Research Assistant

Batch-process Bear notes tagged 「待整理」: find a relevant GIF for each note, append it, then remove the tag.

Prerequisites

• Bear running + grizzly installed + token at ~/.config/grizzly/token
• gifgrep skill available

Workflow

1. Fetch tagged notes

   grizzly open-tag --name "待整理" --enable-callback --json --token-file ~/.config/grizzly/token
   

   Parse the JSON to get note IDs and titles/summaries.

2. For each note: a. Read full content

   grizzly open-note --id "$NOTE_ID" --enable-callback --json
   

   b. Derive a GIF search query — extract 2–3 keywords from the note title or key findings. Prefer concrete nouns and verbs over abstract terms. c. Search for a GIF using the gifgrep skill (or web_search + web_fetch as fallback) with the derived query. d. Append the GIF to the note under a ## Supporting Media heading:

   printf '\n## Supporting Media\n\n![%s](%s)\n' "$ALT_TEXT" "$GIF_URL" \
     | grizzly add-text --id "$NOTE_ID" --mode append --token-file ~/.config/grizzly/token
   

   e. Remove the 「待整理」 tag — replace note content with the tag removed, or use Bear's tag API:

   grizzly create --title "$TITLE" --tag "$OTHER_TAGS" < /dev/null
   # Then delete the old note if needed, or strip the tag from content
   

   Simplest approach: append a line that re-tags the note (Bear removes a tag when it's deleted from the note body). Use sed on the note content to strip #待整理 and re-write via grizzly add-text --mode replace.

3. Report — list processed notes and any failures.

Script

For batch runs, use scripts/process_tagged.sh:

bash scripts/process_tagged.sh

The script reads all notes tagged 「待整理」, extracts keywords, and calls the gifgrep search endpoint. It appends the top GIF result and strips the tag.

Notes

• If no GIF matches, skip the note and log it — don't force irrelevant results.
• Rate-limit Bear API calls (1–2 req/s) to avoid callback timeouts.
• The script is a convenience wrapper; the agent can also drive the workflow step-by-step for more control.