|
1
|
"""Skill: Export knowledge graph as structured notes (Obsidian, Notion).""" |
|
2
|
|
|
3
|
import csv |
|
4
|
import io |
|
5
|
import logging |
|
6
|
from datetime import date |
|
7
|
from pathlib import Path |
|
8
|
from typing import Dict, List, Optional |
|
9
|
|
|
10
|
from video_processor.agent.skills.base import ( |
|
11
|
AgentContext, |
|
12
|
Artifact, |
|
13
|
Skill, |
|
14
|
register_skill, |
|
15
|
) |
|
16
|
|
|
17
|
logger = logging.getLogger(__name__) |
|
18
|
|
|
19
|
|
|
20
|
def _sanitize_filename(name: str) -> str: |
|
21
|
"""Convert a name to a filesystem-safe filename.""" |
|
22
|
return ( |
|
23
|
name.replace("/", "-") |
|
24
|
.replace("\\", "-") |
|
25
|
.replace(":", "-") |
|
26
|
.replace('"', "") |
|
27
|
.replace("?", "") |
|
28
|
.replace("*", "") |
|
29
|
.replace("<", "") |
|
30
|
.replace(">", "") |
|
31
|
.replace("|", "") |
|
32
|
) |
|
33
|
|
|
34
|
|
|
35
|
def _build_indexes(kg_data: dict): |
|
36
|
"""Build lookup structures from knowledge graph data. |
|
37
|
|
|
38
|
Returns (nodes, by_type, node_lookup, outgoing, incoming). |
|
39
|
""" |
|
40
|
nodes = kg_data.get("nodes", []) |
|
41
|
relationships = kg_data.get("relationships", []) |
|
42
|
|
|
43
|
by_type: Dict[str, list] = {} |
|
44
|
node_lookup: Dict[str, dict] = {} |
|
45
|
for node in nodes: |
|
46
|
name = node.get("name", node.get("id", "")) |
|
47
|
ntype = node.get("type", "concept") |
|
48
|
by_type.setdefault(ntype, []).append(node) |
|
49
|
node_lookup[name] = node |
|
50
|
|
|
51
|
outgoing: Dict[str, list] = {} |
|
52
|
incoming: Dict[str, list] = {} |
|
53
|
for rel in relationships: |
|
54
|
src = rel.get("source", "") |
|
55
|
tgt = rel.get("target", "") |
|
56
|
rtype = rel.get("type", "related_to") |
|
57
|
outgoing.setdefault(src, []).append((tgt, rtype)) |
|
58
|
incoming.setdefault(tgt, []).append((src, rtype)) |
|
59
|
|
|
60
|
return nodes, by_type, node_lookup, outgoing, incoming |
|
61
|
|
|
62
|
|
|
63
|
# --------------------------------------------------------------------------- |
|
64
|
# Obsidian export |
|
65
|
# --------------------------------------------------------------------------- |
|
66
|
|
|
67
|
|
|
68
|
def export_to_obsidian( |
|
69
|
kg_data: dict, |
|
70
|
output_dir: Path, |
|
71
|
artifacts: Optional[List[Artifact]] = None, |
|
72
|
) -> List[Path]: |
|
73
|
"""Export knowledge graph as an Obsidian vault. |
|
74
|
|
|
75
|
Creates one ``.md`` file per entity with YAML frontmatter and |
|
76
|
``[[wiki-links]]``, an ``_Index.md`` Map of Content, tag pages per |
|
77
|
entity type, and optional artifact notes. |
|
78
|
""" |
|
79
|
output_dir.mkdir(parents=True, exist_ok=True) |
|
80
|
artifacts = artifacts or [] |
|
81
|
created: List[Path] = [] |
|
82
|
today = date.today().isoformat() |
|
83
|
|
|
84
|
nodes, by_type, node_lookup, outgoing, incoming = _build_indexes(kg_data) |
|
85
|
|
|
86
|
# --- Individual entity notes --- |
|
