feat: collect Bybit orderbook observations for training

tests/test_bybit.py

@@ -125,3 +125,14 @@ def test_websocket_subscribe_uses_configured_kline_interval() -> None:
 
     assert "kline.60.BTCUSDT" in payload
     assert "kline.1.BTCUSDT" not in payload
+
+
+def test_orderbook_level_one_preserves_sizes(make_settings, tmp_path) -> None:
+    client = BybitClient(make_settings(tmp_path))
+    client.public_get = lambda *_args, **_kwargs: {
+        "b": [["100.5", "2.25"]],
+        "a": [["100.7", "1.75"]],
+    }
+
+    assert client.orderbook_level_one("BTCUSDT") == (100.5, 2.25, 100.7, 1.75)
+    assert client.orderbook_top("BTCUSDT") == (100.5, 100.7)

tests/test_config.py

@@ -180,3 +180,12 @@ def test_load_settings_rejects_unknown_fallback_mode(tmp_path, monkeypatch) -> N
 
     with pytest.raises(ValueError, match="TIME_SERIES_FALLBACK_MODE"):
         load_settings(env_file)
+
+
+def test_load_settings_rejects_non_positive_observation_interval(tmp_path, monkeypatch) -> None:
+    monkeypatch.delenv("MARKET_OBSERVATION_SAMPLE_SECONDS", raising=False)
+    env_file = tmp_path / ".env"
+    env_file.write_text("MARKET_OBSERVATION_SAMPLE_SECONDS=0\n", encoding="utf-8")
+
+    with pytest.raises(ValueError, match="MARKET_OBSERVATION_SAMPLE_SECONDS"):
+        load_settings(env_file)

tests/test_dashboard.py

@@ -46,6 +46,8 @@ def test_safe_config_summarizes_torch_forecast_artifact(make_settings, tmp_path)
     assert config["time_series_probe_size_multiplier"] == 0.40
     assert config["time_series_rebound_fallback_enabled"] is True
     assert config["time_series_fallback_mode"] == "trend_macd"
+    assert config["market_observation_enabled"] is True
+    assert config["market_observation_sample_seconds"] == 30.0
     assert config["time_series_model_artifact"] == {
         "available": True,
         "type": "pytorch_recurrent_forecaster",

tests/test_market_data.py

@@ -1,7 +1,8 @@
 from __future__ import annotations
 
-from crypto_spot_bot.market_data import _candles_due, _closed_candles, _is_closed_kline_row
+from crypto_spot_bot.market_data import MarketData, _candles_due, _closed_candles, _is_closed_kline_row
 from crypto_spot_bot.models import Candle
+from crypto_spot_bot.storage import Storage
 
 
 def test_closed_candles_excludes_current_open_interval() -> None:
@@ -26,3 +27,37 @@ def test_rest_candles_refresh_only_after_next_bar_closes() -> None:
 
     assert _candles_due([candle], "1", now_ms=11 * 60_000 + 30_000) is False
     assert _candles_due([candle], "1", now_ms=12 * 60_000) is True
+
+
+def test_orderbook_handler_samples_sizes_and_microstructure(make_settings, tmp_path) -> None:
+    settings = make_settings(
+        tmp_path,
+        market_observation_enabled=True,
+        market_observation_sample_seconds=30.0,
+    )
+    storage = Storage(settings.database_path)
+    market = MarketData(settings, object(), storage)
+
+    market._handle_orderbook(
+        "BTCUSDT",
+        {"b": [["100", "3"]], "a": [["101", "1"]]},
+        source_timestamp_ms=1_789_000_000_000,
+    )
+    market._handle_orderbook(
+        "BTCUSDT",
+        {"b": [["100", "4"]], "a": [["101", "1"]]},
+        source_timestamp_ms=1_789_000_001_000,
+    )
+
+    metrics = market.orderbook_metrics["BTCUSDT"]
+    rows = storage.market_observations_after(symbol="BTCUSDT")
+    assert metrics["bid_size"] == 4.0
+    assert metrics["ask_size"] == 1.0
+    assert metrics["imbalance"] == 0.6
+    assert metrics["microprice"] == 100.8
+    assert len(rows) == 1
+    assert rows[0]["bid_size"] == 3.0
+    assert rows[0]["ask_size"] == 1.0
+    assert rows[0]["imbalance"] == 0.5
+    assert rows[0]["microprice"] == 100.75
+    assert rows[0]["source_timestamp_ms"] == 1_789_000_000_000

tests/test_storage.py

@@ -62,6 +62,57 @@ def test_prune_deletes_only_one_bounded_batch_per_table(tmp_path) -> None:
     assert len(storage.recent_signals(PRUNE_BATCH_SIZE + 10)) == 5
 
 
+def test_market_observation_export_is_symbol_scoped_and_paginated(tmp_path) -> None:
+    storage = Storage(tmp_path / "tradebot.sqlite3")
+    first_id = storage.insert_market_observation(
+        symbol="BTCUSDT",
+        bid_price=100.0,
+        bid_size=2.0,
+        ask_price=101.0,
+        ask_size=1.0,
+        mid_price=100.5,
+        microprice=100.6666666667,
+        spread_bps=99.50248756,
+        imbalance=1 / 3,
+        last_price=100.4,
+        source_timestamp_ms=1_789_000_000_000,
+    )
+    second_id = storage.insert_market_observation(
+        symbol="BTCUSDT",
+        bid_price=101.0,
+        bid_size=1.0,
+        ask_price=102.0,
+        ask_size=1.0,
+        mid_price=101.5,
+        microprice=101.5,
+        spread_bps=98.52216749,
+        imbalance=0.0,
+        last_price=101.4,
+        source_timestamp_ms=1_789_000_030_000,
+    )
+    storage.insert_market_observation(
+        symbol="ETHUSDT",
+        bid_price=10.0,
+        bid_size=1.0,
+        ask_price=11.0,
+        ask_size=1.0,
+        mid_price=10.5,
+        microprice=10.5,
+        spread_bps=952.38095238,
+        imbalance=0.0,
+        last_price=10.4,
+    )
+
+    rows = storage.market_observations_after(
+        symbol="BTCUSDT",
+        after_id=first_id,
+        limit=1,
+    )
+
+    assert [row["id"] for row in rows] == [second_id]
+    assert rows[0]["source_timestamp_ms"] == 1_789_000_030_000
+
+
 def test_runtime_compaction_preserves_durable_state_and_bounds_telemetry(tmp_path) -> None:
     database = tmp_path / "tradebot.sqlite3"
     storage = Storage(database)