Skip to content

Blog

What's new in Aori v0.4.1

Aori v0.4.1 is mostly about making the things already in the app work more smoothly.

There is no single headline feature this time. Instead, we spent this release improving the parts you touch every day: conversations, Personas, memory, tools, approvals, and AI providers. The goal is simple: fewer interruptions and a more dependable Aori across macOS, Windows, Linux, and Android.

Conversations should now feel steadier and more predictable. We worked through a collection of issues around everyday chat behavior so that starting, continuing, and managing conversations is smoother.

These are not flashy changes, but they matter. A personal AI app needs to be reliable in the middle of an ordinary conversation, not only when everything goes exactly as expected.

We also tightened up how Personas and memory behave.

Switching between Personas and continuing conversations with the right context should feel more consistent. Memory-related behavior has been refined too, so Aori can make better use of relevant context without making the experience feel unpredictable.

Tools are useful only when it is clear what Aori is doing and when it needs your permission.

In v0.4.1, tool execution and approval flows are more reliable. The experience should be smoother when Aori needs confirmation before taking an action, and less likely to leave you wondering whether something is still waiting or has already completed.

Different AI providers have their own APIs, models, and occasional quirks. This release improves provider support and makes those integrations more dependable across the app.

You still choose and connect your own provider. The work in v0.4.1 is about making that choice behave more consistently once it is set up.

The same stability work applies across the current macOS, Windows, Linux, and Android builds. Each platform has its own rough edges, and this release smooths out a number of them.

Aori is still in public beta, and there is more polishing to do. v0.4.1 is a practical step in that direction: less about adding another item to the feature list, and more about making Aori nicer to live with every day.

— The Aori Team

What's new in Aori v0.4.0

Aori v0.4.0 is not a giant “everything is solved now” release. It is a quieter one, but it changes the shape of the app in a way that matters.

The focus this time is simple: Aori should feel more like your assistant, and it should be able to work with the notes and knowledge you already have.

That means three things landed together: Personas, Obsidian, and Notion.

Personas are the most visible change.

Until now, Aori mostly behaved like one assistant. That works for basic use, but it is not how people actually use AI day to day. Sometimes you want a practical assistant. Sometimes you want a more casual companion. Sometimes you want a very specific voice for a specific workflow.

In v0.4.0, Aori supports custom Personas. There are Assistant and Friend-style presets to start from, and you can create your own.

You can also switch Personas with slash commands, including from the in-app chat and supported external channels. That matters because the Persona should follow the conversation you are actually having, not be trapped inside one settings screen.

Obsidian support is another big step toward making Aori useful with real personal context.

You can point Aori at an Obsidian vault and let it use that vault as a knowledge source. Aori can read, index, and search supported notes, then bring the relevant pieces into a conversation when you ask for them.

It can also create new notes or append to notes where the feature allows it. The important boundary is that Aori is not meant to autonomously delete or directly rewrite your existing notes. Your vault is still your vault.

This is the kind of feature that is easy to overhype, so I will keep it plain: if you already use Obsidian, Aori can now start to become useful around the notes you already have instead of asking you to move everything somewhere else.

Notion is now supported too.

The idea is similar, but the shape is different because Notion is a cloud workspace. You can connect a Notion account and let Aori use allowed pages or workspace content as a knowledge source.

That makes Aori more useful for people who keep project notes, docs, planning pages, or lightweight databases in Notion. Ask a question, and Aori can use the connected Notion knowledge instead of relying only on the current chat.

Because Notion is a cloud integration, the usual data-flow rules still apply. Aori can access what you authorize, and if you use a cloud AI provider for a request, relevant context may be sent to the provider you selected.

Personas make Aori feel less generic. Obsidian and Notion make it less empty.

That combination is the point of v0.4.0. A personal AI app needs more than a chat box. It needs a way to adapt to the kind of conversation you want to have, and it needs access to the places where your actual context lives.

