Add price signals and cross-symbol signal support by michaelchu · Pull Request #233 · goldspanlabs/optopsy

michaelchu · 2026-02-27T23:34:05Z

Summary

Add 14 new first-class price signals in optopsy/signals/price.py: price level (above/below/cross), gaps, N-period breakouts, daily returns, drawdown/rally from high/low, and consecutive up/down streaks
Add cross-symbol signal support via signal_symbol parameter on build_signal and build_custom_signal, enabling use cases like "use VIX data to trigger entries on SPX options"
Update agent system prompt, tool schemas, and model descriptions to document all new signals and the cross-symbol workflow

Test plan

28 new tests in tests/test_signals_price.py covering all 14 signals + edge cases
Full test suite passes (1862 tests)
ruff check and ruff format clean
ty type checker passes

🤖 Generated with Claude Code

Add 14 new price-related signals in optopsy/signals/price.py: - Price level: price_above, price_below, price_cross_above, price_cross_below - Gaps: gap_up, gap_down - Breakouts: high_of_n_days, low_of_n_days - Returns: daily_return_above, daily_return_below - Drawdown: drawdown_from_high, rally_from_low - Streaks: consecutive_up, consecutive_down Add cross-symbol signal support via signal_symbol parameter on build_signal and build_custom_signal, enabling use cases like "use VIX data to trigger entries on SPX options." Update agent system prompt, tool schemas, and signal descriptions to document all new signals and cross-symbol workflow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Copilot

Pull request overview

Adds a new suite of first-class “price” signals to the optopsy.signals API and extends the UI toolchain to support cross-symbol signal computation (compute signals on one ticker’s OHLCV and apply them to another ticker’s options dataset).

Changes:

Introduces 14 new price-based signals (levels, crossovers, gaps, breakouts, returns, drawdown/rally, streaks) and exports them from optopsy.signals.
Adds signal_symbol support to build_signal / build_custom_signal, including yfinance fetching for an explicit symbol and remapping signal dates to the options dataset symbols.
Updates tool registry/schemas/models and agent prompt docs; adds a new test module covering the new price signals.

Reviewed changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
`tests/test_signals_price.py`	Adds signal-level pytest coverage for the new price signal functions.
`optopsy/signals/price.py`	Implements the new price signals (state + event style signals).
`optopsy/signals/_helpers.py`	Adds `_get_open()` helper used by gap signals.
`optopsy/signals/__init__.py`	Exports new price signals (and `_get_open`) from the package API.
`optopsy/ui/tools/_signals_builder.py`	Adds `signal_symbol` support to `build_signal` / `build_custom_signal` and remaps cross-symbol signal dates.
`optopsy/ui/tools/_helpers.py`	Adds `_fetch_stock_data_for_symbol()` to fetch OHLCV for a single ticker for cross-symbol workflows.
`optopsy/ui/tools/_schemas.py`	Registers the new signal names + params and marks relevant ones as OHLC-requiring.
`optopsy/ui/tools/_models.py`	Documents new signals and adds `signal_symbol` to the tool argument models.
`optopsy/ui/agent.py`	Updates the agent system prompt documentation for the new signals and cross-symbol workflow.

- Strip whitespace from signal_symbol to prevent cache key issues - Gate cross-symbol remapping on needs_stock to avoid duplicating rows when signal_symbol is set on IV or date-only signals Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Extract _fetch_single_symbol_stock from the duplicated fetch logic in _fetch_stock_data_for_signals and _fetch_stock_data_for_symbol - Extract _remap_cross_symbol_dates from the duplicated remap blocks in _handle_build_signal and _handle_build_custom_signal - Extract _STOCK_COLS constant for the shared column list Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Copilot

Pull request overview

Copilot reviewed 9 out of 9 changed files in this pull request and generated 5 comments.

- Add _OPEN_SIGNALS set and validate open column for gap_up/gap_down - Add days >= 1 validation to consecutive_up/consecutive_down - Make gap and high_of_n_days tests assert exact firing indices Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Copilot

Pull request overview

Copilot reviewed 9 out of 9 changed files in this pull request and generated 4 comments.

