feat: production paper trading platform

This commit is contained in:
Курнат Андрей
2026-07-14 22:52:47 +03:00
parent 7186acb9a1
commit 5c4aecfe5f
29 changed files with 396 additions and 564 deletions
+53 -6
View File
@@ -3,6 +3,7 @@ from __future__ import annotations
import base64
import hashlib
import json
from datetime import UTC, datetime, timedelta
import pytest
@@ -16,6 +17,7 @@ def test_training_coordinator_claims_and_completes_job(tmp_path) -> None:
job_id = requested["job"]["id"]
heartbeat = coordinator.heartbeat({"worker_id": "win-1", "name": "DESKTOP-TMFDL0H"})
claimed = coordinator.claim({"worker_id": "win-1", "name": "DESKTOP-TMFDL0H"})
lease_token = claimed["lease_token"]
assert requested["queued"] is True
assert heartbeat["status"]["agent_online"] is True
@@ -25,14 +27,23 @@ def test_training_coordinator_claims_and_completes_job(tmp_path) -> None:
progress = coordinator.progress(
job_id,
{"status": "running", "phase": "training", "progress_percent": 42, "message": "epoch 1"},
{
"status": "running",
"phase": "training",
"progress_percent": 42,
"message": "epoch 1",
"lease_token": lease_token,
},
)
assert progress["job"]["phase"] == "training"
assert progress["job"]["progress_percent"] == 42
assert coordinator.status()["active_job"]["message"] == "epoch 1"
completed = coordinator.complete(job_id, {"success": True, "message": "ok"})
completed = coordinator.complete(
job_id,
{"success": True, "message": "ok", "lease_token": lease_token},
)
assert completed["job"]["status"] == "completed"
assert coordinator.status()["active_job"] is None
@@ -104,7 +115,7 @@ def test_training_coordinator_reports_worker_identity_from_heartbeat(tmp_path) -
def test_training_coordinator_records_rejected_candidate_as_completed_training(tmp_path) -> None:
coordinator = TrainingCoordinator(tmp_path)
job = coordinator.request_retrain({"source": "android"})["job"]
coordinator.claim({"worker_id": "worker-1"})
lease_token = coordinator.claim({"worker_id": "worker-1"})["lease_token"]
completed = coordinator.complete(
job["id"],
@@ -112,6 +123,7 @@ def test_training_coordinator_records_rejected_candidate_as_completed_training(t
"success": True,
"message": "training completed; candidate rejected by quality gate",
"summary": {"accepted": False, "reason": "candidate_failed_honest_validation"},
"lease_token": lease_token,
},
)
@@ -124,7 +136,7 @@ def test_training_coordinator_records_rejected_candidate_as_completed_training(t
def test_training_coordinator_accepts_chunked_artifact_upload(tmp_path) -> None:
coordinator = TrainingCoordinator(tmp_path)
job = coordinator.request_retrain({"source": "test"})["job"]
coordinator.claim({"worker_id": "test-worker"})
lease_token = coordinator.claim({"worker_id": "test-worker"})["lease_token"]
payload = b'{"type":"pytorch_recurrent_forecaster","symbols":{}}\n'
sha256 = hashlib.sha256(payload).hexdigest()
first = payload[:20]
@@ -138,6 +150,7 @@ def test_training_coordinator_accepts_chunked_artifact_upload(tmp_path) -> None:
"total": 2,
"sha256": sha256,
"data_base64": base64.b64encode(first).decode("ascii"),
"lease_token": lease_token,
},
)
part_2 = coordinator.save_artifact_chunk(
@@ -148,6 +161,7 @@ def test_training_coordinator_accepts_chunked_artifact_upload(tmp_path) -> None:
"total": 2,
"sha256": sha256,
"data_base64": base64.b64encode(second).decode("ascii"),
"lease_token": lease_token,
},
)
@@ -199,6 +213,35 @@ def test_running_claimed_job_keeps_agent_online_when_heartbeat_is_stale(tmp_path
assert status["agent_online"] is True
def test_stale_training_lease_is_requeued_and_old_lease_is_rejected(tmp_path) -> None:
coordinator = TrainingCoordinator(tmp_path)
job = coordinator.request_retrain({"source": "android"})["job"]
first_claim = coordinator.claim({"worker_id": "worker-1"})
state_path = tmp_path / "training_coordination.json"
state = json.loads(state_path.read_text(encoding="utf-8"))
state["jobs"][0]["updated_at"] = (
datetime.now(UTC) - timedelta(minutes=11)
).isoformat()
state_path.write_text(json.dumps(state), encoding="utf-8")
second_claim = coordinator.claim({"worker_id": "worker-2"})
assert second_claim["claimed"] is True
assert second_claim["job"]["id"] == job["id"]
assert second_claim["job"]["attempts"] == 2
assert second_claim["lease_token"] != first_claim["lease_token"]
with pytest.raises(ValueError, match="lease"):
coordinator.progress(
job["id"],
{
"phase": "training",
"progress_percent": 10,
"lease_token": first_claim["lease_token"],
},
)
def test_training_upload_rejects_unknown_job(tmp_path) -> None:
coordinator = TrainingCoordinator(tmp_path)
payload = b"{}"
@@ -219,7 +262,7 @@ def test_training_upload_rejects_unknown_job(tmp_path) -> None:
def test_training_bundle_promotes_only_after_successful_guard(tmp_path) -> None:
coordinator = TrainingCoordinator(tmp_path)
job = coordinator.request_retrain({"source": "test"})["job"]
coordinator.claim({"worker_id": "worker-1"})
lease_token = coordinator.claim({"worker_id": "worker-1"})["lease_token"]
model = {
"type": "pytorch_recurrent_forecaster",
"symbols": {
@@ -260,10 +303,14 @@ def test_training_bundle_promotes_only_after_successful_guard(tmp_path) -> None:
"total": 1,
"sha256": hashlib.sha256(payload).hexdigest(),
"data_base64": base64.b64encode(payload).decode("ascii"),
"lease_token": lease_token,
},
)
completed = coordinator.complete(job["id"], {"success": True})
completed = coordinator.complete(
job["id"],
{"success": True, "lease_token": lease_token},
)
assert completed["job"]["status"] == "completed"
assert json.loads((tmp_path / "lstm_forecaster.json").read_text())["symbols"]["BTCUSDT"]