ml4t · stefan-jansen · Jul 7, 2026 · Jun 10, 2026 · Jun 10, 2026 · Jun 10, 2026
diff --git a/.gitignore b/.gitignore
@@ -34,7 +34,11 @@ htmlcov/
 logs/
 site/
 src/ml4t/data/_version.py
+coverage/
 
 # Claude Code (local development only)
 CLAUDE.md
 .claude/
+.agentic_documentation/
+.skills-outputs/
+planning/
diff --git a/README.md b/README.md
@@ -102,6 +102,7 @@ fred = FREDProvider().fetch_series("GDP", "2020-01-01", "2024-12-31")
 
 | Provider | Coverage |
 |----------|----------|
+| Alpaca | US equities + crypto (free IEX feed) |
 | EODHD | 60+ global exchanges |
 | Tiingo | US equities with quality focus |
 | Twelve Data | Multi-asset coverage |

diff --git a/docs/INTEGRATION_TESTING.md b/docs/INTEGRATION_TESTING.md
@@ -170,6 +170,8 @@ pytest tests/integration/ -v -W default
 
 Required secrets:
 ```
+ALPACA_API_KEY
+ALPACA_API_SECRET
 CRYPTOCOMPARE_API_KEY
 DATABENTO_API_KEY
 OANDA_API_KEY

diff --git a/docs/providers/README.md b/docs/providers/README.md
@@ -13,6 +13,7 @@ For detailed pricing, terms, and gap analysis, see [PROVIDER_AUDIT.md](PROVIDER_
 | Provider | API Key | Free Tier | Best For |
 |----------|---------|-----------|----------|
 | [YahooFinance](yahoo.md) | No | Unlimited* | Quick start, US equities |
+| [Alpaca](alpaca.md) | Yes | 200 req/min (IEX feed) | US equities + crypto, intraday |
 | [EODHD](eodhd.md) | Yes | 20 calls/day | Global equities (60+ exchanges) |
 | [Tiingo](tiingo.md) | Yes | 1,000 req/day | US equities alternative |
 | [Finnhub](finnhub.md) | Yes | 60 req/min | Company metrics, real-time |
@@ -73,6 +74,7 @@ For detailed pricing, terms, and gap analysis, see [PROVIDER_AUDIT.md](PROVIDER_
 | Provider | Minute | Hourly | Daily | Options | Fundamentals |
 |----------|--------|--------|-------|---------|--------------|
 | YahooFinance | ✅ (7d) | ✅ | ✅ | ❌* | ❌* |
+| Alpaca | ✅ | ✅ | ✅ | ❌ | ❌ |
 | Databento | ✅ | ✅ | ✅ | ✅ (OPRA) | ❌ |
 | Massive | ✅ | ✅ | ✅ | ✅ | ✅ |
 | EODHD | ❌ | ❌ | ✅ | ✅ ($29.99) | ✅ ($59.99) |
@@ -200,6 +202,9 @@ Create a `.env` file in your project root:
 
 ```bash
 # Free tier providers
+ALPACA_API_KEY=your_key_here
+ALPACA_API_SECRET=your_secret_here
+# (Alpaca's own SDK names APCA_API_KEY_ID / APCA_API_SECRET_KEY also work)
 EODHD_API_KEY=your_key_here
 TIINGO_API_KEY=your_key_here
 FINNHUB_API_KEY=your_key_here
@@ -251,6 +256,7 @@ See [Incremental Updates Guide](../storage/INCREMENTAL_ARCHITECTURE.md) for deta
 | Provider | Split Adjusted | Dividend Adjusted |
 |----------|----------------|-------------------|
 | YahooFinance | ✅ | ✅ |
+| Alpaca | Opt-in (`adjustment=`) | Opt-in (`adjustment=`) |
 | EODHD | ✅ | ✅ |
 | Tiingo | ✅ | ✅ |
 | WikiPrices | ✅ | ✅ |