Copilot · 2026-02-28T00:09:01Z

+_STOCK_COLS = [
+    "underlying_symbol",
+    "quote_date",
+    "open",
+    "high",
+    "low",
+    "close",
+    "volume",
+]
+
+
+def _fetch_single_symbol_stock(
+    symbol: str,
+    padded_start: "date",
+    date_max: "date",
+) -> pd.DataFrame | None:
+    """Fetch and slice OHLCV data for one symbol.
+
+    Reads from (or populates) the yfinance parquet cache, then slices to
+    ``[padded_start, date_max]``.  Returns a DataFrame with ``_STOCK_COLS``
+    or ``None`` on failure.
+    """
+    try:
+        cached = _yf_cache.read(_YF_CACHE_CATEGORY, symbol)
+        cached = _yf_fetch_and_cache(symbol, cached, date_max)
+
+        if cached is None or cached.empty:
+            return None
+
+        result = cached[
+            (pd.to_datetime(cached["date"]).dt.date >= padded_start)
+            & (pd.to_datetime(cached["date"]).dt.date <= date_max)
+        ].rename(columns={"date": "quote_date"})
+        if result.empty:
+            return None
+        return result[_STOCK_COLS]
+    except (OSError, ValueError, KeyError, pd.errors.ParserError) as exc:
+        _log.warning("yfinance fetch failed for %s: %s", symbol, exc)
+        return None


_fetch_single_symbol_stock() always returns result[_STOCK_COLS], but _normalise_yf_df() can legitimately omit optional columns (e.g. volume / open / high / low) when yfinance doesn’t provide them. In that case this selection raises KeyError, causing the fetch to fail even for signals that only need close (and preventing the more specific missing-column validations from triggering). Consider returning the intersection of available columns while enforcing a minimal required set (at least underlying_symbol, quote_date, close), and letting callers validate additional columns as needed.

Valid concern. _fetch_single_symbol_stock now selects only available columns ([c for c in _STOCK_COLS if c in result.columns]) instead of hard-coding all 7, and returns None only if close is missing. This lets signals that only need close work even when yfinance omits optional columns, while letting the caller-level validations (_OHLC_SIGNALS, _OPEN_SIGNALS, _VOLUME_SIGNALS) report specific missing-column errors. Fixed in 79c8c24.

Copilot · 2026-02-28T00:09:02Z

 # Signals that require OHLC data (high, low, close) — need stock data fetch.
 _OHLC_SIGNALS = frozenset(
    {
        "stoch_below",
        "stoch_above",
        "willr_below",
        "willr_above",
        "cci_below",
        "cci_above",
        "uo_above",
        "uo_below",
        "squeeze_on",
        "squeeze_off",
        "ao_above",
        "ao_below",
        "kc_above_upper",
        "kc_below_lower",
        "donchian_above_upper",
        "donchian_below_lower",
        "natr_above",
        "natr_below",
        "massi_above",
        "massi_below",
        "adx_above",
        "adx_below",
        "aroon_cross_above",
        "aroon_cross_below",
        "supertrend_buy",
        "supertrend_sell",
        "psar_buy",
        "psar_sell",
        "chop_above",
        "chop_below",
+        "gap_up",
+        "gap_down",
+        "high_of_n_days",
+        "low_of_n_days",
    }


_OHLC_SIGNALS is documented/validated as requiring high, low, and close, but it now includes gap_up/gap_down and high_of_n_days/low_of_n_days. The new price signals don’t actually need all of high+low (and gap signals don’t use high/low at all), so the build_signal OHLC validation can reject otherwise-computable signals and produce a misleading error. Consider splitting these into more precise requirement sets (e.g. close-only vs open+close vs full OHLC) and validating only the columns each signal truly needs.

Good point. Removed gap_up/gap_down from _OHLC_SIGNALS — they don't need high/low and already have their own _OPEN_SIGNALS validation checking for the open column. high_of_n_days/low_of_n_days remain in _OHLC_SIGNALS since they genuinely use the high/low columns. Fixed in 79c8c24.