87
|
for node in nodes: |
|
88
|
name = node.get("name", node.get("id", "")) |
|
89
|
if not name: |
|
90
|
continue |
|
91
|
ntype = node.get("type", "concept") |
|
92
|
descs = node.get("descriptions", []) |
|
93
|
aliases = node.get("aliases", []) |
|
94
|
|
|
95
|
# YAML frontmatter |
|
96
|
tags_yaml = f" - {ntype}" |
|
97
|
aliases_yaml = "" |
|
98
|
if aliases: |
|
99
|
alias_lines = "\n".join(f" - {a}" for a in aliases) |
|
100
|
aliases_yaml = f"aliases:\n{alias_lines}\n" |
|
101
|
|
|
102
|
frontmatter = f"---\ntype: {ntype}\ntags:\n{tags_yaml}\n{aliases_yaml}date: {today}\n---\n" |
|
103
|
|
|
104
|
parts = [frontmatter, f"# {name}", ""] |
|
105
|
|
|
106
|
# Descriptions |
|
107
|
if descs: |
|
108
|
for d in descs: |
|
109
|
parts.append(f"{d}") |
|
110
|
parts.append("") |
|
111
|
|
|
112
|
# Outgoing relationships |
|
113
|
outs = outgoing.get(name, []) |
|
114
|
if outs: |
|
115
|
parts.append("## Relationships") |
|
116
|
parts.append("") |
|
117
|
for tgt, rtype in outs: |
|
118
|
parts.append(f"- **{rtype}**: [[{tgt}]]") |
|
119
|
parts.append("") |
|
120
|
|
|
121
|
# Incoming relationships |
|
122
|
ins = incoming.get(name, []) |
|
123
|
if ins: |
|
124
|
parts.append("## Referenced by") |
|
125
|
parts.append("") |
|
126
|
for src, rtype in ins: |
|
127
|
parts.append(f"- **{rtype}** from [[{src}]]") |
|
128
|
parts.append("") |
|
129
|
|
|
130
|
filename = _sanitize_filename(name) + ".md" |
|
131
|
path = output_dir / filename |
|
132
|
path.write_text("\n".join(parts), encoding="utf-8") |
|
133
|
created.append(path) |
|
134
|
|
|
135
|
# --- Index note (Map of Content) --- |
|
136
|
index_parts = [ |
|
137
|
"---", |
|
138
|
"type: index", |
|
139
|
"tags:", |
|
140
|
" - MOC", |
|
141
|
f"date: {today}", |
|
142
|
"---", |
|
143
|
"", |
|
144
|
"# Index", |
|
145
|
"", |
|
146
|
f"**{len(nodes)}** entities | **{len(kg_data.get('relationships', []))}** relationships", |
|
147
|
"", |
|
148
|
] |
|
149
|
|
|
150
|
for etype in sorted(by_type.keys()): |
|
151
|
elist = by_type[etype] |
|
152
|
index_parts.append(f"## {etype.title()}") |
|
153
|
index_parts.append("") |
|
154
|
for node in sorted(elist, key=lambda n: n.get("name", "")): |
|
155
|
name = node.get("name", "") |
|
156
|
index_parts.append(f"- [[{name}]]") |
|
157
|
index_parts.append("") |
|
158
|
|
|
159
|
if artifacts: |
|
160
|
index_parts.append("## Artifacts") |
|
161
|
index_parts.append("") |
|
162
|
for art in artifacts: |
|
163
|
index_parts.append(f"- [[{art.name}]]") |
|
164
|
index_parts.append("") |
|
165
|
|
|
166
|
index_path = output_dir / "_Index.md" |
|
167
|
index_path.write_text("\n".join(index_parts), encoding="utf-8") |
|
168
|
created.append(index_path) |
|
169
|
|
|
170
|
# --- Tag pages (one per entity type) --- |
|
171
|
for etype, elist in sorted(by_type.items()): |
|
172
|
tag_parts = [ |
|
173
|
"---", |
|
174
|
"type: tag", |
|
175
|
"tags:", |
|
176
|
f" - {etype}", |
|
177
|
f"date: {today}", |
|
178
|
"---", |
|
179
|
"", |
|
180
|
f"# {etype.title()}", |
|
181
|
"", |
|
182
|
f"All entities of type **{etype}** ({len(elist)}).", |
|
183
|
"", |
|
184
|
] |
|
185
|
for node in sorted(elist, key=lambda n: n.get("name", "")): |
|
186
|
name = node.get("name", "") |
|
187
|
descs = node.get("descriptions", []) |
|
188
|
summary = descs[0] if descs else "" |
|
189
|
tag_parts.append(f"- [[{name}]]" + (f" - {summary}" if summary else "")) |
|
190
|
tag_parts.append("") |
|
191
|
|
|
192
|
tag_filename = f"Tag - {etype.title()}.md" |
|
193
|
tag_path = output_dir / _sanitize_filename(tag_filename) |
|
194
|
tag_path.write_text("\n".join(tag_parts), encoding="utf-8") |
|
195
|
created.append(tag_path) |
|
196
|
|
|
197
|
# --- Artifact notes --- |
|
198
|
for art in artifacts: |
|
199
|
art_parts = [ |
|
200
|
"---", |
|
201
|
"type: artifact", |
|
202
|
f"artifact_type: {art.artifact_type}", |
|
203
|
"tags:", |
|
204
|
" - artifact", |
|
205
|
f" - {art.artifact_type}", |
|
206
|
f"date: {today}", |
|
207
|
"---", |
|
208
|
"", |
|
209
|
f"# {art.name}", |
|
210
|
"", |
|
211
|
art.content, |
|
212
|
"", |
|
213
|
] |
|
214
|
art_filename = _sanitize_filename(art.name) + ".md" |
|
215
|
art_path = output_dir / art_filename |
|
216
|
art_path.write_text("\n".join(art_parts), encoding="utf-8") |
|
217
|
created.append(art_path) |
|
218
|
|
|
219
|
logger.info("Exported %d Obsidian notes to %s", len(created), output_dir) |
|
220
|
return created |
|
221
|
|
|
222
|
|
|
223
|
# --------------------------------------------------------------------------- |
|
224
|
# Notion-compatible markdown export |
|
225
|
# --------------------------------------------------------------------------- |
|
226
|
|
|
227
|
|
|
228
|
def export_to_notion_md( |
|
229
|
kg_data: dict, |
|
230
|
output_dir: Path, |
|
231
|
artifacts: Optional[List[Artifact]] = None, |
|
232
|
) -> List[Path]: |
|
233
|
"""Export knowledge graph as Notion-compatible markdown. |
|
234
|
|
|
235
|
Creates ``.md`` files with Notion-style callout blocks and a |
|
236
|
database-style CSV for bulk import. |
|
237
|
""" |
|
238
|
output_dir.mkdir(parents=True, exist_ok=True) |
|
239
|
artifacts = artifacts or [] |
|
240
|
created: List[Path] = [] |
|
241
|
|
|
242
|
nodes, by_type, node_lookup, outgoing, incoming = _build_indexes(kg_data) |
|
243
|
|
|
244
|
# --- Database CSV --- |
|
245
|
csv_buffer = io.StringIO() |
|
246
|
writer = csv.writer(csv_buffer) |
|
247
|
writer.writerow(["Name", "Type", "Description", "Related To"]) |
|
248
|
|
|
249
|
for node in nodes: |
|
250
|
name = node.get("name", node.get("id", "")) |
|
251
|
ntype = node.get("type", "concept") |
|
252
|
descs = node.get("descriptions", []) |
|
253
|
desc_text = "; ".join(descs[:2]) if descs else "" |
|
254
|
outs = outgoing.get(name, []) |
|
255
|
related = ", ".join(tgt for tgt, _ in outs) if outs else "" |
|
256
|
