Spaces:
Sleeping
Sleeping
File size: 20,314 Bytes
8ed954c d5ba3a3 8ed954c d5ba3a3 8ed954c d5ba3a3 73c184e 65d81ac 109b51d 1358dec d5ba3a3 1358dec d5ba3a3 73c184e 645673f 8ed954c 3c57e36 73c184e 8ed954c ffc1e30 8ed954c a6f44ed 8ed954c 65d81ac 8ed954c 73c184e 6b7b1c1 8ed954c 73c184e 8ed954c 65d81ac 8ed954c 65d81ac 8ed954c ccd5d7c a2930e5 8ed954c 65d81ac d5ba3a3 8ed954c d5ba3a3 a2930e5 8ed954c d5ba3a3 109b51d 8ed954c 73c184e c010f04 8ed954c c010f04 8ed954c d5ba3a3 c010f04 65d81ac 8ed954c d5ba3a3 65d81ac 3c57e36 8ed954c d5ba3a3 3c57e36 8ed954c d5ba3a3 8ed954c 65d81ac d5ba3a3 8ed954c d5ba3a3 65d81ac 8ed954c 73c184e 8ed954c c010f04 8ed954c c010f04 e813ce3 645673f 8ed954c 3ee7903 8ed954c 645673f 3ee7903 645673f ffc1e30 645673f 8ed954c 3ee7903 a051253 ec278c4 8ed954c ec278c4 8ed954c 3ee7903 ec278c4 8ed954c 3ee7903 ec278c4 8ed954c ec278c4 73c184e 8ed954c 4438658 d5ba3a3 e813ce3 645673f 1358dec e813ce3 d5ba3a3 e813ce3 ffc1e30 8ed954c d5ba3a3 645673f ffc1e30 d5ba3a3 8ed954c 65d81ac 8ed954c 73c184e 8ed954c a2930e5 8ed954c 65d81ac 73c184e 8ed954c 73c184e 8ed954c 73c184e a051253 8ed954c 52fef82 73c184e 8ed954c d5ba3a3 8ed954c d5ba3a3 73c184e d5ba3a3 8ed954c 73c184e d5ba3a3 8ed954c 73c184e d5ba3a3 8ed954c 28012a0 645673f 8ed954c 645673f 65d81ac 645673f 8ed954c 645673f d5ba3a3 645673f 65d81ac 645673f | 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 | """Daily automated micro-cap hunter (GitHub Actions cron).
Pipeline per region: Scout -> Gatekeeper -> Analyst -> Email
An outer orchestrator graph dispatches all regions in parallel via the
LangGraph ``Send`` API, then collects results.
The Scout uses a two-pronged discovery approach:
1. yFinance screener for systematic micro-cap filtering
2. Brave Search for trending/momentum signals
3. Quantitative scoring to pick the best candidate
Both feeds are merged, scored, and only the top candidate proceeds to
the expensive LLM analyst step.
"""
import operator
import os
import signal
import time
import warnings
from typing import Annotated, Literal, TypedDict
warnings.filterwarnings("ignore", category=UserWarning, module="pydantic")
from langgraph.graph import StateGraph, START, END
from langgraph.checkpoint.memory import InMemorySaver
from langgraph.types import Command, RetryPolicy, Send
from src.llm import get_llm, get_structured_llm, invoke_with_fallback
from src.finance_tools import (
check_financial_health,
get_insider_sentiment,
get_company_news,
get_basic_financials,
)
from src.portfolio_tracker import record_paper_trade
from src.email_utils import send_email_report
from src.core.logger import get_logger
from src.core.search import brave_search
from src.core.ticker_utils import normalize_price, REGION_SUFFIXES
from src.core.memory import load_seen_tickers, mark_ticker_seen
from src.core.state import AgentState
from src.core.online_eval import log_online_feedback, tag_for_review, get_current_run_id
from src.discovery.screener import screen_microcaps, get_trending_tickers_from_brave
from src.discovery.scoring import rank_candidates
from src.discovery.insider_feed import get_insider_buys
logger = get_logger(__name__)
# --- CONFIGURATION ---
MAX_MARKET_CAP = 500_000_000
MIN_MARKET_CAP = 5_000_000
MAX_PRICE_PER_SHARE = 30.00
MAX_RETRIES = 3
HARD_TIMEOUT_SECONDS = 3000 # 50 min to match GitHub Actions
def _timeout_handler(signum, frame):
raise TimeoutError("Hard timeout reached (50 minutes). Aborting.")