Copilot · 2026-02-28T00:09:02Z

+        result = high_of_n_days(period=5)(data)
+        # Bar 7: close=103, prev 5-bar high max = max(103,99,98,100,104)
+        # Wait — rolling high is on the high column, shifted by 1
+        # Bar 7 rolling(5).max().shift(1) = max of bars 2-6 highs = max(101,103,99,98,100) = 103
+        # close=103 >= 103 → True
+        assert result.iloc[7] == True
+        # Bar 9: close=105, prev 5-bar high = max of bars 4-8 = max(99,98,100,104,103) = 104
+        # 105 >= 104 → True
+        assert result.iloc[9] == True


test_high_of_n_days_breakout only asserts that a couple of bars are True, so it would still pass if high_of_n_days() incorrectly fires on additional bars (false positives). To make the test more regression-proof, consider asserting the full expected boolean vector or the exact list of firing indices for this fixture.

Agreed. Updated to assert exact firing indices: assert fire_indices == [1, 3, 7, 9]. Added per-bar computation comments for clarity. Fixed in 79c8c24.

Copilot · 2026-02-28T00:09:02Z

+        result = low_of_n_days(period=5)(data)
+        # Bar 7: close=96, prev 5-bar low = min of bars 2-6 lows = min(99,101,97,96,98) = 96
+        # 96 <= 96 → True
+        assert result.iloc[7] == True
+


test_low_of_n_days_breakdown currently asserts only one True bar. This can miss regressions where low_of_n_days() produces extra True bars. Consider asserting the exact firing indices (or full boolean vector) for the entire series so the test fails on false positives as well.

Agreed. Updated to assert exact firing indices: assert fire_indices == [4, 5, 7, 8, 9]. Added per-bar computation comments. Fixed in 79c8c24.

- _fetch_single_symbol_stock now selects only available columns from _STOCK_COLS instead of hard-coding all 7, preventing KeyError when yfinance omits optional columns (open/high/low/volume) - Remove gap_up/gap_down from _OHLC_SIGNALS (they need open+close, not high/low); they already have their own _OPEN_SIGNALS validation - Make high_of_n_days and low_of_n_days tests assert exact fire indices Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Copilot

Pull request overview

Copilot reviewed 9 out of 9 changed files in this pull request and generated 1 comment.

michaelchu requested a review from Copilot February 27, 2026 23:34

Copilot AI reviewed Feb 27, 2026

View reviewed changes

Comment thread optopsy/ui/tools/_signals_builder.py

Comment thread optopsy/ui/tools/_helpers.py Outdated

Comment thread optopsy/ui/tools/_signals_builder.py Outdated

Comment thread optopsy/ui/tools/_signals_builder.py

michaelchu requested a review from Copilot February 27, 2026 23:53

Copilot started reviewing on behalf of michaelchu February 27, 2026 23:53 View session

Copilot AI reviewed Feb 27, 2026

View reviewed changes

Comment thread optopsy/ui/tools/_schemas.py Outdated

Comment thread optopsy/ui/tools/_helpers.py

Comment thread tests/test_signals_price.py

Comment thread tests/test_signals_price.py

Comment thread optopsy/signals/price.py

michaelchu requested a review from Copilot February 28, 2026 00:02

Copilot started reviewing on behalf of michaelchu February 28, 2026 00:03 View session

Copilot AI reviewed Feb 28, 2026

View reviewed changes

michaelchu requested a review from Copilot February 28, 2026 00:20

Copilot started reviewing on behalf of michaelchu February 28, 2026 00:20 View session

Copilot AI reviewed Feb 28, 2026

View reviewed changes

Comment thread optopsy/ui/tools/_helpers.py

michaelchu merged commit 45e15b6 into main Feb 28, 2026
12 checks passed

michaelchu deleted the claude/price-signals-cross-symbol branch February 28, 2026 00:35

Conversation

michaelchu commented Feb 27, 2026

Summary

Test plan

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Feb 28, 2026

Choose a reason for hiding this comment

Uh oh!

michaelchu Feb 28, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 28, 2026

Choose a reason for hiding this comment

Uh oh!

michaelchu Feb 28, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 28, 2026

Choose a reason for hiding this comment

Uh oh!

michaelchu Feb 28, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 28, 2026

Choose a reason for hiding this comment

Uh oh!

michaelchu Feb 28, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants