feat: config drift detection across fleet nodes (#686)

scripts/auto-scene-descriptions.py

@@ -1,160 +0,0 @@
-#!/usr/bin/env python3
-"""
-auto-scene-descriptions.py — Generate scene descriptions from image/video assets.
-
-Scans an assets directory, uses vision model to describe each asset,
-outputs training pairs in timmy-config format.
-
-Usage:
-    python3 scripts/auto-scene-descriptions.py --scan ~/assets/
-    python3 scripts/auto-scene-descriptions.py --scan ~/assets/ --output training-data/scene-from-media.jsonl
-    python3 scripts/auto-scene-descriptions.py --scan ~/assets/ --dry-run
-"""
-
-import argparse
-import json
-import os
-import subprocess
-import sys
-from pathlib import Path
-
-IMAGE_EXTS = {".jpg", ".jpeg", ".png", ".gif", ".webp", ".bmp"}
-VIDEO_EXTS = {".mp4", ".webm", ".mov", ".avi", ".mkv"}
-SUPPORTED_EXTS = IMAGE_EXTS | VIDEO_EXTS
-
-
-def scan_assets(directory: str) -> list[Path]:
-    """Find all image/video assets in a directory."""
-    assets = []
-    for root, dirs, files in os.walk(directory):
-        dirs[:] = [d for d in dirs if d not in {".git", "node_modules", "__pycache__"}]
-        for f in sorted(files):
-            ext = Path(f).suffix.lower()
-            if ext in SUPPORTED_EXTS:
-                assets.append(Path(root) / f)
-    return assets
-
-
-def extract_video_frame(video_path: Path) -> Path | None:
-    """Extract a representative frame from a video using ffmpeg."""
-    frame_path = video_path.with_suffix(".frame.jpg")
-    try:
-        subprocess.run(
-            ["ffmpeg", "-i", str(video_path), "-vframes", "1", "-ss", "5", "-y", str(frame_path)],
-            capture_output=True, timeout=30
-        )
-        if frame_path.exists():
-            return frame_path
-    except (subprocess.TimeoutExpired, FileNotFoundError):
-        pass
-    return None
-
-
-def describe_with_ollama(image_path: Path) -> str:
-    """Generate a scene description using local Ollama vision model."""
-    try:
-        result = subprocess.run(
-            ["ollama", "run", "llava", f"Describe this image as a visual scene for a film. Include mood, colors, composition, and camera angle. Be specific and vivid in 2-3 sentences. Image: {image_path}"],
-            capture_output=True, text=True, timeout=60
-        )
-        if result.returncode == 0:
-            return result.stdout.strip()
-    except (subprocess.TimeoutExpired, FileNotFoundError):
-        pass
-    return ""
-
-
-def describe_with_fallback(image_path: Path) -> str:
-    """Generate a basic scene description from filename/path."""
-    name = image_path.stem.replace("_", " ").replace("-", " ")
-    parent = image_path.parent.name.replace("_", " ").replace("-", " ")
-    return f"A scene depicting {name} in a {parent} setting. Visual composition inferred from asset location and naming conventions."
-
-
-def build_training_pair(asset_path: Path, description: str, asset_index: int) -> dict:
-    """Build a training pair in timmy-config scene description format."""
-    name = asset_path.stem
-    parent = asset_path.parent.name
-
-    # Infer mood from path/name keywords
-    mood_keywords = {
-        "dark": "melancholic", "light": "hopeful", "warm": "nostalgic",
-        "cold": "isolated", "bright": "energetic", "sunset": "bittersweet",
-        "night": "mysterious", "morning": "refreshing", "rain": "contemplative",
-    }
-    mood = "neutral"
-    name_lower = name.lower() + parent.lower()
-    for keyword, m in mood_keywords.items():
-        if keyword in name_lower:
-            mood = m
-            break
-
-    return {
-        "song": f"asset-{asset_index:04d}",
-        "beat": 1,
-        "lyric_line": f"[Visual asset: {asset_path.name}]",
-        "scene": {
-            "mood": mood,
-            "colors": ["inferred"],
-            "composition": "frame",
-            "camera": "static",
-            "description": description,
-            "source": "auto-generated",
-            "asset_path": str(asset_path),
-        }
-    }
-
-
-def main():
-    parser = argparse.ArgumentParser(description="Auto-generate scene descriptions from assets")
-    parser.add_argument("--scan", required=True, help="Assets directory to scan")
-    parser.add_argument("--output", default=None, help="Output JSONL path")
-    parser.add_argument("--dry-run", action="store_true", help="Show what would be generated")
-    parser.add_argument("--vision", action="store_true", help="Use Ollama vision model (slow)")
-    args = parser.parse_args()
-
-    assets = scan_assets(args.scan)
-    print(f"Found {len(assets)} assets in {args.scan}", file=sys.stderr)
-
-    if not assets:
-        print("No assets found.", file=sys.stderr)
-        sys.exit(0)
-
-    pairs = []
-    for i, asset in enumerate(assets):
-        if args.vision:
-            # For videos, extract frame first
-            image_for_vision = asset
-            if asset.suffix.lower() in VIDEO_EXTS:
-                frame = extract_video_frame(asset)
-                if frame:
-                    image_for_vision = frame
-
-            description = describe_with_ollama(image_for_vision)
-            if not description:
-                description = describe_with_fallback(asset)
-        else:
-            description = describe_with_fallback(asset)
-
-        pair = build_training_pair(asset, description, i)
-        pairs.append(pair)
-
-        if args.dry_run and i < 5:
-            print(f"  {asset.name}: {description[:80]}")
-
-    if args.dry_run:
-        print(f"\nWould generate {len(pairs)} training pairs")
-        return
-
-    # Write output
-    output_path = args.output or os.path.expanduser("~/.hermes/training-data/scene-from-media.jsonl")
-    os.makedirs(os.path.dirname(output_path), exist_ok=True)
-    with open(output_path, "w") as f:
-        for p in pairs:
-            f.write(json.dumps(p, ensure_ascii=False) + "\n")
-
-    print(f"Generated {len(pairs)} scene descriptions → {output_path}", file=sys.stderr)
-
-
-if __name__ == "__main__":
-    main()

scripts/config_drift_detector.py

@@ -0,0 +1,480 @@
+#!/usr/bin/env python3
+"""
+config_drift_detector.py — Detect config drift across fleet nodes.
+
+Collects config from all wizard nodes via SSH, compares against
+canonical timmy-config golden state, and reports differences.
+
+Usage:
+  python3 scripts/config_drift_detector.py                    # Report only
+  python3 scripts/config_drift_detector.py --auto-sync        # Auto-fix drift with golden state
+  python3 scripts/config_drift_detector.py --node allegro      # Check single node
+  python3 scripts/config_drift_detector.py --json              # JSON output for automation
+
+Exit codes:
+  0 — no drift detected
+  1 — drift detected
+  2 — error (SSH failure, missing deps, etc.)
+"""
+
+import argparse
+import json
+import os
+import subprocess
+import sys
+import tempfile
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import Any
+
+import yaml
+
+# ── Constants ─────────────────────────────────────────────────────────────────
+
+SCRIPT_DIR = Path(__file__).resolve().parent
+REPO_ROOT = SCRIPT_DIR.parent
+ANSIBLE_INVENTORY = REPO_ROOT / "ansible" / "inventory" / "hosts.yml"
+GOLDEN_STATE_PLAYBOOK = REPO_ROOT / "ansible" / "playbooks" / "golden_state.yml"
+
+# Config files to check on each node
+CONFIG_PATHS = [
+    ".hermes/config.yaml",
+    "wizards/{name}/config.yaml",
+]
+
+# Keys that define golden state (from ansible inventory vars)
+GOLDEN_KEYS = [
+    "providers",
+    "provider",
+    "model",
+    "base_url",
+    "api_key_env",
+    "banned_providers",
+    "banned_models_patterns",
+]
+
+
+# ── Data Models ───────────────────────────────────────────────────────────────
+
+@dataclass
+class NodeConfig:
+    name: str
+    host: str
+    configs: dict[str, Any] = field(default_factory=dict)
+    errors: list[str] = field(default_factory=list)
+    reachable: bool = True
+
+
+@dataclass
+class DriftResult:
+    node: str
+    file_path: str
+    diff_type: str  # "missing", "value_mismatch", "key_missing", "extra_key"
+    key: str
+    canonical_value: Any = None
+    node_value: Any = None
+    severity: str = "warning"  # "info", "warning", "critical"
+
+
+# ── Inventory Parsing ─────────────────────────────────────────────────────────
+
+def load_inventory() -> dict:
+    """Load Ansible inventory and extract wizard node definitions."""