# --- NODES ---
def scout_node(state):
"""Two-pronged discovery: yFinance screener + Brave trending, then score."""
region = state.get("region", "USA")
retries = state.get("retry_count", 0)
candidates_queue = state.get("candidates", [])
seen = load_seen_tickers()
# If we still have scored candidates from a previous scout pass, pop next
while candidates_queue:
next_pick = candidates_queue.pop(0)
ticker = next_pick if isinstance(next_pick, str) else next_pick.get("ticker", "")
if ticker and ticker not in seen:
logger.info("Popping next candidate from queue: %s (%d left)", ticker, len(candidates_queue))
mark_ticker_seen(ticker)
return {"ticker": ticker, "candidates": candidates_queue}
logger.info("Skipping %s (seen recently)", ticker)
# Queue is empty — run full discovery
if retries > 0:
logger.info("Retry pause (2s)...")
time.sleep(2)
logger.info("[Attempt %d/%d] Scouting %s micro-caps...", retries + 1, MAX_RETRIES + 1, region)
# Prong 1: Brave Search for trending tickers
trending_tickers = get_trending_tickers_from_brave(region)
logger.info("Brave trending returned %d tickers", len(trending_tickers))
# Prong 2: yFinance screener (merges trending tickers into the pool)
screened = screen_microcaps(region=region, extra_tickers=trending_tickers, max_results=20)
if not screened:
logger.warning("No candidates passed screener for %s", region)
return {"ticker": "NONE", "candidates": []}
# Filter out already-seen tickers
fresh = [c for c in screened if c["ticker"] not in seen]
if not fresh:
logger.warning("All screened candidates already seen for %s", region)
return {"ticker": "NONE", "candidates": []}
# Score and rank
ranked = rank_candidates(fresh, top_n=5)
# Pop the best one, store rest in queue
best = ranked[0]
rest = [c["ticker"] for c in ranked[1:]]
ticker = best["ticker"]
mark_ticker_seen(ticker)
logger.info(
"Target acquired: %s (score=%d) | %d backups queued",
ticker, best.get("score", 0), len(rest),
)
return {"ticker": ticker, "candidates": rest}
def gatekeeper_node(state) -> Command[Literal["analyst", "scout", "email"]]:
"""Validate the candidate against hard financial criteria. Routes via Command."""
import yfinance as yf
ticker = state.get("ticker", "NONE")
retries = state.get("retry_count", 0)
def _fail_route(new_retries: int) -> str:
if new_retries > MAX_RETRIES:
return "email"
return "scout"
if ticker == "NONE":
logger.warning("No ticker to evaluate")
update = {"is_small_cap": False, "retry_count": retries + 1}
return Command(update=update, goto=_fail_route(retries + 1))
logger.info("Gatekeeper evaluating %s...", ticker)
try:
stock = yf.Ticker(ticker)
info = stock.info
mkt_cap = info.get("marketCap", 0) or 0
price = info.get("currentPrice", 0) or info.get("regularMarketPrice", 0) or 0
name = info.get("shortName", ticker)
currency = info.get("currency", "USD")
price = normalize_price(price, ticker, currency)
if price > MAX_PRICE_PER_SHARE:
logger.info("%s rejected — price $%.2f > $%.2f", ticker, price, MAX_PRICE_PER_SHARE)
update = {"is_small_cap": False, "retry_count": retries + 1}
return Command(update=update, goto=_fail_route(retries + 1))
if not (MIN_MARKET_CAP < mkt_cap < MAX_MARKET_CAP):
logger.info("%s rejected — cap $%s out of range", ticker, f"{mkt_cap:,.0f}")
update = {"is_small_cap": False, "retry_count": retries + 1}
return Command(update=update, goto=_fail_route(retries + 1))
health = check_financial_health(ticker, info)
if health["status"] == "FAIL":
logger.info("%s rejected — %s", ticker, health["reason"])
update = {"is_small_cap": False, "retry_count": retries + 1}
return Command(update=update, goto=_fail_route(retries + 1))
sector = health["metrics"].get("sector", "N/A")
logger.info(
"%s PASSED gatekeeper (price=$%.2f | cap=$%s | sector=%s)",
ticker, price, f"{mkt_cap:,.0f}", sector,
)
update = {
"market_cap": mkt_cap,
"is_small_cap": True,
"company_name": name,
"financial_data": info,
}
return Command(update=update, goto="analyst")
except Exception as exc:
logger.error("yFinance error for %s: %s", ticker, exc)
update = {"is_small_cap": False, "retry_count": retries + 1}
return Command(update=update, goto=_fail_route(retries + 1))
def analyst_node(state):
"""Senior Broker analysis with Graham Number, Finnhub data, and insider signals."""
