scitex 2.14.0__py3-none-any.whl → 2.15.3__py3-none-any.whl

scitex/ai/classification/timeseries/_sliding_window_plotting.py

@@ -0,0 +1,369 @@
+#!/usr/bin/env python3
+# Timestamp: "2026-01-24 (ywatanabe)"
+# File: /home/ywatanabe/proj/scitex-python/src/scitex/ai/classification/timeseries/_sliding_window_plotting.py
+
+"""Plotting mixin for TimeSeriesSlidingWindowSplit visualization."""
+
+from __future__ import annotations
+
+from typing import TYPE_CHECKING
+
+import matplotlib.patches as patches
+import matplotlib.pyplot as plt
+import numpy as np
+
+import scitex as stx
+
+if TYPE_CHECKING:
+    from ._sliding_window_core import TimeSeriesSlidingWindowSplitCore
+
+__all__ = ["SlidingWindowPlottingMixin"]
+
+COLORS = stx.plt.color.PARAMS
+
+
+class SlidingWindowPlottingMixin:
+    """Mixin class providing plot_splits visualization for sliding window splitters."""
+
+    def plot_splits(
+        self: TimeSeriesSlidingWindowSplitCore,
+        X,
+        y=None,
+        timestamps=None,
+        figsize=(12, 6),
+        save_path=None,
+    ):
+        """Visualize the sliding window splits as rectangles.
+
+        Shows train (blue), validation (green), and test (red) sets.
+        When val_ratio=0, only shows train and test.
+        When undersampling is enabled, shows dropped samples in gray.
+
+        Parameters
+        ----------
+        X : array-like
+            Training data
+        y : array-like, optional
+            Target variable (required for undersampling visualization)
+        timestamps : array-like, optional
+            Timestamps (if None, uses sample indices)
+        figsize : tuple, default (12, 6)
+            Figure size
+        save_path : str, optional
+            Path to save the plot
+
+        Returns
+        -------
+        fig : matplotlib.figure.Figure
+            The created figure
+        """
+        if timestamps is None:
+            timestamps = np.arange(len(X))
+
+        time_order = np.argsort(timestamps)
+
+        # Get splits WITH undersampling (if enabled)
+        if self.val_ratio > 0:
+            splits = list(self.split_with_val(X, y, timestamps))[:10]
+            split_type = "train-val-test"
+        else:
+            splits = list(self.split(X, y, timestamps))[:10]
+            split_type = "train-test"
+
+        if not splits:
+            raise ValueError("No splits generated")
+
+        # Get splits WITHOUT undersampling to show dropped samples
+        splits_no_undersample = None
+        if self.undersample and y is not None:
+            original_undersample = self.undersample
+            self.undersample = False
+            if self.val_ratio > 0:
+                splits_no_undersample = list(self.split_with_val(X, y, timestamps))[:10]
+            else:
+                splits_no_undersample = list(self.split(X, y, timestamps))[:10]
+            self.undersample = original_undersample
+
+        fig, ax = stx.plt.subplots(figsize=figsize)
+
+        # Plot each fold
+        for fold, split_indices in enumerate(splits):
+            y_pos = fold
+            self._plot_fold_rectangles(ax, split_indices, time_order, y_pos, fold)
+
+        # Plot dropped samples if undersampling
+        if splits_no_undersample is not None:
+            self._plot_dropped_samples(ax, splits, splits_no_undersample, time_order, y)
+
+        # Plot kept samples
+        self._plot_kept_samples(ax, splits, time_order, y)
+
+        # Format plot
+        self._format_plot(ax, X, splits, split_type, y)
+
+        plt.tight_layout()
+
+        if save_path:
+            fig.savefig(save_path, dpi=150, bbox_inches="tight")
+
+        return fig
+
+    def _plot_fold_rectangles(self, ax, split_indices, time_order, y_pos, fold):
+        """Plot train/val/test rectangles for a fold."""
+        if len(split_indices) == 3:
+            train_idx, val_idx, test_idx = split_indices
+            self._plot_train_rect(ax, train_idx, time_order, y_pos, fold)
+            if len(val_idx) > 0:
+                self._plot_val_rect(ax, val_idx, time_order, y_pos, fold)
+            self._plot_test_rect(ax, test_idx, time_order, y_pos, fold, is_3way=True)
+        else:
+            train_idx, test_idx = split_indices
+            self._plot_train_rect(ax, train_idx, time_order, y_pos, fold)
+            self._plot_test_rect(ax, test_idx, time_order, y_pos, fold, is_3way=False)
+
+    def _plot_train_rect(self, ax, train_idx, time_order, y_pos, fold):
+        """Plot training window rectangle."""
+        train_positions = [np.where(time_order == idx)[0][0] for idx in train_idx]
+        if train_positions:
+            train_start = min(train_positions)
+            train_end = max(train_positions)
+            train_rect = patches.Rectangle(
+                (train_start, y_pos - 0.3),
+                train_end - train_start + 1,
+                0.6,
+                linewidth=1,
+                edgecolor="blue",
+                facecolor="lightblue",
+                alpha=0.7,
+                label="Train" if fold == 0 else "",
+            )
+            ax.add_patch(train_rect)
+
+    def _plot_val_rect(self, ax, val_idx, time_order, y_pos, fold):
+        """Plot validation window rectangle."""
+        val_positions = [np.where(time_order == idx)[0][0] for idx in val_idx]
+        if val_positions:
+            val_start = min(val_positions)
+            val_end = max(val_positions)
+            val_rect = patches.Rectangle(
+                (val_start, y_pos - 0.3),
+                val_end - val_start + 1,
+                0.6,
+                linewidth=1,
+                edgecolor="green",
+                facecolor="lightgreen",
+                alpha=0.7,
+                label="Validation" if fold == 0 else "",
+            )
+            ax.add_patch(val_rect)
+
+    def _plot_test_rect(self, ax, test_idx, time_order, y_pos, fold, is_3way):
+        """Plot test window rectangle."""
+        test_positions = [np.where(time_order == idx)[0][0] for idx in test_idx]
+        if test_positions:
+            test_start = min(test_positions)
+            test_end = max(test_positions)
+            if is_3way:
+                edgecolor = COLORS["RGBA_NORM"]["red"]
+                facecolor = COLORS["RGBA_NORM"]["red"]
+            else:
+                edgecolor = "red"
+                facecolor = "lightcoral"
+            test_rect = patches.Rectangle(
+                (test_start, y_pos - 0.3),
+                test_end - test_start + 1,
+                0.6,
+                linewidth=1,
+                edgecolor=edgecolor,
+                facecolor=facecolor,
+                alpha=0.7,
+                label="Test" if fold == 0 else "",
+            )
+            ax.add_patch(test_rect)
+
+    def _plot_dropped_samples(self, ax, splits, splits_no_undersample, time_order, y):
+        """Plot dropped samples from undersampling."""
+        np.random.seed(42)
+        jitter_strength = 0.15
+
+        for fold, split_indices_no_us in enumerate(splits_no_undersample):
+            y_pos = fold
+            split_indices_us = splits[fold]
+
+            if len(split_indices_no_us) == 3:
+                train_idx_no_us, val_idx_no_us, _ = split_indices_no_us
+                train_idx_us, val_idx_us, _ = split_indices_us
+
+                dropped_train = np.setdiff1d(train_idx_no_us, train_idx_us)
+                if len(dropped_train) > 0:
+                    positions = [
+                        np.where(time_order == idx)[0][0] for idx in dropped_train
+                    ]
+                    jitter = np.random.normal(0, jitter_strength, len(positions))
+                    ax.scatter(
+                        positions,
+                        y_pos + jitter,
+                        c="gray",
+                        s=15,
+                        alpha=0.3,
+                        marker="x",
+                        label="Dropped (train)" if fold == 0 else "",
+                        zorder=2,
+                    )
+
+                dropped_val = np.setdiff1d(val_idx_no_us, val_idx_us)
+                if len(dropped_val) > 0:
+                    positions = [
+                        np.where(time_order == idx)[0][0] for idx in dropped_val
+                    ]
+                    jitter = np.random.normal(0, jitter_strength, len(positions))
+                    ax.scatter(
+                        positions,
+                        y_pos + jitter,
+                        c="gray",
+                        s=15,
+                        alpha=0.3,
+                        marker="x",
+                        label="Dropped (val)" if fold == 0 else "",
+                        zorder=2,
+                    )
+            else:
+                train_idx_no_us, _ = split_indices_no_us
+                train_idx_us, _ = split_indices_us
+
+                dropped_train = np.setdiff1d(train_idx_no_us, train_idx_us)
+                if len(dropped_train) > 0:
+                    positions = [
+                        np.where(time_order == idx)[0][0] for idx in dropped_train
+                    ]
+                    jitter = np.random.normal(0, jitter_strength, len(positions))
+                    ax.scatter(
+                        positions,
+                        y_pos + jitter,
+                        c="gray",
+                        s=15,
+                        alpha=0.3,
+                        marker="x",
+                        label="Dropped samples" if fold == 0 else "",
