extract-artifacts

name	extract-artifacts
description	Pull generated documents, canvas docs, code blocks, images and attachments out of an imported ChatGPT conversation and save them as real files. Use when a conversation produced a document, code, or images worth keeping separately.

Extract artifacts

A canvas document quoted inside a transcript is much harder to use than one sitting on disk as a file. This saves them out.

What the normalizer detects

Kind	Where it comes from	Recoverable content
canvas_document	Messages addressed to canmore.create_textdoc / update_textdoc	Full text, name, doc type
code_block	Fenced blocks of 5+ lines in assistant answers	Language and line count; body is in the turn text
generated_image	image_asset_pointer from dalle/image tools	Pointer only — see below
image	Images the user uploaded	Pointer, dimensions
generated_file	Code interpreter output files	Name and file id
attachment	Files the user uploaded	Name, MIME type, size

Listing them

python3 -c "
import json,sys
d=json.load(open(sys.argv[1]))
for a in d['artifacts']:
    print(f\"turn {a['turn_index']:>3}  {a['kind']:<18} {a.get('name') or a.get('pointer') or ''}\")
" <normalized.json>

Saving canvas documents

These carry their full text and are the ones actually worth extracting:

python3 -c "
import json,sys,re,pathlib
d=json.load(open(sys.argv[1])); out=pathlib.Path(sys.argv[2]); out.mkdir(parents=True,exist_ok=True)
n=0
for a in d['artifacts']:
    if a['kind']!='canvas_document' or not a.get('content'): continue
    slug=re.sub(r'[^a-z0-9]+','-',(a.get('name') or 'document').lower()).strip('-')
    ext='.md' if (a.get('doc_type') or '').startswith('document') else '.txt'
    p=out/f\"{a['turn_index']:03d}-{slug}{ext}\"; p.write_text(a['content']); n+=1
    print(p)
print(f'{n} document(s)')
" <normalized.json> <out-dir>

Where a document was created then updated, each operation is a separate artifact and later ones may hold only a diff (updates) rather than the whole text. Prefer the last artifact with non-empty content, and say so if all you have is a diff.

Images and files

file-service://file-... is an internal reference, not a URL — it cannot be fetched with curl and it is not a path. Downloading it needs an authenticated call to the file-download endpoint from inside the browser session, which this plugin does not currently implement. Report images as present-but-not-downloaded rather than implying the pointer is a usable link.

If the user needs the images, the reliable route today is the official account export, which bundles them as real files alongside conversations.json — see import-export-archive.