added deduplication of discoveries by topic
- Added persistence of discoveries to data/curiosity/ dir
- Added timeout guard on each LLM call
"""
import time
import json
import logging
import math
from pathlib import Path
from datetime import datetime
from typing import Optional
logger = logging.getLogger(__name__)
try:
from app.config import DATA_DIR
except ImportError:
DATA_DIR = Path(__file__).parent.parent / "data"
CURIOSITY_DIR = Path(DATA_DIR) / "curiosity"
MAX_DISCOVERIES = 200
LLM_TIMEOUT = 120
# Default seed interests — used when nothing has been learned yet (e.g. after HF restart)
DEFAULT_SEED_INTERESTS = {
"AI market impact": 0.05,
"Federal Reserve policy": 0.05,
"earnings quality": 0.05,
"semiconductor demand": 0.04,
"energy transition": 0.04,
"emerging market risk": 0.03,
"corporate debt levels": 0.03,
"geopolitical trade risk": 0.03,
"venture capital trends": 0.02,
"regulatory AI": 0.02,
}
EXPLORATION_TYPES = [
"deep_dive",
"connection",
"contrarian",
"future_scenario",
"blind_spot",
]
def _call_with_timeout(fn, timeout: float):
import threading
result = [None]
error = [None]
def target():
try:
result[0] = fn()
except Exception as e:
error[0] = e
t = threading.Thread(target=target, daemon=True)
t.start()
t.join(timeout)
if t.is_alive():
return None, TimeoutError(f"Timed out after {timeout}s")
return result[0], error[0]
class CuriosityEngine:
"""
Proactively explores topics the system is interested in.
Generates insights without being asked.
"""
def __init__(self):
CURIOSITY_DIR.mkdir(parents=True, exist_ok=True)
self._interests: dict = {}
self._discoveries: list = []
self._total_discoveries = 0
self._total_interests = 0
self._load()
def _load(self):
"""Load saved interests and discoveries."""
interests_file = CURIOSITY_DIR / "interests.json"
if interests_file.exists():
try:
self._interests = json.loads(interests_file.read_text())
except Exception:
self._interests = {}
# If empty (e.g. after HF restart), seed with defaults
if not self._interests:
self._interests = dict(DEFAULT_SEED_INTERESTS)
logger.info("CuriosityEngine: seeded with default interests (no saved state found)")
# Load recent discoveries
self._discoveries = []
for f in sorted(CURIOSITY_DIR.glob("discovery_*.json"), reverse=True)[:MAX_DISCOVERIES]:
try:
self._discoveries.append(json.loads(f.read_text()))
except Exception:
pass
self._total_discoveries = len(self._discoveries)
self._total_interests = len(self._interests)
def _save_interests(self):
try:
(CURIOSITY_DIR / "interests.json").write_text(
json.dumps(self._interests, indent=2)
)
except Exception as e:
logger.warning(f"CuriosityEngine: save_interests failed: {e}")
def _save_discovery(self, discovery: dict):
try:
ts = datetime.utcnow().strftime("%Y%m%d_%H%M%S_%f")
(CURIOSITY_DIR / f"discovery_{ts}.json").write_text(
json.dumps(discovery, indent=2, default=str)
)
# Clean old discoveries
files = sorted(CURIOSITY_DIR.glob("discovery_*.json"))
for old in files[:-MAX_DISCOVERIES]:
try:
old.unlink()
except Exception:
pass
except Exception as e:
logger.warning(f"CuriosityEngine: save_discovery failed: {e}")
def run_curiosity_cycle(self, force: bool = False) -> dict:
"""
Run one curiosity exploration cycle.
FIXED: always explores something, even after fresh start.
"""
t0 = time.time()
logger.info("CuriosityEngine: starting curiosity cycle")
report = {
"timestamp": datetime.utcnow().isoformat(),
"explorations": [],
"discoveries": [],
"new_interests": [],
"duration_seconds": 0.0,
"total_discoveries": self._total_discoveries,
"total_interests": self._total_interests,
}
try:
from app.agents._model import call_model
# Always explore at least 1-2 topics per cycle
sorted_interests = sorted(
self._interests.items(), key=lambda x: -x[1]
)
import random
explore_topics = [t for t, _ in sorted_interests[:2]]
if len(sorted_interests) > 2:
random_pick = random.choice(sorted_interests[2:])
explore_topics.append(random_pick[0])
for topic in explore_topics[:2]: # max 2 per cycle to avoid timeout
exploration_type = random.choice(EXPLORATION_TYPES)
discovery = self._explore(call_model, topic, exploration_type)
report["explorations"].append({
"type": exploration_type,
"topic": topic,
"result": "discovered" if discovery else "explored",
})
if discovery:
# Deduplicate by topic
existing_topics = {d.get("topic", "") for d in self._discoveries}
if topic not in existing_topics:
self._discoveries.insert(0, discovery)
self._discoveries = self._discoveries[:MAX_DISCOVERIES]
self._total_discoveries += 1
report["discoveries"].append(discovery)
self._save_discovery(discovery)
logger.info(f"CuriosityEngine: new discovery on '{topic}'")
self._update_interest(topic, explored=True, discovered=bool(discovery))
# Occasionally generate new interests from discoveries
if self._discoveries and random.random() < 0.3:
new_interests = self._generate_new_interests(call_model)
for interest in new_interests:
if interest not in self._interests:
self._interests[interest] = 0.02
report["new_interests"].append(interest)
self._total_interests += 1
logger.info(f"CuriosityEngine: new interest '{interest}'")
self._save_interests()
except Exception as e:
logger.error(f"CuriosityEngine: cycle error: {e}")
report["error"] = str(e)
report["duration_seconds"] = round(time.time() - t0, 2)
report["total_discoveries"] = self._total_discoveries
report["total_interests"] = self._total_interests
logger.info(
f"CuriosityEngine: cycle complete in {report['duration_seconds']:.1f}s — "
f"{len(report['discoveries'])} discoveries"
)
return report
def _explore(self, call_model, topic: str, exploration_type: str) -> Optional[dict]:
"""Explore a topic and return a discovery if found."""