+                        zorder=2,
+                    )
+
+    def _plot_kept_samples(self, ax, splits, time_order, y):
+        """Plot kept samples with colors."""
+        np.random.seed(42)
+        jitter_strength = 0.15
+
+        for fold, split_indices in enumerate(splits):
+            y_pos = fold
+
+            if len(split_indices) == 3:
+                train_idx, val_idx, test_idx = split_indices
+                self._scatter_indices(
+                    ax, train_idx, time_order, y_pos, y, fold, "train"
+                )
+                if len(val_idx) > 0:
+                    self._scatter_indices(
+                        ax, val_idx, time_order, y_pos, y, fold, "val"
+                    )
+                self._scatter_indices(ax, test_idx, time_order, y_pos, y, fold, "test")
+            else:
+                train_idx, test_idx = split_indices
+                self._scatter_indices(
+                    ax, train_idx, time_order, y_pos, y, fold, "train"
+                )
+                self._scatter_indices(ax, test_idx, time_order, y_pos, y, fold, "test")
+
+    def _scatter_indices(self, ax, indices, time_order, y_pos, y, fold, set_type):
+        """Scatter plot indices with jittering."""
+        jitter_strength = 0.15
+        positions = [np.where(time_order == idx)[0][0] for idx in indices]
+
+        if not positions:
+            return
+
+        jitter = np.random.normal(0, jitter_strength, len(positions))
+
+        color_map = {
+            "train": ("darkblue", "o", "Train"),
+            "val": ("darkgreen", "^", "Val"),
+            "test": ("darkred", "s", "Test"),
+        }
+
+        if y is not None:
+            class_colors = {
+                "train": (
+                    COLORS["RGBA_NORM"]["blue"],
+                    COLORS["RGBA_NORM"]["lightblue"],
+                ),
+                "val": (COLORS["RGBA_NORM"]["yellow"], COLORS["RGBA_NORM"]["orange"]),
+                "test": (COLORS["RGBA_NORM"]["red"], COLORS["RGBA_NORM"]["brown"]),
+            }
+            colors = [
+                class_colors[set_type][0] if y[idx] == 0 else class_colors[set_type][1]
+                for idx in indices
+            ]
+            ax.scatter(
+                positions,
+                y_pos + jitter,
+                c=colors,
+                s=20,
+                alpha=0.7,
+                marker=color_map[set_type][1],
+                label=f"{color_map[set_type][2]} (class 0)" if fold == 0 else "",
+                zorder=3,
+            )
+        else:
+            ax.scatter(
+                positions,
+                y_pos + jitter,
+                c=color_map[set_type][0],
+                s=20,
+                alpha=0.7,
+                marker=color_map[set_type][1],
+                label=f"{color_map[set_type][2]} points" if fold == 0 else "",
+                zorder=3,
+            )
+
+    def _format_plot(self, ax, X, splits, split_type, y):
+        """Format the plot with labels and legend."""
+        ax.set_ylim(-0.5, len(splits) - 0.5)
+        ax.set_xlim(0, len(X))
+        ax.set_xlabel("Temporal Position (sorted by timestamp)")
+        ax.set_ylabel("Fold")
+
+        gap_text = f", Gap: {self.gap}" if self.gap > 0 else ""
+        val_text = f", Val ratio: {self.val_ratio:.1%}" if self.val_ratio > 0 else ""
+        ax.set_title(
+            f"Sliding Window Split Visualization ({split_type})\\n"
+            f"Window: {self.window_size}, Step: {self.step_size}, Test: {self.test_size}{gap_text}{val_text}\\n"
+            f"Rectangles show windows, dots show actual data points"
+        )
+
+        ax.set_yticks(range(len(splits)))
+        ax.set_yticklabels([f"Fold {i}" for i in range(len(splits))])
+
+        if y is not None:
+            unique_classes, class_counts = np.unique(y, return_counts=True)
+            total_class_info = ", ".join(
+                [
+                    f"Class {cls}: n={count}"
+                    for cls, count in zip(unique_classes, class_counts)
+                ]
+            )
+
+            first_split = splits[0]
+            if len(first_split) == 3:
+                train_idx, val_idx, test_idx = first_split
+                fold_info = f"Fold 0: Train n={len(train_idx)}, Val n={len(val_idx)}, Test n={len(test_idx)}"
+            else:
+                train_idx, test_idx = first_split
+                fold_info = f"Fold 0: Train n={len(train_idx)}, Test n={len(test_idx)}"
+
+            handles, labels = ax.get_legend_handles_labels()
+            legend_title = f"Total: {total_class_info}\\n{fold_info}"
+            ax.legend(handles, labels, loc="upper right", title=legend_title)
+        else:
+            ax.legend(loc="upper right")
+
+
+# EOF

