Skip to content

Releases

Free AI and Agents 1.7.2

Choose a tag to compare

@github-actions github-actions released this 13 Nov 07:02

Full Changelog: 1.7.1...1.7.2

Free AI and Agents 1.7.1

Choose a tag to compare

@github-actions github-actions released this 13 Nov 06:30

Full Changelog: 1.5.5...1.7.1

Free AI and Agents 1.6.0

Choose a tag to compare

@mokelu mokelu released this 06 Nov 03:00

Free AI and Agents 1.6.0

A big feature release — voice, sections, connectors, and more.

What's new

  • Self-hosted voice input (Riemann STT) — the mic now uses our own server by default; per-device keys are auto-issued (no setup, no tokens to manage). Cloud providers remain as alternatives.
  • Sections: Chat / Play / Study / Automation — top-level tabs; Play, Study and Automation render server-fed pages (coming soon placeholders today, wired to load whenever we publish them).
  • Connectors (MCP) — a catalog of hosted connectors with enable/disable toggles, "coming soon" badges, credential hints, and live tool cards. Tools merge into the agent automatically.
  • Quick Prompts — translate, summarize, calculate, explain, ask-about-note now inject into the composer for editing, with an optional (persisted) Autosend toggle.
  • Report a problem menu — Email (reply within the hour), Discord (Beta), GitHub, Reddit (coming soon), feature requests, and Patreon support.
  • i18n — follows Obsidian's language; 15 language packs covering the core UI.
  • Quota UX — offline state, simplified popup, reset display updates.
  • Chat history window — trimmed to 10 exchanges with an intelligent 3–5 sentence rolling summary layer (raw-only, no summary-of-summary drift).
  • Scroll behavior — viewport anchoring: no more auto-jumps when reading mid-chat.

Maintenance

  • New display/chat settings, font stepper in the header, transcribing status strip.

Install from Obsidian's Community plugins, or manually from the releases page.

Free AI and Agents 1.5.5

Choose a tag to compare

@mokelu mokelu released this 27 Oct 15:45

This update makes the app steadier, richer and more capable — and it's the fastest way to unlock your full 5M free tokens per device (up from 1M).

What's new

  • Stable Controls — the model selector, reasoning-effort toggles and settings panel are now dependable; changes take effect immediately and hold across turns.
  • Image display — attachments and image results now render inline in the chat, so you can see what the model is working with instead of a bare link.
  • New models — added Mercury 3 (with Instant / Low / Medium / High thinking modes), Nemotron 3.5 Lightning, Gemma 4 31B, Mistral NeMo, and Qwen3.8 Flash.
  • 5M free tokens per device — the full free tier (up from 1M). Current, older builds may not reflect the new models or the larger quota until you update.

How to update

Obsidian → Settings → Community plugins → Update (or "Browse" → Free AI and Agents → Update). Restart Obsidian afterward to load the new build.

1.5.4

Choose a tag to compare

@mokelu mokelu released this 24 Oct 06:46

Voice input becomes a setting of its own instead of a Groq-only mic, thinking level moves out of the model name, and catalog providers can be added from their deal card.

  • Voice input: until now the mic worked only through Groq. It is now a Voice model setting where the provider and the model are chosen — Whisper Large v3, Whisper Large v3 Turbo, GPT-4o mini Transcribe, GPT-4o Transcribe — and the caption says which one is in use.
  • Thinking level: effort used to be faked by extra model entries such as "Mercury 2.5 (Think: High)". Those entries are retired and migrated away. Models that support it now show level buttons under the selected model, so effort is set in the chat instead of by switching model names.
  • Catalog providers: a provider listed in the catalog can be added from its deal card with "I have a key", instead of being shown with no way in.
  • The Jina caption now reads "the included free allowance" instead of "the included 30,000 tokens".
  • The manifest description was rewritten to say what the plugin does: free monthly tokens with no signup, plus your own key for other providers.

1.5.3

Choose a tag to compare

@mokelu mokelu released this 20 Oct 09:07

Cleaner review report. No feature changes.

  • Removed leftover rgba() fallback declarations that sat next to the modern color-mix() values in the stylesheet; the linter flagged them as duplicate background declarations. color-mix() is supported by every Obsidian version this plugin allows (1.4.0 and newer), so the fallbacks were dead weight. Same appearance, no CSS lint warnings.
  • Added a GitHub Actions workflow that creates a build-provenance attestation for main.js, manifest.json and styles.css on each release, so the release assets can be verified with gh attestation verify.

1.5.2

Choose a tag to compare

@mokelu mokelu released this 20 Oct 08:51

Thinking levels for the free pool, and a real conversation title.

  • Mercury 2.5 is now offered as separate pool models: Think: Medium, Think: High, Think: Low and Think: Off. They are the same model at different reasoning depths, so you can choose speed or depth. Think: Off answers in about a second.
  • The plugin now sends the chosen thinking level to the service, which passes it on to the model provider.
  • The chat header title is now the model-written title, produced by the same model you are chatting with, instead of a guess built from your first message. Until that title arrives, the old placeholder is shown.

1.5.1

Choose a tag to compare

@mokelu mokelu released this 20 Oct 07:41

First public release. Free pool chat, vault agents with approvals, web tools (web_read, web_search), 20+ bring-your-own-key providers, voice typing, local history with restore.