Provider abstraction: providers/ package with BaseDataProvider template

One class per provider (TwelveData, Oanda, Tiingo), all subclassing
BaseDataProvider which owns cache, validation/cleaning, freshness +
boundary guarantees, fetch_range pagination, and streaming stubs —
subclasses implement only _fetch_raw. config.json data_providers.active
selects the live feed; data_provider.py / oanda_provider.py stay as
back-compat shims. Module 16 provider-abstraction tests (253 total green).

Claude-Session: https://claude.ai/code/session_01Mc7XKYyPySVXiUpcN82QEK

backtest/diagnose_filters.py

@@ -51,7 +51,7 @@ from gpu_backtest import (  # noqa: E402
     precompute_indicators,
     _is_xau_blocked_by_calendar,
 )
-from data_provider import ForexDataProvider  # noqa: E402
+from providers import make_data_provider  # noqa: E402
 
 logger = logging.getLogger("diagnose_filters")
 
@@ -206,7 +206,7 @@ def run_diagnosis(
         confirmation_timeframes=list(confirm_tfs),
     )
 
-    provider = ForexDataProvider(platform_cfg)
+    provider = make_data_provider(platform_cfg)
 
     # ── Load data ──
     timeframes = [primary_tf] + list(confirm_tfs)

backtest/gpu_backtest.py

@@ -55,7 +55,7 @@ from config import (  # noqa: E402
     Timeframe,
     XAUUSD,
 )
-from data_provider import ForexDataProvider, make_data_provider  # noqa: E402
+from providers import BaseDataProvider, make_data_provider  # noqa: E402
 from engine import IndicatorSignal, TechnicalAnalysis, TechnicalEngine  # noqa: E402
 from signal_aggregator import SignalAggregator, SignalReport, TradeSetup  # noqa: E402
 
@@ -941,7 +941,7 @@ def _cache_path(symbol: str, tf: Timeframe, start: datetime, end: datetime) -> P
 
 
 def fetch_history(
-    provider: ForexDataProvider,
+    provider: BaseDataProvider,
     symbol: str,
     timeframe: Timeframe,
     start: datetime,
@@ -949,12 +949,13 @@ def fetch_history(
     use_cache: bool = True,
     clean: bool = True,
 ) -> pd.DataFrame:
-    """Fetch OHLCV between start/end with parquet caching and pagination.
+    """Fetch OHLCV between start/end with parquet caching.
 
-    Twelve Data caps responses at 5000 bars; we paginate by date range so
-    multi-month M5/M15 backtests work.
+    Pagination under the provider's per-request bar cap lives in
+    `BaseDataProvider.fetch_range`, so any registered provider can feed
+    the backtester.
 
-    The parquet cache stores the RAW series; `_validate_and_clean` runs on
+    The parquet cache stores the RAW series; `validate_and_clean` runs on
     load. Pass clean=False to get the raw price path — trade resolution must
     use it, because cleaning drops >10σ spike bars (news candles) that hit
     stops in reality. (Caches written before this change already contain
@@ -967,73 +968,16 @@ def fetch_history(
         df = pd.read_parquet(path)
         if df.index.tz is None:
             df.index = df.index.tz_localize("UTC")
-        return provider._validate_and_clean(df, symbol, timeframe) if clean else df
+        return provider.validate_and_clean(df, symbol, timeframe) if clean else df
 
     logger.info(f"Fetching {symbol} {timeframe.value} from {start.date()} to {end.date()}")
 
-    # Bars-per-call estimate to size each window safely under the 5000 cap.
-    bars_per_day = ForexDataProvider.BARS_PER_DAY[timeframe]
-    days_per_chunk = max(1, int(4500 / bars_per_day))
-
-    chunks: List[pd.DataFrame] = []
-    cursor_end = end
-    empty_streak = 0
-    while cursor_end > start:
-        cursor_start = max(start, cursor_end - timedelta(days=days_per_chunk))
-        try:
-            chunk = provider._fetch_twelve_data(
-                symbol=symbol,
-                timeframe=timeframe,
-                outputsize=5000,
-                start_date=cursor_start.strftime("%Y-%m-%d %H:%M:%S"),
-                end_date=cursor_end.strftime("%Y-%m-%d %H:%M:%S"),
-            )
-        except Exception as e:
-            # Twelve Data raises BadRequest for windows with no data (weekends,
-            # holidays, asset gaps). Skip the window, step back, retry — but
-            # bail out after a few consecutive empties to avoid spinning.
-            logger.warning(
-                f"Chunk {cursor_start.date()}–{cursor_end.date()} failed: {e}"
-            )
-            empty_streak += 1
-            if empty_streak >= 3:
-                break
-            cursor_end = cursor_start - timedelta(seconds=1)
-            continue
-
-        if chunk is None or chunk.empty:
-            empty_streak += 1
-            if empty_streak >= 3:
-                break
-            cursor_end = cursor_start - timedelta(seconds=1)
-            continue
-
-        empty_streak = 0
-        chunks.append(chunk)
-        # Step back; dedup on concat.
-        cursor_end = chunk.index[0] - timedelta(seconds=1)
-        if len(chunk) < 2:
-            break
-
-    if not chunks:
-        raise RuntimeError(f"No data fetched for {symbol} {timeframe.value}")
-
-    df = pd.concat(chunks).sort_index()
-    df = df[~df.index.duplicated(keep="last")]
-    # Fetches ending "now" include the still-forming candle — drop it so the
-    # cache only ever contains closed bars.
-    df = provider._drop_forming_candle(df, timeframe)
-    # Trim to requested window. `start`/`end` are already tz-aware UTC
-    # datetimes — wrapping with pd.Timestamp(..., tz=...) would raise on
-    # already-localized inputs.
-    end_ts = pd.Timestamp(end)
-    start_ts = pd.Timestamp(start)
-    df = df[(df.index >= start_ts) & (df.index <= end_ts)]
+    df = provider.fetch_range(symbol, timeframe, start, end)
 
     if use_cache:
         df.to_parquet(path)
         logger.info(f"Cached {len(df)} raw bars to {path.name}")
-    return provider._validate_and_clean(df, symbol, timeframe) if clean else df
+    return provider.validate_and_clean(df, symbol, timeframe) if clean else df
 
 
 # ═════════════════════════════════════════════════════════════════════
@@ -1703,7 +1647,7 @@ def run_backtest(
         # ≥210 bars (outputsize floor). Target data_provider's min_required
         # (210 bars); ×7/5 converts trading days to calendar days. Intraday
         # TFs stay on the base pad, so their books are unchanged.
-        bars_per_day = ForexDataProvider.BARS_PER_DAY[tf]
+        bars_per_day = BaseDataProvider.BARS_PER_DAY[tf]
         min_pad_days = int(np.ceil(210 / bars_per_day * 7 / 5)) + 5
         tf_start = bt_cfg.start - timedelta(days=max(pad_days, min_pad_days))
         df = fetch_history(