ticker = state["ticker"]
info = state.get("financial_data", {})
region = state.get("region", "USA")
logger.info("Analysing %s...", ticker)
price = info.get("currentPrice", 0) or info.get("regularMarketPrice", 0) or 0
eps = info.get("trailingEps", 0) or 0
book_value = info.get("bookValue", 0) or 0
if eps > 0 and book_value > 0:
strategy = "GRAHAM CLASSIC"
valuation = (22.5 * eps * book_value) ** 0.5
thesis = f"Profitable. Graham Value ${valuation:.2f} vs Price ${price:.2f}."
else:
strategy = "DEEP VALUE ASSET PLAY"
valuation = book_value
ratio = price / book_value if book_value > 0 else 0
thesis = f"Unprofitable Miner/Turnaround. Trading at {ratio:.2f}x Book Value."
# Gather context
news = brave_search(f"{ticker} stock analysis catalysts")
# SEC EDGAR ground truth (US equities only)
sec_context = ""
if region == "USA" and "." not in ticker:
from src.sec_edgar import get_sec_filings
try:
sec_context = get_sec_filings.invoke({"ticker": ticker})
except Exception as exc:
logger.warning("SEC EDGAR failed for %s: %s", ticker, exc)
# Build deep-fundamentals context
deep_fundamentals = ""
if region == "USA" and "." not in ticker:
logger.info("Researching Finnhub databases for %s...", ticker)
context = ""
try:
context += get_insider_sentiment.invoke({"ticker": ticker}) + "\n"
context += get_company_news.invoke({"ticker": ticker}) + "\n"
context += get_basic_financials.invoke({"ticker": ticker}) + "\n"
except Exception as exc:
logger.warning("Finnhub tool error for %s: %s", ticker, exc)
insider = get_insider_buys(ticker)
context += f"\nInsider Sentiment (6mo): {insider['sentiment']} | MSPR: {insider['mspr']} | Net Shares: {insider['change']}\n"
deep_fundamentals = f"DEEP FUNDAMENTALS (FINNHUB + INSIDER FEED):\n{context}"
else:
deep_fundamentals = f"NEWS: {str(news)[:1500]}"
# --- Debate or single-LLM path ---
from src.agents.debate import is_debate_enabled, run_debate
from src.models.kelly import get_kelly_stats, calculate_position_size
if is_debate_enabled():
logger.info("Running multi-agent debate for %s...", ticker)
try:
debate_result = run_debate(
ticker=ticker,
company_name=state.get("company_name", ticker),
financial_data_summary=str(info)[:2000],
deep_fundamentals=deep_fundamentals,
sec_context=sec_context,
strategy=strategy,
price=price,
eps=eps,
book_value=book_value,
ebitda=info.get("ebitda", 0) or 0,
)
result = debate_result["_structured_result"]
stats = get_kelly_stats()
result.position_size = calculate_position_size(stats, result.verdict)
result.kelly_win_rate = stats.win_rate
result.kelly_total_trades = stats.total_trades
verdict = result.to_report()
record_paper_trade(ticker, price, verdict, source="Morning Cron",
structured_verdict=result.verdict,
position_size=result.position_size)
_run_id = get_current_run_id()
log_online_feedback(verdict, ticker, run_id=_run_id)
tag_for_review(verdict, ticker, run_id=_run_id)
return {
"final_verdict": verdict, "debate_used": True,
"bull_case": debate_result.get("bull_case", ""),
"bear_case": debate_result.get("bear_case", ""),
}
except Exception as exc:
logger.warning("Debate failed for %s, falling back to single-LLM: %s", ticker, exc)
# --- Single-LLM path (default or debate fallback) ---
prompt = f"""
Act as a Senior Financial Broker evaluating {state.get('company_name', ticker)} ({ticker}).