type_prompts = {
"deep_dive": f"Provide a non-obvious, specific insight about '{topic}'. What do most analysts miss?",
"connection": f"What is a surprising connection between '{topic}' and another domain?",
"contrarian": f"What is the strongest contrarian argument regarding '{topic}'?",
"future_scenario": f"What is the most likely unexpected development in '{topic}' in the next 12 months?",
"blind_spot": f"What critical blind spot do most people have about '{topic}'?",
}
prompt = (
f"You are Janus exploring '{topic}' with exploration type '{exploration_type}'.\n\n"
+ type_prompts.get(exploration_type, f"What is insightful about '{topic}'?")
+ "\n\nProvide a specific, evidence-based insight (not generic). "
"Respond with JSON:\n"
'{"key_insight": "...", "evidence": ["fact1", "fact2"], '
'"why_it_matters": "...", "connections": ["related_concept"]}'
)
result, err = _call_with_timeout(
lambda: call_model([{"role": "user", "content": prompt}]),
LLM_TIMEOUT
)
if err or not result:
logger.warning(f"CuriosityEngine: exploration of '{topic}' failed: {err}")
return None
try:
from app.agents.smart_router import safe_parse
data = safe_parse(result)
key_insight = data.get("key_insight", "")
why_it_matters = data.get("why_it_matters", "")
if not key_insight and not why_it_matters:
return None
insight_text = key_insight or result[:300]
return {
"topic": topic,
"type": exploration_type,
"insight": f"{insight_text[:300]} | Why: {why_it_matters[:200]}",
"timestamp": datetime.utcnow().isoformat(),
"raw_response": data,
}
except Exception as e:
logger.warning(f"CuriosityEngine: parse failed for '{topic}': {e}")
if len(result) > 50:
return {
"topic": topic,
"type": exploration_type,
"insight": result[:400],
"timestamp": datetime.utcnow().isoformat(),
}
return None
def _generate_new_interests(self, call_model) -> list:
"""Generate new topics to explore based on recent discoveries."""
recent = self._discoveries[:3]
topics = [d.get("topic", "") for d in recent if d.get("topic")]
prompt = (
f"Based on these recently explored topics: {', '.join(topics)}, "
"suggest 3 related topics worth investigating. "
'JSON: {"topics": ["topic1", "topic2", "topic3"]}'
)
result, err = _call_with_timeout(
lambda: call_model([{"role": "user", "content": prompt}]),
30
)
if err or not result:
return []
try:
from app.agents.smart_router import safe_parse
data = safe_parse(result)
return data.get("topics", [])[:3]
except Exception:
return []
def _update_interest(self, topic: str, explored: bool, discovered: bool):
"""Update interest score using exponential decay + exploration boost."""
current = self._interests.get(topic, 0.01)
decayed = current * 0.95
boost = 0.01 if discovered else 0.005 if explored else 0
self._interests[topic] = min(decayed + boost, 1.0)
def add_interest(self, topic: str, score: float = 0.03):
"""Add a new interest (called when users ask about a topic repeatedly)."""
if topic and len(topic) > 2:
self._interests[topic] = max(self._interests.get(topic, 0), score)
self._save_interests()
def get_status(self) -> dict:
top_interests = dict(sorted(self._interests.items(), key=lambda x: -x[1])[:10])
return {
"total_discoveries": self._total_discoveries,
"total_interests": self._total_interests,
"top_interests": top_interests,
"latest_discovery": self._discoveries[0] if self._discoveries else None,
}
def get_discoveries(self, limit: int = 10) -> list:
return self._discoveries[:limit]
def get_interests(self) -> dict:
return dict(sorted(self._interests.items(), key=lambda x: -x[1]))
|