personal memory agent
Merge branch 'hopper-5zduu6vy-media-registry'
+44
-40
+2
-1
apps/import/routes.py
+2
-1
apps/import/routes.py
···
13
13
from werkzeug.utils import secure_filename
14
14
15
15
from convey import emit, state
16
+
from media import MEDIA_EXTENSIONS
16
17
from think.detect_created import detect_created
17
18
from think.importers.utils import (
18
19
build_import_info,
···
120
121
"input_type": "file",
121
122
"upload_prompt": "Upload an audio file (.m4a, .mp3, .wav)",
122
123
"has_guide": False,
123
-
"accept": ".m4a,.mp3,.wav,.ogg,.webm",
124
+
"accept": ",".join(sorted(MEDIA_EXTENSIONS)),
124
125
},
125
126
{
126
127
"name": "document",
+25
media.py
+25
media.py
···
1
+
# SPDX-License-Identifier: AGPL-3.0-only
2
+
# Copyright (c) 2026 sol pbc
3
+
4
+
"""Media format registry - single source of truth for extensions, MIME types, and kind."""
5
+
6
+
FORMATS = [
7
+
(".flac", "audio/flac", "audio"),
8
+
(".opus", "audio/opus", "audio"),
9
+
(".ogg", "audio/ogg", "audio"),
10
+
(".m4a", "audio/mp4", "audio"),
11
+
(".mp3", "audio/mpeg", "audio"),
12
+
(".wav", "audio/wav", "audio"),
13
+
(".webm", "video/webm", "video"),
14
+
(".mp4", "video/mp4", "video"),
15
+
(".mov", "video/quicktime", "video"),
16
+
]
17
+
18
+
AUDIO_EXTENSIONS: frozenset[str] = frozenset(
19
+
ext for ext, _, kind in FORMATS if kind == "audio"
20
+
)
21
+
VIDEO_EXTENSIONS: frozenset[str] = frozenset(
22
+
ext for ext, _, kind in FORMATS if kind == "video"
23
+
)
24
+
MEDIA_EXTENSIONS: frozenset[str] = frozenset(ext for ext, _, _ in FORMATS)
25
+
MIME_TYPES: dict[str, str] = {ext: mime for ext, mime, _ in FORMATS}
+1
-3
observe/transcribe/main.py
+1
-3
observe/transcribe/main.py
···
54
54
55
55
import numpy as np
56
56
57
+
from media import AUDIO_EXTENSIONS as SUPPORTED_AUDIO_FORMATS
57
58
from observe.transcribe import (
58
59
BACKEND_REGISTRY,
59
60
get_backend,
···
98
99
99
100
# Number of recent entity names to load for transcription context
100
101
ENTITY_NAMES_LIMIT = 40
101
-
102
-
# Supported audio file formats for transcription
103
-
SUPPORTED_AUDIO_FORMATS = {".flac", ".m4a", ".mp3", ".ogg", ".opus", ".wav"}
104
102
105
103
# Module-level voice encoder cache
106
104
_voice_encoder = None
+4
-2
observe/utils.py
+4
-2
observe/utils.py
···
16
16
import numpy as np
17
17
import soundfile as sf
18
18
19
+
from media import AUDIO_EXTENSIONS as _AUDIO_EXTENSIONS
20
+
from media import VIDEO_EXTENSIONS as _VIDEO_EXTENSIONS
19
21
from think.utils import day_path
20
22
21
23
logger = logging.getLogger(__name__)
···
23
25
# Standard sample rate for audio processing
24
26
SAMPLE_RATE = 16000
25
27
26
-
VIDEO_EXTENSIONS = (".webm", ".mp4", ".mov")
27
-
AUDIO_EXTENSIONS = (".flac", ".ogg", ".m4a", ".opus", ".mp3")
28
+
VIDEO_EXTENSIONS = tuple(_VIDEO_EXTENSIONS)
29
