k0880k commited on
Commit
fd79ed3
·
1 Parent(s): be4a882

fix(postmeet): retain YouTube source fields for template hydration

Browse files
Files changed (1) hide show
  1. tools/meeting_source_fields.py +167 -0
tools/meeting_source_fields.py ADDED
@@ -0,0 +1,167 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python3
2
+ """Canonical source-field projection for PostMeet.
3
+
4
+ PostMeet resolves media before it understands a meeting. The media resolver is
5
+ intentionally small, but a publish target such as a Vlog/YOOtheme template may
6
+ already expose richer source fields (channel, thumbnail, description, publish
7
+ date, etc.). This adapter preserves those facts as data so downstream template
8
+ hydration does not have to scrape the source again or invent values.
9
+
10
+ For YouTube-like sources we ask yt-dlp for metadata only (never the media bytes).
11
+ For every other provider the semantic ``source_media`` descriptor remains the
12
+ source of truth.
13
+ """
14
+ from __future__ import annotations
15
+
16
+ import datetime as _dt
17
+ import json
18
+ import shutil
19
+ import subprocess
20
+ from typing import Any, Dict, Mapping, Optional
21
+
22
+
23
+ YTDLP_FIELDS = (
24
+ "id", "title", "description", "thumbnail", "duration", "timestamp",
25
+ "release_timestamp", "upload_date", "live_status", "channel", "channel_id",
26
+ "channel_url", "uploader", "uploader_id", "uploader_url", "view_count",
27
+ "like_count", "comment_count", "tags", "categories", "license",
28
+ "webpage_url", "original_url",
29
+ )
30
+
31
+
32
+ def _ytdlp(url: str) -> Dict[str, Any]:
33
+ exe = shutil.which("yt-dlp")
34
+ if not exe or not url:
35
+ return {}
36
+ try:
37
+ proc = subprocess.run(
38
+ [exe, "--dump-single-json", "--skip-download", "--no-warnings",
39
+ "--no-playlist", url],
40
+ capture_output=True, text=True, timeout=90,
41
+ )
42
+ if proc.returncode != 0 or not proc.stdout.strip():
43
+ return {}
44
+ data = json.loads(proc.stdout)
45
+ return data if isinstance(data, dict) else {}
46
+ except Exception:
47
+ return {}
48
+
49
+
50
+ def _iso_timestamp(value: Any) -> str:
51
+ try:
52
+ if value in (None, ""):
53
+ return ""
54
+ return _dt.datetime.fromtimestamp(float(value), tz=_dt.timezone.utc).isoformat().replace("+00:00", "Z")
55
+ except (TypeError, ValueError, OSError, OverflowError):
56
+ return ""
57
+
58
+
59
+ def _upload_date(value: Any) -> str:
60
+ raw = str(value or "").strip()
61
+ if len(raw) == 8 and raw.isdigit():
62
+ return f"{raw[:4]}-{raw[4:6]}-{raw[6:8]}"
63
+ return raw
64
+
65
+
66
+ def duration_label(seconds: Any) -> str:
67
+ try:
68
+ total = max(0, int(round(float(seconds))))
69
+ except (TypeError, ValueError):
70
+ return ""
71
+ h, rem = divmod(total, 3600)
72
+ m, s = divmod(rem, 60)
73
+ return f"{h:d}:{m:02d}:{s:02d}" if h else f"{m:d}:{s:02d}"
74
+
75
+
76
+ def from_metadata(source_media: Mapping[str, Any], metadata: Optional[Mapping[str, Any]] = None) -> Dict[str, Any]:
77
+ """Build the bounded, template-facing source field set.
78
+
79
+ ``metadata`` is injectable so tests and non-YouTube adapters can use the
80
+ exact same projection without shelling out to yt-dlp.