+    if not ANSIBLE_INVENTORY.exists():
+        print(f"ERROR: Inventory not found at {ANSIBLE_INVENTORY}", file=sys.stderr)
+        sys.exit(2)
+
+    with open(ANSIBLE_INVENTORY) as f:
+        inventory = yaml.safe_load(f)
+
+    wizards = inventory.get("all", {}).get("children", {}).get("wizards", {}).get("hosts", {})
+    global_vars = inventory.get("all", {}).get("vars", {})
+
+    nodes = {}
+    for name, config in wizards.items():
+        nodes[name] = {
+            "host": config.get("ansible_host", "localhost"),
+            "user": config.get("ansible_user", ""),
+            "wizard_name": config.get("wizard_name", name),
+            "hermes_home": config.get("hermes_home", "~/.hermes"),
+            "wizard_home": config.get("wizard_home", f"~/wizards/{name}"),
+            "machine_type": config.get("machine_type", "unknown"),
+        }
+
+    return nodes, global_vars
+
+
+def load_golden_state(inventory_vars: dict) -> dict:
+    """Extract golden state from inventory vars."""
+    golden = {
+        "providers": inventory_vars.get("golden_state_providers", []),
+        "banned_providers": inventory_vars.get("banned_providers", []),
+        "banned_models_patterns": inventory_vars.get("banned_models_patterns", []),
+    }
+    return golden
+
+
+# ── SSH Collection ────────────────────────────────────────────────────────────
+
+def ssh_collect(node_name: str, node_info: dict, timeout: int = 15) -> NodeConfig:
+    """SSH into a node and collect config files."""
+    host = node_info["host"]
+    user = node_info.get("user", "")
+    hermes_home = node_info.get("hermes_home", "~/.hermes")
+    wizard_home = node_info.get("wizard_home", f"~/wizards/{node_name}")
+
+    result = NodeConfig(name=node_name, host=host)
+
+    # Build SSH target
+    if host in ("localhost", "127.0.0.1"):
+        ssh_target = None  # local
+    else:
+        ssh_target = f"{user}@{host}" if user else host
+
+    # Collect each config path
+    for path_template in CONFIG_PATHS:
+        # Resolve path template
+        remote_path = path_template.replace("{name}", node_name)
+        if not remote_path.startswith("/"):
+            # Resolve relative to home
+            if "wizards/" in remote_path:
+                full_path = f"{wizard_home}/config.yaml"
+            else:
+                full_path = f"{hermes_home}/config.yaml" if ".hermes" in remote_path else f"~/{remote_path}"
+        else:
+            full_path = remote_path
+
+        config_content = _remote_cat(ssh_target, full_path, timeout)
+        if config_content is not None:
+            try:
+                parsed = yaml.safe_load(config_content)
+                if parsed:
+                    result.configs[full_path] = parsed
+            except yaml.YAMLError as e:
+                result.errors.append(f"YAML parse error in {full_path}: {e}")
+        # Don't flag missing files as errors — some paths may not exist on all nodes
+
+    # Also collect banned provider scan
+    banned_check = _remote_grep(
+        ssh_target,
+        hermes_home,
+        r"anthropic|claude-sonnet|claude-opus|claude-haiku",
+        timeout
+    )
+    if banned_check:
+        result.configs["__banned_scan__"] = banned_check
+
+    return result
+
+
+def _remote_cat(ssh_target: str | None, path: str, timeout: int) -> str | None:
+    """Cat a file remotely (or locally)."""
+    if ssh_target is None:
+        cmd = ["cat", path]
+    else:
+        cmd = ["ssh", "-o", "ConnectTimeout=5", "-o", "StrictHostKeyChecking=no",
+               ssh_target, f"cat {path}"]
+
+    try:
+        proc = subprocess.run(cmd, capture_output=True, text=True, timeout=timeout)
+        if proc.returncode == 0:
+            return proc.stdout
+    except subprocess.TimeoutExpired:
+        pass
+    except FileNotFoundError:
+        pass
+    return None
+
+
+def _remote_grep(ssh_target: str | None, base_path: str, pattern: str, timeout: int) -> dict:
+    """Grep for banned patterns in config files."""
