cocktail peanut

Pinokio 3.3.0

Pinokio is an AI automation engine that orchestrates programs that are not 100% reliable (shell, conda, pip, etc.). Reliability is important.

Today releasing 3.3.0. A major update fixing most of the core reliability issues that have existed from the beginning.

The official install link has been updated to 3.3.0.

Get 3.3.0 here: https://pinokio.computer

1. Major Fix Update

Going into 2025, I finally decided to sit down and fix all the core issues that have been plaguing Pinokio users.

- homebrew fix (mac)
- conda fix
- pip fix
- linux fix
- playwright fix
- built-in python RPC server (uvicorn) fix
- "run in admin" mode fix

2. Core Engine Major Upgrades

The core parts of Pinokio are:

1. Shell execution & streaming & parsing engine
2. Web terminal

Pinokio 3.3.0 has updated both of them (which haven't been updated since the beginning) so most of the bugs related to the core modules are gone.

3. Autolaunch Apps & Scripts

You know how you can set "start programs" on your OS to automatically launch?

You can now configure any script to automatically run when Pinokio launches. Just set PINOKIO_SCRIPT_AUTOLAUNCH=<SCRIPT_PATH>

Example ComfyUI autolaunch:

4. Use ANY version of Python

By default Pinokio comes with Python 3.10. This works for most apps but in some cases you need Python 3.11 or 3.12.

Thanks to the built-in UV, you now can use any version of python. As easy as specifying a "venv_python" attribute when calling shell

5. "next" RPC attribute

A new API has been added to the RPC spec: "next".

This is like the "jump" API https://program.pinokio.computer/#/?id=jump but you don't need to make additional calls just to jump. Just specify where to jump to after the current step.

Pinokio script is Turing complete.

6. Info API

Realtime overview of Pinokio processes:

1. pinokio version
2. which scripts/steps are running
3. which shells are running and what they are displaying
4. local variables
5. installed apps
6. local sharing endpoints

Accessible at http://localhost/pinokio/info

The info API is very comprehensive, you can use it to build:

1. 3rd party apps that use Pinokio powered apps as backend
2. Your custom pinokio via PSP

more on psp https://x.com/cocktailpeanut/status/1869409330241339894

Read the full release note for Pinokio 3.3.0 here: https://github.com/pinokiocomputer/pinokio/releases/tag/3.3.0

Pinokio 3.3.0 has more low-level updates than any other past Pinokio updates, fixing A LOT of bugs.

But if you have any new issues with 3.3.0 please post on pinokio-support channel on Discord. I'll be closely monitoring for the next couple of days.

https://discord.com/channels/1121039057993089076/1131623060315844710

Several Linux users have reported installation issues, so I've worked on a hotfix.

Just published 3.3.7 and this is the official latest version on the homepage as well. it should all work now on all OS https://x.com/cocktailpeanut/status/1882233549476204954

Made another hotfix 3.3.8 after some people updated issues with Fluxgym installs. This update applies to all users, not just linux.

You may already have this version since i published yesterday, but check your version just to make sure.

https://x.com/cocktailpeanut/status/1882521907649757326

Pinokio 3.6

The show must go on. Releasing Pinokio 3.6 today.

This will probably be the last major update until the next major update, and I strongly recommend getting it.

Fixes a lot of important components and minimizes chances of running into edge cases.

You can get the latest version by:

1. Going to pinokio home http://pinokio.computer
2. Download
3. Just install (No need to uninstall the old version), and it will overwrite only the pinokio binary, while keeping all the installed apps and data in tact.

Port 42000

Pinokio started out with port 4200, then switched to 42000, and then 80. And then port 80 or 42000 depending on which port is available. But this is needlessly complicated and confusing.

Starting with 3.6, Pinokio port has been officially fixed to 42000.

The idea behind port 80 was to make the URL much more user friendly (localhost vs. localhost:42000) but I actually have a better idea for this, so no need for all this complexity.

Pinokio port is 42000 on ALL platforms. http://localhost:42000

UV Fix

There are many great things about UV but one really powerful feature is you can use any version of python in a virtual environment. This feature was incorporated into Pinokio script but I found that there were some quirks with UV that introduced unexpected behaviors.

If you ever ran into "python version" issue when installing any app (like Open-WebUI for example), this was. it.

This has been fixed and UV now fully creates and uses isolated Python inside Pinokio, which makes everything deterministic and predictable.

Good Bye Permission Errors

If you're a Windows user, you have probably run into "EACCESS Permission Error" sometimes.

Normally this shouldn't happen but sometimes they happen, and it's because the code doesn't have WRITE access. It's a simple issue that's quite annoying when it happens.