writer.writerow([name, ntype, desc_text, related]) |
|
257
|
|
|
258
|
csv_path = output_dir / "entities_database.csv" |
|
259
|
csv_path.write_text(csv_buffer.getvalue(), encoding="utf-8") |
|
260
|
created.append(csv_path) |
|
261
|
|
|
262
|
# --- Individual entity pages --- |
|
263
|
for node in nodes: |
|
264
|
name = node.get("name", node.get("id", "")) |
|
265
|
if not name: |
|
266
|
continue |
|
267
|
ntype = node.get("type", "concept") |
|
268
|
descs = node.get("descriptions", []) |
|
269
|
|
|
270
|
type_emoji = { |
|
271
|
"person": "person", |
|
272
|
"technology": "computer", |
|
273
|
"organization": "building", |
|
274
|
"concept": "bulb", |
|
275
|
"event": "calendar", |
|
276
|
"location": "round_pushpin", |
|
277
|
} |
|
278
|
emoji = type_emoji.get(ntype, "bulb") |
|
279
|
|
|
280
|
parts = [ |
|
281
|
f"# {name}", |
|
282
|
"", |
|
283
|
f"> :{emoji}: **Type:** {ntype}", |
|
284
|
"", |
|
285
|
] |
|
286
|
|
|
287
|
if descs: |
|
288
|
parts.append("## Description") |
|
289
|
parts.append("") |
|
290
|
for d in descs: |
|
291
|
parts.append(f"{d}") |
|
292
|
parts.append("") |
|
293
|
|
|
294
|
# Properties callout |
|
295
|
properties = node.get("properties", {}) |
|
296
|
if properties: |
|
297
|
parts.append("> :memo: **Properties**") |
|
298
|
for k, v in properties.items(): |
|
299
|
parts.append(f"> - **{k}:** {v}") |
|
300
|
parts.append("") |
|
301
|
|
|
302
|
# Outgoing relationships |
|
303
|
outs = outgoing.get(name, []) |
|
304
|
if outs: |
|
305
|
parts.append("## Relationships") |
|
306
|
parts.append("") |
|
307
|
parts.append("| Target | Relationship |") |
|
308
|
parts.append("|--------|-------------|") |
|
309
|
for tgt, rtype in outs: |
|
310
|
parts.append(f"| {tgt} | {rtype} |") |
|
311
|
parts.append("") |
|
312
|
|
|
313
|
# Incoming relationships |
|
314
|
ins = incoming.get(name, []) |
|
315
|
if ins: |
|
316
|
parts.append("## Referenced by") |
|
317
|
parts.append("") |
|
318
|
parts.append("| Source | Relationship |") |
|
319
|
parts.append("|--------|-------------|") |
|
320
|
for src, rtype in ins: |
|
321
|
parts.append(f"| {src} | {rtype} |") |
|
322
|
parts.append("") |
|
323
|
|
|
324
|
filename = _sanitize_filename(name) + ".md" |
|
325
|
path = output_dir / filename |
|
326
|
path.write_text("\n".join(parts), encoding="utf-8") |
|
327
|
created.append(path) |
|
328
|
|
|
329
|
# --- Overview page --- |
|
330
|
overview_parts = [ |
|
331
|
"# Knowledge Graph Overview", |
|
332
|
"", |
|
333
|
f"> :bar_chart: **Stats:** {len(nodes)} entities, " |
|
334
|
f"{len(kg_data.get('relationships', []))} relationships", |
|
335
|
"", |
|
336
|
"## Entity Types", |
|
337
|
"", |
|
338
|
] |
|
339
|
for etype in sorted(by_type.keys()): |
|
340
|
elist = by_type[etype] |
|
341
|
overview_parts.append(f"### {etype.title()} ({len(elist)})") |
|
342
|
overview_parts.append("") |
|
343
|
for node in sorted(elist, key=lambda n: n.get("name", "")): |