Aori is still in public beta. Some parts are rough, and release builds are still being prepared. But this release makes the direction clearer: local-first where it can be, connected where you choose, and personal without trying to own all of your data.

— The Aori Team

How I Use Aori Every Day

One of the most common questions we get is: “What do you actually use Aori for?”

It’s one thing to read a list of features, but another to see how a personal agent actually fits into the messy, unpredictable flow of a real day. Here is a typical day with the current Aori release.

My day doesn’t start with me checking every app manually; it starts with Aori checking in with me. I have a Routine set up that fires every weekday morning at 8:00 AM.

Aori sends a message to my private Telegram bot. It summarizes the context I have connected, highlights a few things I asked it to watch, and gives me a simple starting point for the day.

While I’m out for a walk, I remember I need to reschedule a meeting. I don’t stop and pull up my calendar app. Instead, I send a quick voice message to my Aori bot on Telegram:

“Aori, can you move my 2 PM sync with Robert to tomorrow at 10 AM and send him an email letting him know?”

Because voice input is connected, Aori transcribes the message and turns it into a clear request. I can then continue the conversation from Telegram or move back into the in-app chat without losing the thread.

1:00 PM — The “Second Brain” at Work

Section titled “1:00 PM — The “Second Brain” at Work”

During a deep work session on my desktop, I come across an interesting idea. I open the Aori chat and say:

“Aori, make a note: We should explore using vector-based retrieval for the new documentation engine.”

Aori can create or append a note in the connected folder or Obsidian vault. Later, if I ask, “What were my ideas for the documentation engine?”, it can search the connected knowledge source and bring the relevant note back into the conversation.

I have a deadline today at 5:00 PM. Aori knows this because I added it to my to-do list earlier. Around 4:00 PM, I get a push notification on my Android phone:

“Hey, just a heads-up that you have that documentation deadline in an hour. Anything I can help you wrap up?”

This isn’t a generic alarm; it’s a context-aware nudge from my agent. I ask it to quickly summarize the last few emails from the team about the project so I have all the context I need to finish.

As I’m winding down, I ask Aori for a quick summary of my day:

“Aori, what did we get done today?”

Aori lists the calendar changes, the emails it prepared, the notes I made, and the follow-ups we handled. It feels less like a log and more like a brief conversation with a helpful assistant. It’s a great way to close out the day with a sense of accomplishment.

The reason Aori fits the workflow is simple: it meets me where I am. I don’t have to go to a specific “productivity app” to get things done. Whether I’m on my laptop, my phone, in Telegram, WhatsApp, or in-app chat, my agent is available with the files, notes, and Personas I have chosen to connect.

Everyone’s workflow is different. Aori doesn’t force you into a specific system; it adapts to how you describe your needs.

Ready to build your own workflow? Recommendation for new users: Start with Groq — it’s free, fast, and requires no credit card.

Get Started with Routines

— The Aori Team

Which LLM Should I Use With Aori?

One of Aori’s most powerful features is that it is model-agnostic. You “Bring Your Own Key,” which means you aren’t locked into a single provider’s ecosystem. But with so many options—Groq, OpenAI, Anthropic, DeepSeek, Ollama—how do you choose?

Here’s our guide to the best LLM setups for the current Aori release.

If you’re just getting started with Aori, there is no better choice than Groq.

Groq offers incredibly fast inference and a generous free tier that is perfect for daily agent use. It supports high-quality open-source models like Llama 3.3 70B, which is more than capable of handling Aori’s routines, memory extraction, and general chat.

Recommendation for new users: Start with Groq — it’s free, fast, and requires no credit card.

When you need the absolute best in reasoning, nuance, and long-context performance, Anthropic’s Claude series is the gold standard.

  • Claude 3.5 Sonnet: The ideal “power user” model. It excels at complex task planning and following intricate instructions.
  • Claude 3.5 Haiku: A faster, more affordable alternative that still maintains high intelligence for daily tasks.

Anthropic typically offers a $5 credit on signup, which can go a long way with Aori.