scitex/audio/README.md

@@ -27,6 +27,8 @@ scitex audio speak "Hello world"
 scitex audio speak "Bonjour" --backend gtts --voice fr
 scitex audio backends       # List available backends
 scitex audio check          # Check audio status (WSL)
+scitex audio relay          # Start HTTP relay server (for remote audio)
+scitex audio serve          # Start MCP server
 ```
 
 ## MCP Server
@@ -46,75 +48,77 @@ Add to `~/.claude/mcp.json`:
 }
 ```
 
-### Remote Audio (SSH tunnel)
+### Remote Audio (HTTP Relay)
 
-Enable remote agents to play audio on local speakers.
+Enable remote agents to play audio on local speakers using a simple HTTP relay.
 
 **Architecture:**
 ```
 ┌─────────────────────────┐              ┌─────────────────────────┐
 │  Remote (e.g., NAS)     │              │  Local (WSL/Windows)    │
 │                         │              │                         │
-│  Claude Agent ──────────┼─ SSH ───────▶│  scitex audio serve     │
-│  connects to            │ RemoteForward│  -t sse --port 8084     │
-│  localhost:8084         │              │         │               │
-│                         │              │         ▼               │
+│  Claude Agent uses      │              │  scitex audio relay     │
+│  audio_speak_relay ─────┼─ SSH ───────▶│  --port 31293           │
+│                         │ Reverse      │         │               │
+│  localhost:31293        │ Tunnel       │         ▼               │
 │                         │              │     🔊 Speakers         │
 └─────────────────────────┘              └─────────────────────────┘
 ```
 