HARD DATA: Price: ${price} | EPS: {eps} | Book/Share: {book_value} | EBITDA: {info.get('ebitda', 0)}
QUANTITATIVE THESIS: {thesis}
"""
if sec_context:
prompt += f"\n{sec_context}\n"
prompt += f"\n{deep_fundamentals}\n"
prompt += f"""
Your task is to write a highly structured investment memo combining strict {strategy} math with qualitative analysis and recent insider behavior/news. Do not use fluff or buzzwords.
Format your response EXACTLY like this:
### THE QUANTITATIVE BASE (Graham / Asset Play)
* State the current Price vs the calculated {strategy} valuation.
* Briefly explain if the math supports a margin of safety.
### THE LYNCH PITCH (Why I would own this)
* **The Core Action:** In one sentence, what are insiders doing (buying/selling/neutral)?
* **The Catalyst:** Based on the news, what is the ONE simple reason this stock could run?
### THE MUNGER INVERT (How I could lose money)
* **Structural Weakness:** What is the most likely way an investor loses money here based on fundamentals/news?
* **The Bear Evidence:** What exact metric, news, or math would prove the bear case right?
### FINAL VERDICT
STRONG BUY / BUY / WATCH / AVOID (Choose one, followed by a 1-sentence bottom line).
"""
try:
import warnings
from src.models.verdict import InvestmentVerdict
structured_llm = get_structured_llm().with_structured_output(InvestmentVerdict)
result = structured_llm.invoke(prompt)
stats = get_kelly_stats()
result.position_size = calculate_position_size(stats, result.verdict)
result.kelly_win_rate = stats.win_rate
result.kelly_total_trades = stats.total_trades
verdict = result.to_report()
record_paper_trade(ticker, price, verdict, source="Morning Cron",
structured_verdict=result.verdict,
position_size=result.position_size)
_run_id = get_current_run_id()
log_online_feedback(verdict, ticker, run_id=_run_id)
tag_for_review(verdict, ticker, run_id=_run_id)
except Exception as exc:
logger.warning("Structured output failed for %s, falling back to plain LLM: %s", ticker, exc)
try:
verdict = invoke_with_fallback(prompt)
stats = get_kelly_stats()
v_upper = verdict.upper()
verdict_type = "AVOID"
if "STRONG BUY" in v_upper:
verdict_type = "STRONG BUY"
elif "BUY" in v_upper:
verdict_type = "BUY"
elif "WATCH" in v_upper:
verdict_type = "WATCH"
pos = calculate_position_size(stats, verdict_type)
if pos > 0:
verdict += (
f"\n\n### POSITION SIZING (Kelly Criterion)\n"
f"**Recommended allocation: {pos:.1f}% of portfolio**"
)
record_paper_trade(ticker, price, verdict, source="Morning Cron",
position_size=pos)
_run_id = get_current_run_id()
log_online_feedback(verdict, ticker, run_id=_run_id, is_fallback=True)
tag_for_review(verdict, ticker, run_id=_run_id, is_fallback=True)
except Exception as exc2:
logger.error("LLM analysis failed for %s: %s", ticker, exc2)
verdict = f"LLM analysis unavailable: {exc2}"
return {"final_verdict": verdict}
def email_node(state):
"""Send the analysis or failure report to the team."""
region = state.get("region", "Global")
ticker = state.get("ticker", "Unknown")
verdict = state.get("final_verdict", "No Verdict")
if not state.get("is_small_cap"):
logger.info("Sending failure report for %s...", region)
subject = f"Hunt Failed: {region}"
body = f"Found no suitable micro-caps under ${MAX_PRICE_PER_SHARE} in {region} after {MAX_RETRIES + 1} attempts."
