Spaces:
Running
Running
Commit History
-
Upload folder using huggingface_hub
Verified signaturea7e09b5 -
Upload folder using huggingface_hub
Verified signaturea8ad61a -
Upload folder using huggingface_hub
Verified signaturea179080
-
Picks panels: rank on recent activity; tighten the insider screen
Super Investor Activity - Rank on net change in the latest 13F quarter (2x new + adds - trims - 2x exits) instead of gross buying; n_gurus tiebreak dropped, ticker as a final deterministic tiebreak. Detail line now shows trims as well as exits. Derives the net figure from the four existing columns when the snapshot predates guru_activity_net. Insider Activity - Rank on the 30-day window (insider_*_30d) when the snapshot ships it, else the 90-day fields, so the code is safe to deploy before or after the new snapshot lands. - Display minimums: >= $100k aggregate and >= $25k per buyer (drops small-ticket and employee-plan-style buying, e.g. TSM's 30 buyers x $0.3M). - 13F coverage filter is n_holders > 0, not notna (KBDC/BPRE had 0 holders). - Shows "Latest buy: <date>". LLM screen - Insider prompt: judge only the listed fraud criteria; a single buyer or a stock's reputation is not suspicious when the transactions look genuine (the screen had excluded LILA, KBDC and FTK fo…50f32cd
-
Fix standing margin balance, restore the 20-name book, harden the panels
Executor - Size targets off (1 - CASH_BUFFER) x equity instead of 100%, and add a cash-restore pass over deadband-skipped overweights. A fully-invested target plus a one-sided min-trade deadband left the paper account on a permanent margin balance (-1,712 on 98.5k, gross 101.7%) that no rebalance corrected. - Place sells first and WAIT for terminal status before buys, so buys are funded by settled sells rather than briefly by margin. - Drop the no-op ticker .replace(".", "."). Book - Add MIN_POS_W (1%) floor alternating with the 15% cap. The 13F size column spans ~140x, so the tail rounded below the $493 min trade (BSX $26, PODD $14, DINO $71, VEEV $235) and four target names were never bought: a "20-stock" fund holding 16. Effective breadth 6 -> 10.4. - Extend SPY with live Alpaca bars for the regime test, so the fund's biggest switch no longer waits on a bundle that can be 9+ days stale. - Memoise compute_book for 15 min; the real rebalance forces a fresh book. Pick…ea01671
-
Upload services/supervisor.py with huggingface_hub
Verified signature7ec49f5
-
Upload services/news.py with huggingface_hub
Verified signature0d1670f -
Upload services/news.py with huggingface_hub
Verified signatureaca3394 -
Upload services/insider.py with huggingface_hub
Verified signatured874111 -
Upload services/superinvestor.py with huggingface_hub
Verified signature65391da
-
Add Gemini (gemini-3.6-flash) as preferred picks screener
Verified signaturee5e019b -
Slipstream fund + AI picks — initial deploy
Verified signature134649f