+    if ssh_target is None:
+        cmd = ["grep", "-rn", "-i", pattern, base_path, "--include=*.yaml", "--include=*.yml"]
+    else:
+        cmd = ["ssh", "-o", "ConnectTimeout=5", "-o", "StrictHostKeyChecking=no",
+               ssh_target, f"grep -rn -i '{pattern}' {base_path} --include='*.yaml' --include='*.yml' 2>/dev/null || true"]
+
+    try:
+        proc = subprocess.run(cmd, capture_output=True, text=True, timeout=timeout)
+        if proc.stdout.strip():
+            lines = proc.stdout.strip().split("\n")
+            return {"matches": lines, "count": len(lines)}
+    except subprocess.TimeoutExpired:
+        pass
+    return {}
+
+
+# ── Drift Detection ───────────────────────────────────────────────────────────
+
+def detect_drift(nodes: list[NodeConfig], golden: dict) -> list[DriftResult]:
+    """Compare each node's config against golden state."""
+    results = []
+
+    for node in nodes:
+        if not node.reachable:
+            continue
+
+        # Check for banned providers
+        banned_scan = node.configs.get("__banned_scan__", {})
+        if banned_scan.get("count", 0) > 0:
+            for match in banned_scan.get("matches", []):
+                results.append(DriftResult(
+                    node=node.name,
+                    file_path="(config files)",
+                    diff_type="banned_provider_found",
+                    key="banned_provider_reference",
+                    node_value=match,
+                    severity="critical"
+                ))
+
+        # Check each config file
+        for path, config in node.configs.items():
+            if path == "__banned_scan__":
+                continue
+
+            # Check provider chain
+            if isinstance(config, dict):
+                node_providers = _extract_provider_chain(config)
+                golden_providers = golden.get("providers", [])
+
+                if node_providers and golden_providers:
+                    # Compare provider names in order
+                    node_names = [p.get("name", "") for p in node_providers]
+                    golden_names = [p.get("name", "") for p in golden_providers]
+
+                    if node_names != golden_names:
+                        results.append(DriftResult(
+                            node=node.name,
+                            file_path=path,
+                            diff_type="value_mismatch",
+                            key="provider_chain",
+                            canonical_value=golden_names,
+                            node_value=node_names,
+                            severity="critical"
+                        ))
+
+                    # Check for banned providers in node config
+                    for banned in golden.get("banned_providers", []):
+                        for provider in node_providers:
+                            prov_name = provider.get("name", "").lower()
+                            prov_model = provider.get("model", "").lower()
+                            if banned in prov_name or banned in prov_model:
+                                results.append(DriftResult(
+                                    node=node.name,
+                                    file_path=path,
+                                    diff_type="banned_provider_found",
+                                    key=f"provider.{provider.get('name', 'unknown')}",
+                                    node_value=provider,
+                                    severity="critical"
+                                ))
+
+                # Check for missing critical keys
+                critical_keys = ["display", "providers", "tools", "delegation"]
+                for key in critical_keys:
+                    if key not in config and key in str(config):
+                        results.append(DriftResult(
+                            node=node.name,
+                            file_path=path,
+                            diff_type="key_missing",
+                            key=key,
+                            canonical_value="(present in golden state)",
+                            severity="warning"
+                        ))
+
+    return results
+
+
+def _extract_provider_chain(config: dict) -> list[dict]:
+    """Extract provider list from a config dict (handles multiple formats)."""
+    # Direct providers key
+    if "providers" in config:
+        providers = config["providers"]
+        if isinstance(providers, list):
+            return providers
+
+    # Nested in display or model config
+    for key in ["model", "inference", "llm"]:
+        if key in config and isinstance(config[key], dict):
+            if "providers" in config[key]:
+                return config[key]["providers"]
+
+    # Single provider format
+    if "provider" in config and "model" in config:
+        return [{"name": config["provider"], "model": config["model"]}]
+
+    return []
+
+
+# ── Auto-Sync ─────────────────────────────────────────────────────────────────
+
+def auto_sync(drifts: list[DriftResult], nodes: list[NodeConfig]) -> list[str]:
+    """Auto-sync drifted nodes using golden state playbook."""
+    actions = []
+
+    drifted_nodes = set(d.node for d in drifts if d.severity == "critical")
+    if not drifted_nodes:
+        actions.append("No critical drift to sync.")
+        return actions
+
+    for node_name in drifted_nodes:
+        node_info = next((n for n in nodes if n.name == node_name), None)
+        if not node_info:
+            continue
+
+        actions.append(f"[{node_name}] Running golden state sync...")
