Release 2026-10-05: row graphs for 64, 128 and 256 tokens, shared-state pairs, final Gated DeltaNet kernel
90285c9 verified | """Example: one support ticket, three questions (choice, noul, score), answered on the CPU. | |
| Run from the repository root: | |
| pip install -r host/requirements-host.txt | |
| python examples/run_example.py # prints the /v1/systemone response | |
| python examples/run_example.py --check # also compares it with examples/run_example.expected.json | |
| python examples/run_example.py --mode row # only the row graphs (default: auto) | |
| The expected file holds the response without "latency_ms" (a timing, different on every run). The ticket and every | |
| name in it are invented. In the default mode ("auto") the host runs the three questions through the shared-state pair | |
| when its file is present (the state once, then one question step per question: 128 + 3 x 64 = 320 positions, against | |
| 3 x 256 = 768 through the row graphs) and through the row graphs otherwise; "row" uses only the row graphs (here L256), | |
| "pair" only the pair. | |
| """ | |
| import argparse | |
| import json | |
| import sys | |
| from pathlib import Path | |
| ROOT = Path(__file__).resolve().parents[1] | |
| sys.path.insert(0, str(ROOT / "host")) | |
| from kev_litert import KevLiteRT # noqa: E402 | |
| REQUEST = { | |
| "state": "Ticket #48213, opened by Mara Quellen.\n\nHi, I ordered the Thistlebeam desk lamp (order TB-20931) on " | |
| "September 14 and was charged twice on my card: two identical charges of 64.90 appeared on September 15. " | |
| "The lamp arrived yesterday and works perfectly, no complaints there. Could you refund the duplicate " | |
| "charge? I need it sorted before my card statement closes on Friday. No rush on anything else, and thanks " | |
| "for the lovely lamp.", | |
| "questions": { | |
| "team": {"type": "choice", "instructions": "Which team should handle this ticket?", | |
| "criteria": {"billing": "Charges, refunds and invoices", | |
| "shipping": "Deliveries, tracking and lost parcels", | |
| "returns": "Exchanges and sending a product back", | |
| "technical": "Product faults and setup help"}}, | |
| "deadline": {"type": "noul", "instructions": "Does the customer ask for action by a specific deadline?", | |
| "criteria": {"true": "The ticket names a day or date by which something must happen", | |
| "false": "No deadline is stated"}}, | |
| "mood": {"type": "score", "instructions": "How upset is the customer?", "criteria": ["Calm", "Annoyed", "Angry"]}, | |
| }, | |
| } | |
| def main(): | |
| ap = argparse.ArgumentParser() | |
| ap.add_argument("--check", action="store_true", help="compare with examples/run_example.expected.json") | |
| ap.add_argument("--mode", choices=["auto", "row", "pair"], default="auto", help="the host's route (default auto)") | |
| a = ap.parse_args() | |
| with KevLiteRT.from_dir(ROOT, mode=a.mode) as kev: | |
| response = kev.decide(REQUEST) | |
| print(json.dumps(response, indent=2, ensure_ascii=False)) | |
| if a.check: | |
| expected = json.loads((ROOT / "examples/run_example.expected.json").read_text(encoding="utf-8")) | |
| got = {k: v for k, v in response.items() if k != "latency_ms"} | |
| print("matches run_example.expected.json" if got == expected else | |
| f"DIFFERS from run_example.expected.json:\n{json.dumps(expected, indent=2, ensure_ascii=False)}") | |
| if got != expected: | |
| raise SystemExit(1) | |
| if __name__ == "__main__": | |
| main() | |