diff --git a/docs/providers/alpaca.md b/docs/providers/alpaca.md
@@ -0,0 +1,151 @@
+# Alpaca Provider
+
+**Provider**: `AlpacaDataProvider`
+**Website**: [alpaca.markets](https://alpaca.markets)
+**API Key**: Required (key + secret pair)
+**Free Tier**: 200 requests/min, real-time IEX feed
+
+---
+
+## Overview
+
+Alpaca provides long-history, high-frequency US market data across equities and
+crypto over a single historical REST API. One provider serves both asset
+classes: plain tickers route to the stock bars endpoint, `BASE/QUOTE` symbols
+route to the crypto bars endpoint.
+
+**Best For**: Free US intraday equities, US crypto bars
+
+**Pricing**:
+| Tier | Price | Features |
+|------|-------|----------|
+| Basic | $0/mo | 200 req/min, real-time IEX feed, no recent-15-min SIP access |
+| Algo Trader Plus | $99/mo | 10,000 req/min, full SIP (consolidated tape) |
+
+---
+
+## Quick Start
+
+```python
+import os
+os.environ["ALPACA_API_KEY"] = "your_key_here"
+os.environ["ALPACA_API_SECRET"] = "your_secret_here"
+
+from ml4t.data.providers import AlpacaDataProvider
+
+provider = AlpacaDataProvider()
+
+# US stocks
+df = provider.fetch_ohlcv("AAPL", "2024-01-01", "2024-12-01")
+
+# Crypto (BASE/QUOTE symbol routes to the crypto endpoint)
+df = provider.fetch_ohlcv("BTC/USD", "2024-01-01", "2024-01-31")
+
+# Intraday with RFC-3339 datetime bounds
+df = provider.fetch_ohlcv(
+    "BTC/USD", "2024-01-01T00:00:00Z", "2024-01-01T01:00:00Z", frequency="minute"
+)
+
+# Multi-year intraday backfills with minute multiples (5m / 15m / 30m)
+df = provider.fetch_ohlcv("AAPL", "2021-01-01", "2024-12-31", frequency="15m")
+
+provider.close()
+```
+
+Async usage:
+
+```python
+async with AlpacaDataProvider() as provider:
+    df = await provider.fetch_ohlcv_async("AAPL", "2024-01-01", "2024-12-01")
+```
+
+---
+
+## Symbol Format
+
+| Asset class | Format | Examples |
+|-------------|--------|----------|
+| US stocks | Plain ticker | AAPL, MSFT |
+| Crypto | BASE/QUOTE | BTC/USD, ETH/USD |
+
+Symbols are uppercased into requests and into the output `symbol` column; the
+crypto slash is preserved (e.g. `BTC/USD`).
+
+---
+
+## Supported Frequencies
+
+| Frequency | Availability |
+|-----------|--------------|
+| `daily` / `1d` | ✅ |
+| `hourly` / `1h` | ✅ |
+| `minute` / `1m` | ✅ |
+| `5m` / `5minute` | ✅ |
+| `15m` / `15minute` | ✅ |
+| `30m` / `30minute` | ✅ |
+
+`start`/`end` accept `YYYY-MM-DD` dates or RFC-3339 datetimes (both inclusive);
+datetime bounds are the natural shape for sub-day minute/hour windows.
+
+---
+
+## Feeds and Adjustment
+
+- `feed="iex"` (default): the free feed, real-time but served from a single
+  exchange (IEX, roughly 2-3% of US volume).
+- `feed="sip"`: the consolidated tape (100% of US volume, all exchanges).
+  Real-time on paid plans; the free Basic plan can still query SIP bars, only
+  not the most recent 15 minutes.
+- `feed="otc"` (over-the-counter) and `feed="boats"` (Blue Ocean overnight) are
+  also accepted. Any other value raises `DataValidationError` at construction.
+- `adjustment="raw"` (default, Alpaca's own default): stock bars are **not**
+  adjusted for splits or dividends. Pass `adjustment="split"`, `"dividend"`, or
+  `"all"` for adjusted bars. Crypto has no adjustment concept.
+
+```python
+provider = AlpacaDataProvider(feed="sip", adjustment="all")
+```
+
+---
+
+## API Key Setup
+
+```bash
+# .env file (project convention)
+ALPACA_API_KEY=your_key_here
+ALPACA_API_SECRET=your_secret_here
+```
+
+Alpaca's own SDK/CLI names `APCA_API_KEY_ID` / `APCA_API_SECRET_KEY` are also
+accepted. Get a free key at [alpaca.markets](https://alpaca.markets/).
+
+---
+
+## Rate Limits
+
+| Tier | Limit |
+|------|-------|
+| Basic (free) | 200 req/min |
+| Algo Trader Plus | 10,000 req/min |
+
+The provider throttles client-side to 200 req/min by default (override with the
+`rate_limit` constructor argument), honors 429 `Retry-After`/rate-limit-reset
+headers, and retries transient failures per pagination page.
+
+---
+
+## Not Yet Implemented
+
+| Feature | Priority |
+|---------|----------|
+| Quotes / trades (tick) endpoints | MEDIUM |
+| Options bars | LOW |
+| News API | LOW |
+
+---
+
+## See Also
+
+- [Alpaca Market Data docs](https://docs.alpaca.markets/us/docs/about-market-data-api)
+- [Provider README](README.md)
+- [Provider Audit](PROVIDER_AUDIT.md)
diff --git a/examples/01_microstructure_tick_analysis.py b/examples/01_microstructure_tick_analysis.py
@@ -21,12 +21,10 @@
 import matplotlib.pyplot as plt
 import numpy as np
 import polars as pl
+
 from ml4t.engineer.bars import DollarBar, VolumeBar
 from ml4t.engineer.features import microstructure as ms
 from ml4t.engineer.labeling import BarrierConfig, triple_barrier_labels
-
-# Import qeval modules
-# Import qfeatures modules
 from ml4t.evaluation import Evaluator, Tier
 from ml4t.evaluation.splitters import PurgedWalkForwardCV
 

