PREREQUISITE: Read deepvista-shared, deepvista-vistabase, and deepvista-notes.
Search, retrieve, and analyze notes from the knowledge base to surface insights, patterns, and summaries.
Search for relevant notes using a query derived from the user's request: bash
deepvista card +search "<topic or keyword>" --type note --limit 20
List recent notes if no specific topic was given, to get a broad view: bash
deepvista notes list --limit 20
Fetch full content for the most relevant notes (pick IDs from search/list results): bash
deepvista notes get <note_id>
Repeat for each note you need to read in full.
Analyze and synthesize — read the content and identify:
Present findings to the user as a structured summary.
Optionally save the analysis back as a new note (confirm with user first): bash
deepvista notes create --title "Analysis: <topic> — <date>" --content "<synthesis>"
[!CAUTION] Write command — confirm with user before saving.
Use card +search with specific keywords rather than listing everything — it uses hybrid vector+keyword search and returns the most relevant results.
Filter by type to stay focused on notes: deepvista card +search "<query>" --type note.
For time-bounded analysis ("notes from this week"), use notes list and filter by created_at in the JSON output.
Use chat +send to ask the AI agent to synthesize across a large set of notes:
bash
deepvista chat +send "Here are my recent notes: <paste content>. What are the key themes?"
bash
# Find all notes about a project
deepvista card +search "project alpha" --type note --limit 15
# Get full content of a note
deepvista notes get note_abc123
# Save analysis as a new note
deepvista notes create \
--title "Weekly Themes — 2026-04-02" \
--content "## Key Themes\n- Theme 1\n- Theme 2\n\n## Open Questions\n- ..."