|
344
|
name = node.get("name", "") |
|
345
|
overview_parts.append(f"- {name}") |
|
346
|
overview_parts.append("") |
|
347
|
|
|
348
|
if artifacts: |
|
349
|
overview_parts.append("## Artifacts") |
|
350
|
overview_parts.append("") |
|
351
|
for art in artifacts: |
|
352
|
overview_parts.append(f"- **{art.name}** ({art.artifact_type})") |
|
353
|
overview_parts.append("") |
|
354
|
|
|
355
|
overview_path = output_dir / "Overview.md" |
|
356
|
overview_path.write_text("\n".join(overview_parts), encoding="utf-8") |
|
357
|
created.append(overview_path) |
|
358
|
|
|
359
|
# --- Artifact pages --- |
|
360
|
for art in artifacts: |
|
361
|
art_parts = [ |
|
362
|
f"# {art.name}", |
|
363
|
"", |
|
364
|
f"> :page_facing_up: **Type:** {art.artifact_type} | **Format:** {art.format}", |
|
365
|
"", |
|
366
|
art.content, |
|
367
|
"", |
|
368
|
] |
|
369
|
art_filename = _sanitize_filename(art.name) + ".md" |
|
370
|
art_path = output_dir / art_filename |
|
371
|
art_path.write_text("\n".join(art_parts), encoding="utf-8") |
|
372
|
created.append(art_path) |
|
373
|
|
|
374
|
logger.info("Exported %d Notion markdown files to %s", len(created), output_dir) |
|
375
|
return created |
|
376
|
|
|
377
|
|
|
378
|
# --------------------------------------------------------------------------- |
|
379
|
# Skill class |
|
380
|
# --------------------------------------------------------------------------- |
|
381
|
|
|
382
|
|
|
383
|
class NotesExportSkill(Skill): |
|
384
|
"""Export knowledge graph as structured notes (Obsidian, Notion). |
|
385
|
|
|
386
|
For GitHub wiki export, see the ``wiki_generator`` skill. |
|
387
|
""" |
|
388
|
|
|
389
|
name = "notes_export" |
|
390
|
description = "Export knowledge graph as structured notes (Obsidian, Notion)" |
|
391
|
|
|
392
|
def execute(self, context: AgentContext, **kwargs) -> Artifact: |
|
393
|
fmt = kwargs.get("format", "obsidian") |
|
394
|
output_dir = Path(kwargs.get("output_dir", f"notes_export_{fmt}")) |
|
395
|
kg_data = context.knowledge_graph.to_dict() |
|
396
|
artifacts = context.artifacts or [] |
|
397
|
|
|
398
|
if fmt == "notion": |
|
399
|
created = export_to_notion_md(kg_data, output_dir, artifacts=artifacts) |
|
400
|
else: |
|
401
|
created = export_to_obsidian(kg_data, output_dir, artifacts=artifacts) |
|
402
|
|
|
403
|
file_list = "\n".join(f"- {p.name}" for p in created) |
|
404
|
summary = f"Exported {len(created)} {fmt} notes to `{output_dir}`:\n\n{file_list}" |
|
405
|
|
|
406
|
return Artifact( |
|
407
|
name=f"Notes Export ({fmt.title()})", |
|
408
|
content=summary, |
|
409
|
artifact_type="notes_export", |
|
410
|
format="markdown", |
|
411
|
metadata={ |
|
412
|
"output_dir": str(output_dir), |
|
413
|
"format": fmt, |
|
414
|
"file_count": len(created), |
|
415
|
"files": [str(p) for p in created], |
|
416
|
}, |
|
417
|
) |
|
418
|
|
|
419
|
|
|
420
|
register_skill(NotesExportSkill()) |
|
421
|
|