else:
logger.info("Sending analysis for %s...", ticker)
subject = f"Micro-Cap Found ({region}): {ticker}"
body = (
f"<h1>{ticker}</h1>"
f"<h3>Cap: ${state.get('market_cap', 0):,.0f}</h3>"
f"<hr>{verdict.replace(chr(10), '<br>')}"
)
team = [
{"name": "Cisco", "email": os.getenv("EMAIL_CISCO"), "key": os.getenv("RESEND_API_KEY_CISCO")},
{"name": "Raul", "email": os.getenv("EMAIL_RAUL"), "key": os.getenv("RESEND_API_KEY_RAUL")},
{"name": "David", "email": os.getenv("EMAIL_DAVID"), "key": os.getenv("RESEND_API_KEY_DAVID")},
]
for member in team:
if member["email"] and member["key"]:
try:
send_email_report(subject, body, member["email"], member["key"])
except Exception as exc:
logger.error("Email to %s failed: %s", member["name"], exc)
return {}
# ---------------------------------------------------------------------------
# Per-region subgraph (scout -> gatekeeper -> analyst -> email)
# ---------------------------------------------------------------------------
_api_retry = RetryPolicy(max_attempts=3, initial_interval=2.0)
_region_workflow = StateGraph(AgentState)
_region_workflow.add_node("scout", scout_node, retry=_api_retry)
_region_workflow.add_node("gatekeeper", gatekeeper_node, retry=_api_retry)
_region_workflow.add_node("analyst", analyst_node, retry=_api_retry)
_region_workflow.add_node("email", email_node, retry=_api_retry)
_region_workflow.add_edge(START, "scout")
_region_workflow.add_edge("scout", "gatekeeper")
_region_workflow.add_edge("analyst", "email")
_region_workflow.add_edge("email", END)
_region_app = _region_workflow.compile(checkpointer=InMemorySaver())
# ---------------------------------------------------------------------------
# Orchestrator graph (parallel fan-out via Send API)
# ---------------------------------------------------------------------------
class GlobalHunterState(TypedDict, total=False):
regions: list[str]
region_results: Annotated[list, operator.add]
def dispatch_regions(state: GlobalHunterState):
"""Fan-out: emit one Send per region so they run in parallel."""
return [
Send("hunt_region", {"region": r})
for r in state["regions"]
]
def hunt_region(state) -> dict:
"""Invoke the full per-region pipeline and report back."""
region = state.get("region", "USA")
logger.info("--- Initiating hunt for %s ---", region)
try:
config = {
"configurable": {"thread_id": f"hunt-{region.lower()}"},
"recursion_limit": 30,
}
_region_app.invoke(
{"region": region, "retry_count": 0, "ticker": ""},
config,
)
logger.info("%s hunt complete.", region)
return {"region_results": [{"region": region, "success": True}]}
except Exception as exc:
logger.error("Error in %s: %s", region, exc, exc_info=True)
return {"region_results": [{"region": region, "success": False, "error": str(exc)}]}
_orchestrator = StateGraph(GlobalHunterState)
_orchestrator.add_node("hunt_region", hunt_region, retry=_api_retry)
_orchestrator.add_conditional_edges(START, dispatch_regions, ["hunt_region"])
_orchestrator.add_edge("hunt_region", END)
app = _orchestrator.compile(checkpointer=InMemorySaver())
# ---------------------------------------------------------------------------
# Execution
# ---------------------------------------------------------------------------
if __name__ == "__main__":
catalyst_ticker = os.getenv("CATALYST_TICKER", "").strip()
if catalyst_ticker:
logger.info("Catalyst alert mode — analysing %s only", catalyst_ticker)
try:
signal.signal(signal.SIGALRM, _timeout_handler)
signal.alarm(HARD_TIMEOUT_SECONDS)
except (AttributeError, ValueError):
pass
config = {
"configurable": {"thread_id": f"catalyst-{catalyst_ticker.lower()}"},
"recursion_limit": 30,
}
_region_app.invoke(
{"region": "USA", "retry_count": 0, "ticker": catalyst_ticker},
config,
)
logger.info("Catalyst analysis complete for %s.", catalyst_ticker)
else:
logger.info("Starting Global Micro-Cap Hunter (Screener + Brave Edition)...")
try:
signal.signal(signal.SIGALRM, _timeout_handler)
signal.alarm(HARD_TIMEOUT_SECONDS)
logger.info("Timeout set: %ds", HARD_TIMEOUT_SECONDS)
except (AttributeError, ValueError):
logger.info("SIGALRM not available on this platform")
regions = ["USA", "UK", "Canada", "Australia"]
config = {"configurable": {"thread_id": "global-hunt"}, "recursion_limit": 30}
result = app.invoke({"regions": regions}, config)
for entry in result.get("region_results", []):
status = "OK" if entry.get("success") else f"FAILED: {entry.get('error', 'unknown')}"
logger.info("Region %s: %s", entry.get("region"), status)
logger.info("Global mission complete.")
|