I got sick of this error (and you too probably), and starting with Pinokio 3.6, everything that happens inside will NEVER run into this issue because:

1. Everything is installed under ~/pinokio (Pinokio Home)
2. Now the Pinokio Home folder is given FULL write access.

Visual Studio Build Tools Auto-install Fix

Visual Studio is the ONLY piece of software that is NOT installed inside Pinokio home. It's installed under your Program Files folder.

Because of exception, there were various edge case scenarios where people would get stuck in a "install visual studio build tools" loop.

A lot of this has been resolved, and even if you are one of the few who still have this issue, now Pinokio displays instructions on how you can resolve this (Basically uninstall existing Visual Studio Build tools 2019 and reinstall Pinokio)

Dealing with Visual Studio has been quite a journey and hope this fixes it all for you. Otherwise, follow the instructions suggested in Pinokio.

Fix for Various Other Bugs Introduced in 3.X

When I first started working on Pinokio 3, I really didn't think it would suck this much time and energy out of me. 

I had a lot of plans for January but ended up spending the entire month of January dealing with issues introduced in v3.

I think I've reached a point where we can finally move forward, and things are much more stable and reliable than ever.

This means things will move much faster now.

More details https://github.com/pinokiocomputer/pinokio/releases/tag/3.6.0

Local Autonomous Agent for All of Us

Browser-Use Web UI is a gradio app that uses LLMs to automatically surf the web to achieve tasks. All you need to do is just tell it what to do. It works like magic.

And now anyone can run it locally with 1 click.

It runs on ALL OS (Mac, Windows, Linux)

It works with ALL LLMs (OpenAI, Anthropic, Deepseek, Gemini, Ollama, Mistral, ..)

It "just works". 

Here's the official github https://github.com/browser-use/web-ui

Agents in 2025

One reason why I really tried hard to NATIVELY integrate Playwright into Pinokio despite the complexity (introduced in v3.x) was:

We will see a lot of innovation around agents in 2025, and wanted to make sure it's as easy as 1-click for all of them.

You will need Pinokio 3.6 (The latest version) to run this. https://x.com/cocktailpeanut/status/1885721335961129089

DeeperHermes

DeepHermes can act as a simple AI agent or a deep thinking AI, just by passing a system prompt.

But what if we can go DEEPER? The LLM decides on its own how to respond by...asking itself! (No system prompts)

I made a gradio app powered by @lmstudio, it works!

What's going on above?

1. First &amp; Second questions are simple, so the model acts as a basic agent ("capital of china", "3+5")
2. Third question is complex, requires resasoning. So starts with [THINKING] then ends with an [ANSWER].

All on its own!

How it works:

Step 1. When the user asks a question, DeepHermes asks itself which mode it should respond with, based on the question. This generates the relevant system prompt.

Step 2. The system prompt from step 1 is used to actually ask the original question.

Here's the github https://github.com/cocktailpeanut/deeperhermes

Also put it on Pinokio so you can also easily play with it with 1-click. https://pinokio.computer/item?uri=https://github.com/cocktailpeanut/deeperhermes

Since the LLM inference is powered by LMStudio, it works on ALL machines and OS. 

also posted on reddit https://www.reddit.com/r/LocalLLaMA/comments/1ipj9ux/i_took_nous_deephermes_and_made_it_autodecide_how/

Tips on Generating Timestamped Lyrics for DiffRhythm, for FREE.

The DiffRhythm Gradio app has an "LLM Generate LRC" tab that lets you paste in your own lyrics and get a timestamped lyrics text back. This requires OpenAI API key.

But you really don't need one. Let me share how.

Just go to any LLM service like ChatGPT, Gemini, Claude, etc. and enter:

[YOUR LYRICS GOES HERE]

This is the lyrics of a song. Each line is a line of lyrics. I now want to timestamp each line of lyrics of this song to get LRC. I hope that the timestamp allocation should be reasonably inferred based on the song tag, the emotion of the lyrics, and the rhythm, rather than mechanically allocated according to the length of the lyrics. The timestamp of the first line of lyrics should take into account the length of the prelude to avoid the lyrics starting directly from [00:00.00]. Output the lyrics strictly in LRC format, with each line in the format of [mm:ss.xx] lyrics content. The final result only outputs LRC, no other explanation is required.

And it will give you the timestamped lyrics.

This system prompt is EXACTLY what the gradio app uses behind the scenes, and you don't need an API key for that.

Once you get the timestamped lyrics, paste it into the lyrics field in the "Music Generate" tab, and generate. 