-**Step 1: Local machine - Start audio server**
+**Step 1: Local machine - Start relay server**
 ```bash
-scitex audio serve -t sse --port 8084
+scitex audio relay --port 31293
 ```
 
-**Step 2: SSH config - Add RemoteForward**
+**Step 2: SSH with reverse tunnel**
+```bash
+ssh -R 31293:localhost:31293 remote-server
+```
 
-In `~/.ssh/config` (on local machine):
+Or add to `~/.ssh/config`:
 ```
 Host nas
   HostName 192.168.x.x
   User youruser
-  RemoteForward 8084 127.0.0.1:8084  # Audio: remote -> local speakers
+  RemoteForward 31293 127.0.0.1:31293
 ```
 
-**Step 3: Remote machine - MCP config**
+**Step 3: Remote agent uses relay**
 
-On remote machine, add to `~/.claude/mcp.json`:
-```json
-{
-  "mcpServers": {
-    "scitex-audio-remote": {
-      "type": "sse",
-      "url": "http://localhost:8084/sse"
-    }
-  }
-}
-```
+The `audio_speak_relay` MCP tool auto-detects:
+1. `SCITEX_AUDIO_RELAY_URL` env var
+2. Localhost:31293 (SSH reverse tunnel)
+3. SSH_CLIENT IP (auto-detected from SSH session)
 
-**Step 4: Connect via SSH**
-```bash
-ssh nas  # RemoteForward creates tunnel automatically
-```
+### Environment Variables
 
-Now Claude agents on the remote machine can use `mcp__scitex-audio-remote__speak` to play audio on your local speakers.
+| Variable | Default | Description |
+|----------|---------|-------------|
+| `SCITEX_AUDIO_PORT` | 31293 | Server/relay port |
+| `SCITEX_AUDIO_MODE` | auto | `local`, `remote`, or `auto` |
+| `SCITEX_AUDIO_RELAY_URL` | (auto) | Full relay URL |
+| `SCITEX_AUDIO_RELAY_HOST` | (none) | Relay host |
+| `SCITEX_AUDIO_RELAY_PORT` | 31293 | Relay port |
 
 ### Server Transports
 
 | Transport | Command | Use Case |
 |-----------|---------|----------|
 | stdio | `scitex audio serve` | Claude Desktop (default) |
-| sse | `scitex audio serve -t sse --port 8084` | Remote agents via SSH |
-| http | `scitex audio serve -t http --port 8084` | HTTP clients |
+| sse | `scitex audio serve -t sse --port 31293` | Remote MCP agents |
+| http | `scitex audio serve -t http --port 31293` | HTTP MCP clients |
+| relay | `scitex audio relay --port 31293` | Simple HTTP relay |
 
-### Tools
+### MCP Tools
 
 | Tool | Description |
 |------|-------------|
-| `speak` | Text to speech (supports `rate`, `speed` params) |
-| `list_backends` | Show available backends |
-| `check_audio_status` | Check WSL audio connectivity |
-| `announce_context` | Announce current directory and git branch |
+| `audio_speak` | Text to speech (plays on server) |
+| `audio_speak_local` | TTS on server machine |
+| `audio_speak_relay` | TTS via relay (remote playback) |
+| `audio_list_backends` | Show available backends |
+| `audio_check_audio_status` | Check WSL audio connectivity |
+| `audio_announce_context` | Announce current directory and git branch |
 
 ## Backends
 
@@ -126,4 +130,4 @@ Now Claude agents on the remote machine can use `mcp__scitex-audio-remote__speak
 
 ## Cross-Process Locking
 
-The MCP server uses FIFO locking to ensure only one audio plays at a time across all Claude Code sessions. This prevents audio overlap when multiple agents are running.
+The relay server uses FIFO locking to ensure only one audio plays at a time across all Claude Code sessions. This prevents audio overlap when multiple agents are running.