Plugin: Jarvis (AI assistant) - also on mobile [v0.13.0, 2026-05-04]

Thank you for taking the time to address my concern. The information is invaluable. I knew there was more I could get out of Jarvis but didn't really understand how.

Joplin is a wonderful app, but Jarvis is the icing on the cake.

May you have a wonderful day.
Don

shikuz,

I was wondering whether the full note is inserted into the database. From your description, I get the impression the full note is inserted. Then as chats evolve, appropriate chunks of relevant notes augment the user's query. Is this correct?

Don

Indeed Don, the full note (and extracted OCR) is inserted into the database, but it is first chunked, and each chunk is indexed independently so that the most relevant chunks are added to the chat.

So I understand, are you saying there is 2 copies of each note in the database? One copy being the full text and a second version that covers the full note but is a series of chunks. Is this correct?

Don

The database in Jarvis only references where the chunks are within notes (such as the note ID, line number), so there is only a single copy of the content of your notes. The database is a SQLite, so you can open it and have a look around if you'd like.

Thank you. That's good to know.

Have a great day,
Don

happy to announce this special release! :partying_face:

v0.12.0

  • added: mobile support (#62)
    • all code migrated, and most commands accessible on mobile / web
    • new database based on note properties (setup guide)
      • syncs between devices and backed up with notes
      • more memory efficient with in-memory cache and Q8 quantization
    • device profile setting with platform-aware tuning
  • added: model management dialog
    • manage your embedding models: view stats, delete, or switch models
  • added: setting: 'None' embedding model option
    • disable embedding features (similar to the existing 'None' chat model option)
  • added: setting: toolbar button visibility
    • most commands are now accessible from the toolbar
    • configure which Jarvis toolbar buttons to display (chat, notes, find, edit, autocomplete, annotate)
  • added: release notes dialog
  • improved: upgraded models list
    • gpt-5.2
    • claude-4.5
  • improved: memory management
  • improved: decrease package size and plugin loading speed
  • improved: truncate long error messages from APIs
  • improved: truncate long lines in text blocks during chunking
  • improved: strip Jarvis-generated blocks from context in annotations, chat, auto-complete, and embeddings
  • improved: skip related notes search for excluded notes
  • improved: preserve embeddings for notes in excluded folders
  • improved: progress display with stage messages during database updates

This is an absolute bangers. I will begin testing after the New Year’s. Deep gratitude!

Hi, Love this so far but I’m trying to work out how to get rid of the related notes panel and so far all I’ve come up with is renaming it to GOAWAY (it didn’t work)
please help

Hi @ItMeCorban, try to look under the Tools -> Jarvis menu for the command Toggle related notes panel.

Amazing thank you! That’s exactly what I was looking for, didn’t notice the separate Jarvis menu before, that’ll come in handy :slight_smile:

v0.13.0

jarvis-chat-panel

  • graduated: note properties database out of experimental (still opt-in)
  • added: chat panel for notes (#69, by @developerzohaib786)
    • see screenshot
    • added: chat / notes mode toggle in the chat panel (Shift+Tab)
  • added: hybrid retrieval (semantic + keyword) for chat with your notes (default: on)
  • added: LLM query decomposition for chat with your notes (default: on)
  • added: multi-chunk search for the related notes panel (default: on)
    • returns multiple blocks per note ranked by MaxSim, with keyword reranking on the search box
  • added: inter-plugin API (#72)
    • other plugins can call jarvis.api.status and jarvis.api.search to query Jarvis programmatically (see discussion)
    • see docs for the reference
  • added: settings: Chat: Panel visible, Notes: Panel visible, Notes: Keyword search weight, Notes: Keyword search RRF k, Notes: LLM query decomposition, Notes: Multi-chunk search
  • removed: setting Notes: Number of nearest blocks to add
  • improved: reference-style note citations
  • improved: upgraded models list (gpt-5.5, claude-opus-4-7, claude-sonnet-4-6, gemini-embedding-2)
  • improved: migrated to the new Google GenAI SDK (closes #64)
  • improved: OpenAI-compatible endpoints accept api-key auth for Azure-style gateways (#66, by @developerzohaib786)
  • improved: removed model communication test on startup (closes #65)
  • improved: chat context grouping by note and sorted by original order
  • improved: error surfacing
  • improved: notes context preview dialog aligned with chat panel style
  • fixed: Anthropic models on mobile (opt in to browser access for the WebView)
  • fixed: block attachment (prev/next/nearest) inactive in note properties database mode (#71)
  • fixed: folder-excluded notes triggered a sync ping-pong in note properties database mode
  • fixed: model dropdown labels reordered for readability on small screens
  • perf: bulk-fetch settings via joplin.settings.values()