How one can Use AgentTrove: Streaming 1.7M Agentic Traces and Constructing a Clear ShareGPT SFT Dataset in Python
def is_success(row): res = (row.get(“consequence”) or “”).decrease() if res in (“resolved”, “success”, “go”, “handed”, “appropriate”): return True rw = row.get(“reward”) attempt: return float(rw) >= 1.0 besides (TypeError, ValueError): return False out_path = “agenttrove_clean_sft.jsonl” stored, scanned, SCAN, KEEP = 0, 0, 1500, 200 print(f”n⏳ Scanning as much as {SCAN} rows, protecting as much as {KEEP} profitable…