Migrating from Pinokio

Both Pinokio and Thumper-Run let you run AI applications locally. They take different approaches: Pinokio uses script-based automation, while Thumper uses declarative manifests with built-in security, encryption, and cross-device sync. This guide covers what carries over, what doesn’t, and how to make the switch.

Overview

Pinokio is a script-based local AI app launcher. You write or download JSON scripts that automate git clone, pip install, and launch sequences. It’s flexible and community-driven.

Thumper uses .thumper.yaml manifest files that declare an app’s dependencies, models, GPU requirements, and launch configuration. The runtime handles installation, process management, GPU detection, and model downloads automatically.

The key differences are in the layers above the app itself: Thumper adds CRDT-based state management, device pairing, a curated app catalog with security reviews, and a planned private end-to-end encryption model.

What Transfers

Models (Automatic)

If you have models in your HuggingFace cache (~/.cache/huggingface/hub/), Thumper discovers them automatically. They appear in the Local Cache tab on the Models page — no re-download needed. This covers most Stable Diffusion checkpoints, LoRAs, VAEs, and LLM weights downloaded through Pinokio apps.

Ollama Models (Automatic)

Ollama models are stored in Ollama’s own cache directory and persist independently. If you had Ollama installed for Pinokio apps, those same models are available to Thumper immediately.

Pinokio Scripts (Do Not Transfer)

Pinokio’s JSON scripts are not compatible with Thumper manifests. The two systems use different schemas, different installation strategies, and different process management. You’ll install apps fresh from the Thumper catalog.

App Data and Settings

App-specific data (chat history, generated images, custom workflows) depends on the app. Some apps store data in standard locations that carry over; others use app-specific directories that differ between Pinokio and Thumper installs.

ComfyUI workflows, Open WebUI chat exports, and SillyTavern character cards are portable files you can import into fresh Thumper installations.

Feature Comparison

FeaturePinokioThumper-Run
App catalogCommunity scripts (open)Curated catalog (reviewed manifests)
Install methodJSON scripts (imperative).thumper.yaml manifests (declarative)
GPU auto-configurationScript-dependentAutomatic detection + platform patches
Security / encryptionNoneServer-assisted CRDT sync, device pairing, planned private E2EE
Cross-device syncNot supportedCRDT-based server-assisted sync
Model managementManual per-appShared model store, auto-discovery, symlinks
Offline modeApps work offline after installFull offline mode with CRDT state persistence
App updatesRe-run install scriptManaged updates with rollback
Custom / community appsWrite a JSON scriptWrite a .thumper.yaml manifest

Migration Steps

For ComfyUI Users

  1. Install ComfyUI from the Thumper catalog — models in your HuggingFace cache are reused automatically
  2. Export your workflows as .json files from Pinokio’s ComfyUI (Save button in the UI)
  3. Import them into the Thumper ComfyUI instance (drag-and-drop or Load button)
  4. Missing models are resolved automatically via Thumper’s 5-step model resolution

For Open WebUI Users

  1. In Pinokio’s Open WebUI, go to Settings → Export and download your chat history
  2. Install Open WebUI from the Thumper catalog — Ollama models carry over automatically
  3. Import your chat history via Settings → Import in the Thumper Open WebUI instance

General Migration

  1. Install Thumper-Run (desktop app or web shell)
  2. Browse the catalog and install your apps — models are auto-discovered from the HuggingFace cache
  3. Export any app-specific data (chats, workflows, characters) from Pinokio before uninstalling
  4. Optionally remove Pinokio’s app directories once you’ve verified everything works in Thumper

FAQ

Can I run both Pinokio and Thumper side by side?

Yes. They use separate app directories and separate process managers. The only shared resources are the HuggingFace cache and Ollama — both of which handle concurrent access safely.

Will I lose my models?

No. Models in the HuggingFace cache (~/.cache/huggingface/hub/) are not touched by either platform. Ollama models persist in Ollama’s own directory. Thumper discovers both caches automatically.

Can I use my existing Pinokio scripts in Thumper?

Not directly. Pinokio scripts and Thumper manifests are different formats with different capabilities. However, converting a Pinokio script to a .thumper.yaml is straightforward for most apps — the App Developer Guide covers the manifest format in detail.

What if an app I use is not in the Thumper catalog?

You can write a .thumper.yaml manifest for any app and submit it to the catalog. See the App Developer Guide for the manifest specification and contribution process.

Key Takeaways

Your models are safe. HuggingFace cache and Ollama models persist independently of both Pinokio and Thumper. You can migrate at your own pace without re-downloading anything.