1.3.4: Bug Fix Update

1. Remove constant logging: a recent release added a feature that keeps the most up-to-date system info to be fed into API calls. But this was not optimal because it kept refreshing the machine every few seconds even when nothing is running. Removed this and instead queries the system info only right before making an RPC call, which has the same effect but no need to keep querying.
2. Registry fix: Some people experienced issues with recent releases where Pinokio kept saying the "registry" module is not installed. This was because of some terminal parsing issue, and has been fixed.
3. Proxy port: Do NOT use the default 8000 to create proxies. Instead, the proxy ports start from 42420. When a proxy.start API call is made, Pinokio scans open ports starting from 42420 and starts a proxy using the first discovered port.

Pinokio 1.3.0 : Neuron

Pinokio script is now a full fledged AI scripting language. With this release, you can script ANYTHING AI related: installing, running, orchestrating, etc.

Also, the scripts can interact with other scripts in a distributed manner, while saving disk space and memory.

1. Modules
2. Distributed URI
3. Conditional Execution
4. Jump
5. Programmable Conda + Venv
6. Gradio Script

Also in v1.3.0,

• Windows Canary Support: removed the windows canary restriction (previously windows canary had some bugs that made it impossible to run pinokio, but it has been fixed, so update to the latest canary version of windows, and update pinokio to this version, and it should work now)
• Robust Linux CUDA Support: supported with the new programmable conda + venv script
• Minimal mode: minimize/maximize the terminal to hide/show the sidebar
• Various Bug fixes

1. Modules

With the new "script" API, scripts can call other scripts, process its return values, and even access local memory:

1. download another script on the fly
2. start/stop other scripts
3. call and respond to other scripts

all programmatically.

2. Distributed URI

Any script can call any other script modules through the distributed git URI scheme, fully making use of:

1. published git uri
2. commit hash
3. git branch
4. file path

Simply specifying a git URI will automatically download and execute the script.

3. Conditional Execution

Every step in a Pinokio run script is a JSON RPC call, complete with "method" and "params". But often you want to run a step ONLY IF certain conditions are met (ex: run commands for certain OS/GPU/Arch).

The "when" attribute lets you achieve this now.

4. Jump

Officially introducing "jump" API, and an "id" attribute for each RPC call, which lets you jump to any index or an id within the run script. Even supports passing parameters when jumping.

5. Programmable Conda/Venv API

With just one line in the JSON script, you can:

1. create a conda/venv if it doesn't exist
2. activate a conda/venv environment

Can handle pretty much every scenario for installing & running AI apps in isolated environments, saving storage.

6. Gradio Script

Every gradio app has a built-in API and a JS/Python API client that can interact with it.

Pinokio script has introduced a new "gradio.predict" API that lets you make requests to ANY gradio app, using a JSON based script.

Great for building agents & workflows.