+
AUDIO_EXTENSIONS = tuple(_AUDIO_EXTENSIONS)
28
30
29
31
30
32
def audio_to_flac_bytes(audio: np.ndarray, sample_rate: int) -> bytes:
+1
-1
pyproject.toml
+1
-1
pyproject.toml
+1
-1
tests/test_retention.py
+1
-1
tests/test_retention.py
···
26
26
27
27
class TestIsRawMedia:
28
28
def test_audio_extensions(self, tmp_path):
29
-
for ext in (".flac", ".opus", ".ogg", ".m4a"):
29
+
for ext in (".flac", ".opus", ".ogg", ".m4a", ".mp3", ".wav"):
30
30
p = tmp_path / f"audio{ext}"
31
31
p.touch()
32
32
assert is_raw_media(p), f"{ext} should be raw media"
+4
-5
think/retention.py
+4
-5
think/retention.py
···
23
23
from pathlib import Path
24
24
from typing import Any
25
25
26
+
from media import AUDIO_EXTENSIONS as RAW_AUDIO_EXTENSIONS
27
+
from media import MEDIA_EXTENSIONS as RAW_MEDIA_EXTENSIONS
28
+
from media import VIDEO_EXTENSIONS as RAW_VIDEO_EXTENSIONS
26
29
from think.utils import day_dirs, get_journal, iter_segments
27
30
28
31
logger = logging.getLogger(__name__)
···
31
34
# Raw media file identification
32
35
# ---------------------------------------------------------------------------
33
36
34
-
RAW_AUDIO_EXTENSIONS = frozenset({".flac", ".opus", ".ogg", ".m4a"})
35
-
RAW_VIDEO_EXTENSIONS = frozenset({".webm", ".mov", ".mp4"})
36
-
RAW_MEDIA_EXTENSIONS = RAW_AUDIO_EXTENSIONS | RAW_VIDEO_EXTENSIONS
37
-
38
37
39
38
def is_raw_media(path: Path) -> bool:
40
39
"""Check if a file is raw media (layer 1 capture).
41
40
42
-
Raw media: *.flac, *.opus, *.ogg, *.m4a (audio),
41
+
Raw media: *.flac, *.opus, *.ogg, *.m4a, *.mp3, *.wav (audio),
43
42
*.webm, *.mov, *.mp4 (video), monitor_*_diff.png (screen diffs).
44
43
"""
45
44
if path.suffix.lower() in RAW_MEDIA_EXTENSIONS:
+4
-18
think/utils.py
+4
-18
think/utils.py
···
26
26
from dotenv import load_dotenv
27
27
from timefhuman import timefhuman
28
28
29
+
from media import MIME_TYPES
30
+
29
31
DATE_RE = re.compile(r"\d{8}")
30
32
_journal_path_cache: str | None = None
31
33
···
800
802
if not rel:
801
803
raise ValueError(f"No 'raw' field found in metadata for {name}")
802
804
803
-
# Determine MIME type from raw file extension
804
-
if rel.endswith(".flac"):
805
-
mime = "audio/flac"
806
-
elif rel.endswith(".ogg"):
807
-
mime = "audio/ogg"
808
-
elif rel.endswith(".m4a"):
809
-
mime = "audio/mp4"
810
-
elif rel.endswith(".png"):
811
-
mime = "image/png"
812
-
elif rel.endswith(".webm"):
813
-
mime = "video/webm"
814
-
elif rel.endswith(".mp4"):
815
-
mime = "video/mp4"
816
-
elif rel.endswith(".mov"):
817
-
mime = "video/quicktime"
818
-
else:
819
-
# Default fallback for unknown types
820
-
mime = "application/octet-stream"
805
+
suffix = Path(rel).suffix.lower()
806
+
mime = {**MIME_TYPES, ".png": "image/png"}.get(suffix, "application/octet-stream")
821
807
822
808
return rel, mime, meta
823
809