diff --git a/examples/02_cross_sectional_nasdaq100.py b/examples/02_cross_sectional_nasdaq100.py
@@ -22,11 +22,10 @@
 from pathlib import Path
 
 import matplotlib.pyplot as plt
-
-# Import qfeatures modules
-import ml4t.engineer as qf
 import polars as pl
 import seaborn as sns
+
+import ml4t.engineer as qf
 from ml4t.evaluation import Evaluator, Tier
 from ml4t.evaluation.evaluation.stats import deflated_sharpe_ratio
 from ml4t.evaluation.evaluation.viz import create_factor_tearsheet

diff --git a/mkdocs.yml b/mkdocs.yml
@@ -124,6 +124,7 @@ nav:
   - Providers:
     - providers/index.md
     - Yahoo Finance: providers/yahoo.md
+    - Alpaca: providers/alpaca.md
     - CoinGecko: providers/coingecko.md
     - FRED: providers/fred.md
     - Fama-French: providers/fama_french.md

diff --git a/pyproject.toml b/pyproject.toml
@@ -236,6 +236,13 @@ ignore = [
     "B904",  # exception chaining (not always needed)
 ]
 
+[tool.ruff.lint.isort]
+# Treat the whole ml4t namespace (data, engineer, evaluation, ...) as
+# first-party so its imports always form their own third section after the
+# standard library and third-party blocks, even for ml4t distributions
+# installed from outside this repo.
+known-first-party = ["ml4t"]
+
 [tool.ruff.lint.per-file-ignores]
 "tests/*" = ["ARG001", "ARG002", "ARG005", "B017", "F821", "SIM117"]  # Test patterns
 "src/ml4t/data/futures/continuous_downloader.py" = ["ARG002"]  # Public API placeholder

diff --git a/src/ml4t/data/config/models.py b/src/ml4t/data/config/models.py
@@ -53,6 +53,7 @@ class ProviderType(str, Enum):
     """Provider type enumeration."""
 
     YAHOO = "yahoo"
+    ALPACA = "alpaca"
     BINANCE = "binance"
     CRYPTOCOMPARE = "cryptocompare"
     DATABENTO = "databento"

diff --git a/src/ml4t/data/config/validator.py b/src/ml4t/data/config/validator.py
@@ -65,11 +65,21 @@ def _validate_providers(self) -> None:
             provider_names.add(provider.name)
 
             # Check API keys for providers that need them
-            if provider.type in ["massive", "polygon", "cryptocompare"] and not provider.api_key:
+            if (
+                provider.type in ["massive", "polygon", "cryptocompare", "alpaca"]
+                and not provider.api_key
+            ):
                 self.warnings.append(
                     f"Provider {provider.name} ({provider.type}) may require an API key"
                 )
 
+            # Alpaca authenticates with a key/secret pair, so a missing secret
+            # is just as fatal as a missing key.
+            if provider.type == "alpaca" and not provider.api_secret:
+                self.warnings.append(
+                    f"Provider {provider.name} (alpaca) requires api_secret as well as api_key"
+                )
+
             # Validate rate limits
             if provider.rate_limit and provider.rate_limit.requests_per_second <= 0:
                 self.errors.append(

diff --git a/src/ml4t/data/managers/config_manager.py b/src/ml4t/data/managers/config_manager.py
@@ -40,8 +40,14 @@ class ConfigManager:
         >>> print(config_mgr.config["providers"]["yahoo"])
     """
 
-    # Environment variable to provider config mapping
+    # Environment variable to provider config mapping. Later entries win when
+    # two variables map to the same field, so the APCA_* aliases (Alpaca's own
+    # SDK/CLI names) come before the ALPACA_* project-convention names.
     ENV_MAPPING = {
+        "APCA_API_KEY_ID": ("alpaca", "api_key"),
+        "APCA_API_SECRET_KEY": ("alpaca", "api_secret"),
+        "ALPACA_API_KEY": ("alpaca", "api_key"),
+        "ALPACA_API_SECRET": ("alpaca", "api_secret"),
         "CRYPTOCOMPARE_API_KEY": ("cryptocompare", "api_key"),
         "DATABENTO_API_KEY": ("databento", "api_key"),
         "POLYGON_API_KEY": ("massive", "api_key"),