OpenAI remains a popular choice for its reliability and wide range of features, including native vision support.

  • GPT-5.x (Flagship): The most capable model for highly complex agent tasks and deep reasoning.
  • GPT-4o (Popular): A versatile, high-performance model that balances speed and intelligence perfectly.
  • GPT-4o mini: An extremely cost-effective option for simple queries and high-volume tasks.

OpenAI is a strong choice if you want broad model coverage, native vision support where available, and predictable API behavior across many task types.

If you want to run models from your own desktop machine, Ollama is the option to look at.

Ollama runs open-source models (like Llama 3.2 or Mistral) on your computer’s hardware. Aori Desktop connects to your running Ollama server. Aori does not bundle local models or provide mobile local inference in this launch.

DeepSeek has quickly become a favorite for users who want flagship-level performance at a fraction of the cost. Their reasoning models are surprisingly strong, making them an excellent choice for heavy daily use on a budget.

ProviderFree Tier?SpeedQualityBest For
GroqYesExtremeHighGetting started, daily use
Anthropic$5 CreditFastEliteNuance, complex reasoning
OpenAI$5 CreditFastEliteFlagship power (GPT-5.x), vision
DeepSeekYesMediumHighHigh-volume tasks, budget power
OllamaFree local runtimeVariableGoodDesktop local models
OpenRouterNoVariableMixedAccessing niche models

Don’t overthink it. Most users should start with Groq. It’s the easiest way to see what Aori can do without spending a dime.

As your needs grow—if you find yourself needing more nuanced reasoning or longer context windows—you can switch to Claude 3.5 Sonnet or GPT-5.x in seconds via Aori’s settings. Your memory and routines will follow you wherever you go.

View all Provider Guides

— The Aori Team

Introducing Aori — your personal AI agent

We’re excited to share the first public version of Aori: v0.1.0.

Over the last few years, we’ve seen an explosion of powerful large language models. But for most of us, using them still means opening a separate website, pasting context again, and treating the assistant like a destination instead of part of the workflow.

We built Aori because we wanted a more personal assistant experience: local-first, connected to the places you already chat, and under your control.

Aori is a cross-platform personal AI agent designed to run as a desktop and Android app. You can chat with Aori in-app, from Telegram, or from WhatsApp. On desktop, you can also connect local knowledge sources such as a folder or Obsidian vault.

Aori is not an Aori-hosted cloud assistant. You bring your own AI provider key from providers such as Groq, Anthropic, OpenAI, Gemini, DeepSeek, or OpenRouter. Aori uses that key to talk to the model you choose. On desktop, Aori can also connect to Ollama.

1. It meets you where you already talk Aori supports in-app conversations, Telegram, and WhatsApp. It keeps chats and threads separate, so context from one conversation does not get mixed into another.

2. It works with your files and notes Aori can use connected knowledge sources such as Local Folder, Obsidian, and Notion. It can read, index, and search connected notes, and it can create or append files in places you explicitly grant access to.

3. It lets you shape the assistant Personas let you switch between assistant-style and friend-style presets, or your own custom behavior. You can switch Personas with slash commands across channels, including in-app chat.

Aori v0.1.0 is an early public release. It focuses on the core experience:

  • Desktop and Android app support.
  • In-app chat with multiple conversations.
  • Telegram and WhatsApp channels.
  • Bring-your-own-key AI provider setup.
  • Personas and slash-command switching.
  • Knowledge sources including Local Folder, Obsidian, and Notion.
  • Voice input via supported speech-to-text providers.

The fastest path is:

  1. Download Aori for your platform.
  2. Get a free API key from Groq. Recommendation for new users: Start with Groq — it’s free, fast, and requires no credit card.
  3. Paste your key into Aori and say “Hello.”

We think the future of AI is not just bigger models, but better personal agents: tools that work across your chats, files, and everyday context without forcing everything into one hosted service.

Download Aori or read the current docs to get started.

— The Aori Team