brain-search

Search across personal atoms (~/brain/atoms/) and team vault (~/brain-team/) for relevant knowledge. Results are labeled by layer.

Steps

1. Read config and query

cat ~/.claude/brain/config.json

If no query was provided as argument, ask:

"What are you looking for? Enter keywords or a phrase:"

Call this QUERY.

2. Detect Obsidian CLI

which obsidian 2>/dev/null && echo "cli-available" || echo "cli-not-found"

Store result as HAS_OBSIDIAN_CLI. Use CLI when available; fall back to grep + atoms-index otherwise.

3. Search personal atoms

If HAS_OBSIDIAN_CLI:

VAULT_NAME=$(basename <vault_path>)
obsidian vault="$VAULT_NAME" search query="<QUERY>" limit=20 2>/dev/null

If that returns results, read any matching files for full content. If the CLI errors or returns nothing, fall through to grep.

Grep fallback (always works):

grep -ril "<QUERY>" <vault_path>/atoms/ 2>/dev/null

For each matching file, read it.

Also score via atoms-index (fast keyword match, works regardless of CLI):

python3 -c "
import json, re
from pathlib import Path
index_path = Path.home() / '.claude/brain/atoms-index.json'
if not index_path.exists():
    exit()
index = json.loads(index_path.read_text())
query_words = set(re.findall(r'[a-z][a-z0-9_-]{2,}', '<QUERY>'.lower()))
results = []
for atom in index:
    score = len(query_words & set(atom['keywords']))
    if score >= 1:
        results.append((score, atom['slug'], atom['path'], atom.get('reference_count', 1)))
results.sort(reverse=True)
for score, slug, path, rc in results[:10]:
    print(f'{score:3d} matches  [{rc}x referenced]  {slug}  {path}')
"

Merge results from CLI/grep and atoms-index, deduplicate by path, sort by relevance.

4. Search team vault (if enabled)

Check config for team.enabled. If true:

If HAS_OBSIDIAN_CLI and team vault is registered as an Obsidian vault:

TEAM_VAULT_NAME=$(basename <team_vault_path>)
obsidian vault="$TEAM_VAULT_NAME" search query="<QUERY>" limit=20 2>/dev/null