+
+        # Run ansible-playbook for this node
+        cmd = [
+            "ansible-playbook",
+            str(GOLDEN_STATE_PLAYBOOK),
+            "-i", str(ANSIBLE_INVENTORY),
+            "-l", node_name,
+            "--tags", "golden",
+        ]
+
+        try:
+            proc = subprocess.run(
+                cmd, capture_output=True, text=True, timeout=120,
+                cwd=str(REPO_ROOT)
+            )
+            if proc.returncode == 0:
+                actions.append(f"[{node_name}] Sync completed successfully.")
+            else:
+                actions.append(f"[{node_name}] Sync FAILED: {proc.stderr[:200]}")
+        except subprocess.TimeoutExpired:
+            actions.append(f"[{node_name}] Sync timed out after 120s.")
+        except FileNotFoundError:
+            actions.append(f"[{node_name}] ansible-playbook not found. Install Ansible or run manually.")
+
+    return actions
+
+
+# ── Reporting ─────────────────────────────────────────────────────────────────
+
+def print_report(drifts: list[DriftResult], nodes: list[NodeConfig], golden: dict):
+    """Print human-readable drift report."""
+    print("=" * 70)
+    print("CONFIG DRIFT DETECTION REPORT")
+    print("=" * 70)
+    print()
+
+    # Summary
+    reachable = sum(1 for n in nodes if n.reachable)
+    print(f"Nodes checked: {len(nodes)} (reachable: {reachable})")
+    print(f"Golden state providers: {' → '.join(p['name'] for p in golden.get('providers', []))}")
+    print(f"Banned providers: {', '.join(golden.get('banned_providers', []))}")
+    print()
+
+    if not drifts:
+        print("[OK] No config drift detected. All nodes match golden state.")
+        return
+
+    # Group by node
+    by_node: dict[str, list[DriftResult]] = {}
+    for d in drifts:
+        by_node.setdefault(d.node, []).append(d)
+
+    for node_name, node_drifts in sorted(by_node.items()):
+        print(f"--- {node_name} ---")
+        for d in node_drifts:
+            severity_icon = {"critical": "[!!]", "warning": "[!]", "info": "[i]"}.get(d.severity, "[?]")
+            print(f"  {severity_icon} {d.diff_type}: {d.key}")
+            if d.canonical_value is not None:
+                print(f"       canonical: {d.canonical_value}")
+            if d.node_value is not None:
+                print(f"       actual:    {d.node_value}")
+        print()
+
+    # Severity summary
+    critical = sum(1 for d in drifts if d.severity == "critical")
+    warning = sum(1 for d in drifts if d.severity == "warning")
+    print(f"Total: {len(drifts)} drift(s) — {critical} critical, {warning} warning")
+
+
+def print_json_report(drifts: list[DriftResult], nodes: list[NodeConfig], golden: dict):
+    """Print JSON report for automation."""
+    report = {
+        "nodes_checked": len(nodes),
+        "reachable": sum(1 for n in nodes if n.reachable),
+        "golden_providers": [p["name"] for p in golden.get("providers", [])],
+        "drift_count": len(drifts),
+        "critical_count": sum(1 for d in drifts if d.severity == "critical"),
+        "drifts": [
+            {
+                "node": d.node,
+                "file": d.file_path,
+                "type": d.diff_type,
+                "key": d.key,
+                "canonical": d.canonical_value,
+                "actual": d.node_value,
+                "severity": d.severity,
+            }
+            for d in drifts
+        ],
+    }
+    print(json.dumps(report, indent=2, default=str))
+
+
+# ── CLI ───────────────────────────────────────────────────────────────────────
+
+def main():
+    parser = argparse.ArgumentParser(description="Detect config drift across fleet nodes")
+    parser.add_argument("--node", help="Check only this node")
+    parser.add_argument("--auto-sync", action="store_true", help="Auto-fix critical drift with golden state")
+    parser.add_argument("--json", action="store_true", help="JSON output")
+    parser.add_argument("--timeout", type=int, default=15, help="SSH timeout per node (seconds)")
+    args = parser.parse_args()
+
+    # Load inventory
+    print("Loading inventory...", file=sys.stderr)
+    node_defs, global_vars = load_inventory()
+    golden = load_golden_state(global_vars)
+
+    # Filter to single node if requested
+    if args.node:
+        if args.node not in node_defs:
+            print(f"ERROR: Node '{args.node}' not in inventory. Available: {', '.join(node_defs.keys())}")
+            sys.exit(2)
+        node_defs = {args.node: node_defs[args.node]}
+
+    # Collect configs from each node
+    print(f"Collecting configs from {len(node_defs)} node(s)...", file=sys.stderr)
+    nodes = []
+    for name, info in node_defs.items():
+        print(f"  {name} ({info['host']})...", file=sys.stderr, end=" ", flush=True)
+        node_config = ssh_collect(name, info, timeout=args.timeout)
+        if node_config.reachable:
+            print(f"OK ({len(node_config.configs)} files)", file=sys.stderr)
+        else:
+            print("UNREACHABLE", file=sys.stderr)
+        nodes.append(node_config)
+
+    # Detect drift
+    print("\nAnalyzing drift...", file=sys.stderr)
+    drifts = detect_drift(nodes, golden)
+
+    # Output
+    if args.json:
+        print_json_report(drifts, nodes, golden)
+    else:
+        print()
+        print_report(drifts, nodes, golden)
+
+    # Auto-sync if requested
+    if args.auto_sync and drifts:
+        print("\n--- AUTO-SYNC ---")
+        actions = auto_sync(drifts, nodes)
+        for a in actions:
+            print(a)
+
+    # Exit code
+    if any(d.severity == "critical" for d in drifts):
+        sys.exit(1)
+    elif drifts:
+        sys.exit(1)
+    else:
+        sys.exit(0)
+
+
+if __name__ == "__main__":
+    main()

scripts/generate_scene_descriptions.py

@@ -1,191 +0,0 @@
-#!/usr/bin/env python3
-"""
-Auto-generate scene descriptions from image/video assets.
-
-Scans a directory for media files, generates scene descriptions using
-a local vision model (Ollama), and outputs training pairs in JSONL format.
-
-Usage:
-    python scripts/generate_scene_descriptions.py --input ./assets --output training-data/scene-descriptions-auto.jsonl
-    python scripts/generate_scene_descriptions.py --input ./assets --model gemma3:12b --limit 50
-    python scripts/generate_scene_descriptions.py --dry-run  # List files without generating
-"""
-
-import argparse
-import json
-import os
-import subprocess
-import sys
-import time
-from datetime import datetime, timezone
-from pathlib import Path
-from typing import Optional
-
-# Supported media extensions
-IMAGE_EXTS = {".jpg", ".jpeg", ".png", ".gif", ".webp", ".bmp"}
-VIDEO_EXTS = {".mp4", ".webm", ".mov", ".avi", ".mkv"}
-ALL_EXTS = IMAGE_EXTS | VIDEO_EXTS
-
-# Vision model prompt template
-SCENE_PROMPT = """Describe this image for a visual scene database. Output JSON:
-{
-  "mood": "one of: calm, energetic, dark, warm, cool, chaotic, serene, tense, joyful, melancholic",
-  "colors": ["dominant color 1", "dominant color 2", "dominant color 3"],
-  "composition": "one of: close-up, wide-shot, medium-shot, low-angle, high-angle, bird-eye, profile, over-shoulder",
-  "camera": "one of: static, slow-pan, tracking, handheld, crane, dolly, steady, locked-off",
-  "lighting": "one of: natural, artificial, mixed, dramatic, soft, harsh, backlit",
-  "description": "2-3 sentence visual description of the scene"
-}
-
-Be specific. Describe what you see, not what you imagine."""
-
-
-def scan_media(input_dir: str) -> list[Path]:
-    """Scan directory for media files."""
-    media_files = []
-    input_path = Path(input_dir)
-    if not input_path.exists():
-        print(f"Error: {input_dir} does not exist", file=sys.stderr)
-        return media_files
-
-    for ext in sorted(ALL_EXTS):
-        media_files.extend(input_path.rglob(f"*{ext}"))
-        media_files.extend(input_path.rglob(f"*{ext.upper()}"))
-
-    return sorted(set(media_files))
-
-
-def extract_video_frame(video_path: Path, output_path: Path) -> bool:
-    """Extract a representative frame from a video."""
-    try:
-        subprocess.run(
-            ["ffmpeg", "-i", str(video_path), "-vframes", "1",
-             "-q:v", "2", str(output_path), "-y"],
-            capture_output=True, timeout=30,
-        )
-        return output_path.exists()
-    except Exception:
-        return False
-
-
-def describe_image(image_path: Path, model: str = "gemma3:12b",
-                   ollama_url: str = "http://localhost:11434") -> Optional[dict]:
-    """Generate scene description using Ollama vision model."""