81
+ """
82
+ media = dict(source_media or {})
83
+ info = dict(metadata or {})
84
+
85
+ channel = str(info.get("channel") or info.get("uploader") or media.get("creator") or "")
86
+ canonical = str(info.get("webpage_url") or info.get("original_url") or media.get("canonical_ref") or media.get("input_ref") or "")
87
+ seconds = info.get("duration")
88
+ if seconds in (None, ""):
89
+ seconds = media.get("duration_seconds")
90
+
91
+ published_at = _iso_timestamp(info.get("release_timestamp") or info.get("timestamp"))
92
+ upload_date = _upload_date(info.get("upload_date"))
93
+ if not published_at:
94
+ published_at = upload_date
95
+
96
+ fields: Dict[str, Any] = {
97
+ "provider": str(media.get("provider") or ""),
98
+ "source_url": canonical,
99
+ "video_url": canonical,
100
+ "video_id": str(info.get("id") or ""),
101
+ "video_title": str(info.get("title") or media.get("title") or ""),
102
+ "description": str(info.get("description") or ""),
103
+ "thumbnail": str(info.get("thumbnail") or ""),
104
+ "duration_seconds": seconds,
105
+ "duration": duration_label(seconds),
106
+ "published_at": published_at,
107
+ "upload_date": upload_date,
108
+ "live_status": str(info.get("live_status") or ""),
109
+ "channel": channel,
110
+ "channel_name": channel,
111
+ "channel_id": str(info.get("channel_id") or info.get("uploader_id") or ""),
112
+ "channel_url": str(info.get("channel_url") or info.get("uploader_url") or ""),
113
+ "uploader": str(info.get("uploader") or channel),
114
+ "uploader_id": str(info.get("uploader_id") or ""),
115
+ "uploader_url": str(info.get("uploader_url") or ""),
116
+ "view_count": info.get("view_count"),
117
+ "like_count": info.get("like_count"),
118
+ "comment_count": info.get("comment_count"),
119
+ "tags": list(info.get("tags") or []),
120
+ "categories": list(info.get("categories") or []),
121
+ "license": str(info.get("license") or ""),
122
+ }
123
+ # Keep the object deterministic and JSON-safe while retaining explicit zero
124
+ # counts. Empty strings/lists are useful because the template can declare a
125
+ # field without us pretending a value exists.
126
+ return fields
127
+
128
+
129
+ def extract(source_media: Mapping[str, Any], source_ref: str = "") -> Dict[str, Any]:
130
+ media = dict(source_media or {})
131
+ provider = str(media.get("provider") or "").lower()
132
+ metadata: Dict[str, Any] = {}
133
+ if provider == "youtube":
134
+ raw = _ytdlp(str(source_ref or media.get("canonical_ref") or media.get("input_ref") or ""))
135
+ metadata = {k: raw.get(k) for k in YTDLP_FIELDS if k in raw}
136
+ return from_metadata(media, metadata)
137
+
138
+
139
+ def apply(payload: Dict[str, Any], source_media: Mapping[str, Any], source_ref: str = "") -> Dict[str, Any]:
140
+ """Attach source fields and convenient canonical aliases to ``payload``.
141
+
142
+ Existing meeting facts win. Source metadata fills blanks; it never replaces
143
+ a title/summary/poster that a later meeting specialist has established.
144
+ """
145
+ fields = extract(source_media, source_ref)
146
+ payload["source_fields"] = fields
147
+
148
+ aliases = {
149
+ "poster": fields.get("thumbnail"),
150
+ "duration": fields.get("duration"),
151
+ "duration_seconds": fields.get("duration_seconds"),
152
+ "channel": fields.get("channel"),
153
+ "channel_name": fields.get("channel_name"),
154
+ "channel_id": fields.get("channel_id"),
155
+ "channel_url": fields.get("channel_url"),
156
+ "video_id": fields.get("video_id"),
157
+ "source_url": fields.get("source_url"),
158
+ "source_title": fields.get("video_title"),
159
+ "source_description": fields.get("description"),
160
+ "source_published_at": fields.get("published_at"),
161
+ "source_provider": fields.get("provider"),
162
+ }
163
+ for key, value in aliases.items():
164
+ if payload.get(key) in (None, "", []):
165
+ if value not in (None, "", []):
166
+ payload[key] = value
167
+ return fields