personal memory agent
Refactor screen describe utilities
Centralize segment/suffix parsing in describe and reuse video extension lists across observe and stats. Update category docs/comments to reflect current paths and legacy formatter support.
+40
-31
+1
observe/categories/README.md
+1
observe/categories/README.md
···
66
66
2. For categories with prompts here, a follow-up request extracts detailed content
67
67
3. Results are stored in JSONL under the category name (e.g., `"meeting": {...}`)
68
68
4. `observe/screen.py` formats JSONL to markdown, using custom formatters when available
69
+
5. The formatter also supports legacy keys (`meeting_analysis`, `extracted_text`) for older data
+24
-16
observe/describe.py
+24
-16
observe/describe.py
···
38
38
CATEGORY = "category" # Category-specific follow-up
39
39
40
40
41
+
def _segment_and_suffix(media_path: Path) -> tuple[str, str]:
42
+
"""Return segment key and descriptive suffix for a media path."""
43
+
from observe.utils import extract_descriptive_suffix
44
+
from think.utils import segment_key
45
+
46
+
segment = segment_key(media_path.stem)
47
+
if segment is None:
48
+
raise ValueError(
49
+
f"Invalid video filename: {media_path.stem} (must be HHMMSS_LEN format)"
50
+
)
51
+
try:
52
+
suffix = extract_descriptive_suffix(media_path.stem)
53
+
except ValueError as exc:
54
+
raise ValueError(
55
+
f"Invalid video filename: {media_path.stem} (must be HHMMSS_LEN format)"
56
+
) from exc
57
+
return segment, suffix
58
+
59
+
41
60
def _discover_category_prompts() -> dict[str, dict]:
42
61
"""
43
62
Discover available category prompts from categories/ directory.
···
264
283
265
284
def _move_to_segment(self, media_path: Path) -> Path:
266
285
"""Move media file to its segment and return new path."""
267
-
from observe.utils import extract_descriptive_suffix
268
-
from think.utils import segment_key
269
-
270
-
segment = segment_key(media_path.stem)
271
-
if segment is None:
272
-
raise ValueError(f"Invalid media filename: {media_path.stem}")
273
-
suffix = extract_descriptive_suffix(media_path.stem)
286
+
segment, suffix = _segment_and_suffix(media_path)
274
287
segment_dir = media_path.parent / segment
275
288
try:
276
289
segment_dir.mkdir(exist_ok=True)
···
696
709
suffix = None
697
710
if not args.frames_only:
698
711
# Extract segment and suffix for output naming
699
-
from observe.utils import extract_descriptive_suffix
700
-
from think.utils import segment_key
701
-
702
-
segment = segment_key(video_path.stem)
703
-
if segment is None:
704
-
parser.error(
705
-
f"Invalid video filename: {video_path.stem} (must be HHMMSS_LEN format)"
706
-
)
707
-
suffix = extract_descriptive_suffix(video_path.stem)
712
+
try:
713
+
segment, suffix = _segment_and_suffix(video_path)
714
+
except ValueError as exc:
715
+
parser.error(str(exc))
708
716
segment_dir = video_path.parent / segment
709
717
segment_dir.mkdir(exist_ok=True)
710
718
# Output JSONL matches input filename pattern (e.g., center_DP-3_screen.jsonl)
+1
-1
observe/screen.py
+1
-1
observe/screen.py
+5
-6
observe/sense.py
+5
-6
observe/sense.py
···
20
20
from watchdog.events import FileSystemEventHandler
21
21
from watchdog.observers import Observer
22
22
23
+
from observe.utils import VIDEO_EXTENSIONS
23
24
from think.callosum import CallosumConnection
24
25
from think.runner import ManagedProcess as RunnerManagedProcess
25
26
from think.utils import day_path, setup_cli
···
580
581
unprocessed = []
581
582
unprocessed.extend(sorted(p.name for p in day_dir.glob("*.flac")))
582
583
unprocessed.extend(sorted(p.name for p in day_dir.glob("*.m4a")))
583
-
unprocessed.extend(sorted(p.name for p in day_dir.glob("*.webm")))
584
-
unprocessed.extend(sorted(p.name for p in day_dir.glob("*.mp4")))
585
-
unprocessed.extend(sorted(p.name for p in day_dir.glob("*.mov")))
584
+
for ext in VIDEO_EXTENSIONS:
585
+
unprocessed.extend(sorted(p.name for p in day_dir.glob(f"*{ext}")))
586
586
587
587
return {"processed": processed, "unprocessed": unprocessed}
588
588
···
616
616
sensor.register("*.m4a", "transcribe", ["observe-transcribe", "{file}"])
617
617
618
618
# Video files: any HHMMSS_*.webm, HHMMSS_*.mp4, HHMMSS_*.mov in day root
619
-
sensor.register("*.webm", "describe", ["observe-describe", "{file}"])
620
-
sensor.register("*.mp4", "describe", ["observe-describe", "{file}"])
621
-
sensor.register("*.mov", "describe", ["observe-describe", "{file}"])
619
+
for ext in VIDEO_EXTENSIONS:
620
+
sensor.register(f"*{ext}", "describe", ["observe-describe", "{file}"])
622
621
623
622
if args.day:
624
623
# Batch mode: process specific day
+2
observe/utils.py
+2
observe/utils.py
+7
-8
think/journal_stats.py
+7
-8
think/journal_stats.py
···
7
7
from pathlib import Path
8
8
from typing import Dict
9
9
10
-
from observe.utils import load_analysis_frames
10
+
from observe.utils import VIDEO_EXTENSIONS, load_analysis_frames
11
11
from think.insight import scan_day as insight_scan_day
12
12
from think.utils import day_dirs, setup_cli
13
13
···
43
43
files.extend(day_dir.glob("*/screen.jsonl"))
44
44
files.extend(day_dir.glob("*/*_screen.jsonl")) # Split screen files
45
45
files.extend(day_dir.glob("*/raw.flac"))
46
-
files.extend(day_dir.glob("*/screen.webm"))
46
+
for ext in VIDEO_EXTENSIONS:
47
+
files.extend(day_dir.glob(f"*/screen{ext}"))
47
48
# Check day root for unprocessed files
48
49
files.extend(day_dir.glob("*_raw.flac"))
49
50
files.extend(day_dir.glob("*_raw.m4a"))
50
-
files.extend(day_dir.glob("*_screen.webm"))
51
-
files.extend(day_dir.glob("*_screen.mp4"))
52
-
files.extend(day_dir.glob("*_screen.mov"))
51
+
for ext in VIDEO_EXTENSIONS:
52
+
files.extend(day_dir.glob(f"*_screen{ext}"))
53
53
54
54
insights = day_dir / "insights"
55
55
if insights.is_dir():
···
246
246
# --- Unprocessed files ---
247
247
unprocessed = list(day_dir.glob("*_raw.flac"))
248
248
unprocessed.extend(day_dir.glob("*_raw.m4a"))
249
-
unprocessed.extend(day_dir.glob("*_screen.webm"))
250
-
unprocessed.extend(day_dir.glob("*_screen.mp4"))
251
-
unprocessed.extend(day_dir.glob("*_screen.mov"))
249
+
for ext in VIDEO_EXTENSIONS:
250
+
unprocessed.extend(day_dir.glob(f"*_screen{ext}"))
252
251
stats["unprocessed_files"] = len(unprocessed)
253
252
254
253
# --- Insight summaries ---