-    try:
-        import base64 as b64
-        with open(image_path, "rb") as f:
-            image_b64 = b64.b64encode(f.read()).decode()
-
-        import urllib.request
-        req = urllib.request.Request(
-            f"{ollama_url}/api/generate",
-            data=json.dumps({
-                "model": model,
-                "prompt": SCENE_PROMPT,
-                "images": [image_b64],
-                "stream": False,
-                "options": {"temperature": 0.3, "num_predict": 512}
-            }).encode(),
-            headers={"Content-Type": "application/json"},
-        )
-        resp = urllib.request.urlopen(req, timeout=120)
-        data = json.loads(resp.read())
-        response_text = data.get("response", "")
-
-        # Parse JSON from response
-        import re
-        json_match = re.search(r"\{[\s\S]*\}", response_text)
-        if json_match:
-            return json.loads(json_match.group())
-
-        return {"description": response_text[:500], "mood": "unknown",
-                "colors": [], "composition": "unknown", "camera": "unknown", "lighting": "unknown"}
-    except Exception as e:
-        print(f"  Error describing {image_path.name}: {e}", file=sys.stderr)
-        return None
-
-
-def generate_training_pairs(media_files: list[Path], model: str, ollama_url: str,
-                           limit: int = 0, dry_run: bool = False) -> list[dict]:
-    """Generate training pairs from media files."""
-    pairs = []
-    files = media_files[:limit] if limit > 0 else media_files
-
-    print(f"Processing {len(files)} files...", file=sys.stderr)
-
-    for i, media_path in enumerate(files):
-        print(f"  [{i+1}/{len(files)}] {media_path.name}...", file=sys.stderr, end=" ")
-
-        if dry_run:
-            print("(dry run)", file=sys.stderr)
-            pairs.append({"source": str(media_path), "status": "dry-run"})
-            continue
-
-        is_video = media_path.suffix.lower() in VIDEO_EXTS
-        work_path = media_path
-
-        if is_video:
-            # Extract frame for video
-            frame_path = media_path.with_suffix(".frame.jpg")
-            if extract_video_frame(media_path, frame_path):
-                work_path = frame_path
-            else:
-                print("SKIP (frame extraction failed)", file=sys.stderr)
-                continue
-
-        description = describe_image(work_path, model, ollama_url)
-        if description:
-            pair = {
-                "source": str(media_path),
-                "media_type": "video" if is_video else "image",
-                "description": description,
-                "model": model,
-                "generated_at": datetime.now(timezone.utc).isoformat(),
-            }
-            pairs.append(pair)
-            print("OK", file=sys.stderr)
-        else:
-            print("FAIL", file=sys.stderr)
-
-        # Cleanup temp frame
-        if is_video and work_path != media_path:
-            try:
-                work_path.unlink()
-            except Exception:
-                pass
-
-    return pairs
-
-
-def main():
-    parser = argparse.ArgumentParser(description="Auto-generate scene descriptions from media")
-    parser.add_argument("--input", "-i", required=True, help="Input directory with media files")
-    parser.add_argument("--output", "-o", default="training-data/scene-descriptions-auto.jsonl")
-    parser.add_argument("--model", "-m", default="gemma3:12b", help="Ollama model name")
-    parser.add_argument("--ollama-url", default="http://localhost:11434")
-    parser.add_argument("--limit", "-l", type=int, default=0, help="Max files to process (0=all)")
-    parser.add_argument("--dry-run", action="store_true", help="List files without generating")
-    args = parser.parse_args()
-
-    media_files = scan_media(args.input)
-    print(f"Found {len(media_files)} media files", file=sys.stderr)
-
-    if not media_files:
-        print("No media files found.", file=sys.stderr)
-        sys.exit(1)
-
-    pairs = generate_training_pairs(media_files, args.model, args.ollama_url,
-                                    args.limit, args.dry_run)
-
-    # Write output
-    output_path = Path(args.output)
-    output_path.parent.mkdir(parents=True, exist_ok=True)
-    with open(output_path, "w") as f:
-        for pair in pairs:
-            f.write(json.dumps(pair, ensure_ascii=False) + "\n")
-
-    print(f"\nWrote {len(pairs)} pairs to {output_path}", file=sys.stderr)
-
-
-if __name__ == "__main__":
-    main()