API Interface
A refined API interface for conversing with Aīris through Anthropic, OpenAI, and local Ollama models — direct, customisable, and built for depth of exchange.
Features
Anthropic, OpenAI, and local Ollama support
Choose the provider that suits your needs — Anthropic, OpenAI, or a local Ollama model — and connect with your own credentials through the same refined interface.
Parameter Control
Creativity, focus, system prompts, and token limits are all visible and adjustable. Nothing is hidden behind default settings.
Session Threads & Export
Organize conversations into threads, preserve them as you work, and export the results to plain text, JSON, or Markdown.
Usage Tracking
Live token and cost estimates keep spending in view as you experiment or build.
×
Using Ollama (free, runs on your computer)
Ollama lets you run AI models privately on your own machine — no account needed, no data sent anywhere. Choose your system below for step-by-step instructions.
Step 1 — Install Ollama
If you already have Homebrew, open Terminal and paste this:
brew install ollamaNot sure what Homebrew is? Paste this larger block instead — it installs Homebrew first, then Ollama. It is safe, takes about two minutes, and may ask for your Mac password once.
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)" && brew install ollamaStep 2 — Quit the Ollama app from the menu bar
If Ollama is already installed and showing in your top menu bar, right-click its icon and choose Quit Ollama before continuing. The app version does not allow Instrumenta to connect to it — you need to start it from Terminal instead.
Step 3 — Start Ollama with web access enabled
In Terminal, paste this exactly:
OLLAMA_ORIGINS=* ollama serveThis starts Ollama with permission for Instrumenta (running in your browser) to talk to it. Keep this Terminal window open while you use Aīris Chat.
Step 4 — Choose a model
Open a second Terminal window and pull a model. These are our recommended models and their respective 'pull' commands. You only need to do this once for each model. After that it is stored on your machine.
Gemma 3 · 1B
Good for simple, everyday tasks — quick questions, short summaries, basic writing help. The lightest option; runs on almost any modern laptop with 4 GB of RAM or more.
ollama pull gemma3:1bGemma 3 · 4B
A solid all-rounder for everyday use — handles longer conversations, light analysis, and creative tasks well. Suited to Apple M series machines with 8 GB RAM or higher.
ollama pull gemma3:4bGemma 3 · 12B
Noticeably better at complex reasoning, detailed writing, and nuanced conversations. Suited to Apple M series machines with 16 GB of RAM or higher.
ollama pull gemma3:12bGemma 4 Uncensored · Q6 (2B)
The lighter uncensored option — no content restrictions, good for unrestricted creative writing or shorter conversations. Suited to a Intel-based Apple machines or Apple M series machines with 8 GB RAM or higher.
ollama pull hf.co/HauhauCS/Gemma-4-E2B-Uncensored-HauhauCS-Aggressive:Q6_K_PGemma 4 Uncensored · Q6 (4B)
The more capable uncensored model — better for longer, more involved conversations with no restrictions. Suited to Apple M series machines with 8 GB RAM or higher.
ollama pull hf.co/HauhauCS/Gemma-4-E4B-Uncensored-HauhauCS-Aggressive:Q6_K_PGemma 4 Uncensored · Q6 (26B)
A higher-capacity uncensored option for demanding creative work, longer conversations, and more complex prompts. Suited to Apple M series machines with 32 GB of RAM or higher.
ollama pull hf.co/HauhauCS/Gemma4-26B-A4B-Uncensored-HauhauCS-Balanced:Q6_K_PStep 1 — Install Ollama
Open PowerShell as administrator and paste this:
winget install Ollama.OllamaIf Winget is not available, download the installer directly from ollama.com and run it.
Step 2 — Quit Ollama from the system tray
If Ollama is running in the system tray (bottom-right of your taskbar), right-click its icon and choose Quit. You need to start it from the command line so that Instrumenta can connect to it.
Step 3 — Start Ollama with web access enabled
Open PowerShell and paste this:
$env:OLLAMA_ORIGINS="*"; ollama serveKeep this window open while you use Aīris Chat.
Step 4 — Choose a model
Open a second PowerShell window and pull a model. These are our recommended models and their respective 'pull' commands. You only need to do this once for each model. After that it is stored on your machine.
Gemma 3 · 1B
Good for simple, everyday tasks — quick questions, short summaries, basic writing help. The lightest option; runs on almost any Windows laptop with 4 GB of RAM or more.
ollama pull gemma3:1bGemma 3 · 4B
A solid all-rounder for everyday use — handles longer conversations, light analysis, and creative tasks well. Suited to Windows machines with 8 GB of RAM or higher .
ollama pull gemma3:4bGemma 3 · 12B
Noticeably better at complex reasoning, detailed writing, and nuanced conversations. Suited to a Windows laptop or desktop with 16 GB of RAM or higher.
ollama pull gemma3:12bGemma 4 Uncensored · Q6 (2B)
The lighter uncensored option — no content restrictions, good for unrestricted creative writing or shorter conversations. Runs comfortably on Windows laptops with 8 GB of RAM or higher.
ollama pull hf.co/HauhauCS/Gemma-4-E2B-Uncensored-HauhauCS-Aggressive:Q6_K_PGemma 4 Uncensored · Q6 (4B)
The more capable uncensored model — better for longer, more involved conversations with no restrictions. Suited to Windows laptops with 8 GB of RAM or higher.
ollama pull hf.co/HauhauCS/Gemma-4-E4B-Uncensored-HauhauCS-Aggressive:Q6_K_PGemma 4 Uncensored · Q6 (26B)
A higher-capacity uncensored option for demanding creative work, longer conversations, and more complex prompts. Suited to Windows systems with 32 GB of RAM, or a dedicated 12 GB GPU.
ollama pull hf.co/HauhauCS/Gemma4-26B-A4B-Uncensored-HauhauCS-Balanced:Q6_K_PStep 5 — Connect Ollama inside Aīris Chat
Open Aīris Chat, choose the Ollama provider in settings, keep the default server URL, and press 'test key'. Select the model you downloaded. That is all — Instrumenta will connect to your machine directly.
Performance note
Performance varies widely based on how many applications you have open and your specific device type. The best way to know is through trial and error and testing models yourself. However, please note that attempting to run a model that exceeds your device's capacity may render it unresponsive, requiring a forced restart to recover.
Using OpenAI instead
Go to platform.openai.com, sign in, create an API key, and paste it into the OpenAI settings in Aīris Chat. You will be charged per use by OpenAI directly — Instrumenta never sees your billing.
Using Claude instead
Go to console.anthropic.com, sign in, create an API key, and paste it into the Claude settings in Aīris Chat. You will be charged per use by Anthropic directly — Instrumenta never sees your billing.