L00P.AI / RASPBERRY PI 500+
An AI workspace.
On your Pi.
We are preparing a free installer and configurator for Raspberry Pi 500+. It will help you choose agents, models, archive access and sign-in methods. Every component should be understandable before installation.
First, a free installer on official Raspberry Pi OS; later, a L00P.AI system variant that opens the same configurator after first login. We own a US-layout Pi 500+, but the installer and test results are not ready; European layouts will be checked before release.
ONE COMPUTER, TWO PATHS
Install knowingly.
Configure your way.
QUICK · SELF-SERVICE
Hermes Agent + Codex CLI
A recommended starting point. The installer will check the Pi, show every system change, guide vendor sign-in and separate the archive from outputs. The setup still needs device acceptance.
ADVANCED · WITH US OR ON YOUR OWN
Hermes / OpenClaw + Codex / Claude Code
Choose an open agent and a coding agent, models, permissions, costs and access boundaries. Guided advanced setup and organisational adaptation can be paid L00P.AI services.
ChatGPT and Claude passwords do not pass through L00P.AI: vendor apps handle sign-in. API keys will be optional after the local secret store is accepted. The installer itself will be free. Codex · Claude Code.
WHAT BRINGS IT TOGETHER
A folder of recordings becomes a starting point.
In the planned first-run workflow, you will choose an archive directory. Originals will remain read-only; transcripts, indexes and exports will go to a separate folder. Words will retain their recording timestamps, so important answers can be checked by listening.
Recording
The original and its provenance.
Words and time
Karaoke transcription and metadata.
Your workspace
Listening, retrieval and agent workflows.
Explore the existing interface demonstration. It shows how Szpieg+ works; readiness of the Raspberry Pi port will be assessed separately. Open the demo →
Choose where the model runs.
ON YOUR COMPUTER
Local audio and small models
Plan: Szpieg+, whisper.cpp / sherpa-onnx, Ollama or llama.cpp. After model download, this path is intended to work offline. Acceptance includes a disconnected test. Agents will receive access only to selected tools and directories.
WITH YOUR OWN ACCOUNT
Claude Code · ChatGPT / Codex
A client application does not include Claude or GPT model weights. Cloud services require Internet access, your own account and a suitable plan or API billing. Sending material to the cloud must be the user’s deliberate choice.
Hermes Agent and OpenClaw connect tools to a chosen model provider. Running an agent on a Pi does not mean a large model runs locally. OpenClaw’s Raspberry Pi guidance recommends cloud models. OpenClaw / Raspberry Pi.
ARM64: what is documented and what still needs testing?
Claude Code, Codex CLI, Hermes and Ollama have Linux ARM64 installation paths. ChatGPT now offers a Linux ARM64 preview, and LM Studio documents Linux ARM64 requirements. Raspberry Pi OS compatibility still needs to be checked on an actual Pi 500+. Browser access is also planned for ChatGPT.
Claude Code · Codex CLI · ChatGPT Linux · Hermes · Ollama · LM Studio
AUTHORS AND TERMS
A free installer.
Visible tool provenance.
Our contribution is integration, configuration and radio production workflow. Each tool’s authors retain their rights. Below are installer candidates, sources and terms. This is neither an installation receipt nor proof of operation on the Pi.
| Component | Licence / terms | Installation path |
|---|---|---|
| Raspberry Pi OS Desktop 64-bit | Per-package; firmware separatelyLicence · Source | We check the user’s official installation. |
| FFmpeg | LGPL/GPL depending on build and librariesLicence · Source | Installer after version review and Pi tests. |
| whisper.cpp | MIT core; dependencies separatelyLicence · Source | Installer after version review and Pi tests. |
| llama.cpp | MIT core; dependencies separatelyLicence · Source | Installer after version review and Pi tests. |
| sherpa-onnx | Apache-2.0 core; ONNX Runtime and other dependencies separatelyLicence · Source | Installer after version review and Pi tests. |
| Codex CLI | Apache-2.0 source; service access separateLicence · Source | Installer after version review and Pi tests. |
| Ollama | MIT core; model licenses separateLicence · Source | An option selected by the user. |
| Hermes Agent | MIT core; skills and dependencies separatelyLicence · Source | An option selected by the user. |
| OpenClaw | MIT core and third-party noticesLicence · Source | An option selected by the user. |
| Claude Code | Proprietary Anthropic terms; no confirmed L00P redistribution rightLicence · Source | Install and sign in directly with the vendor. |
| ChatGPT browser / Linux preview | Proprietary OpenAI terms; browser access preferredLicence · Source | Install and sign in directly with the vendor. |
| LM Studio | Proprietary; app terms restrict distributionLicence · Source | Install and sign in directly with the vendor. |
| OpenAI Whisper model weights | MIT code and weightsLicence · Source | An option selected by the user. |
| Qwen3-ASR 0.6B weights | Apache-2.0 upstream modelLicence · Source | An option selected by the user. |
| Parakeet TDT 0.6B v3 weights | CC-BY-4.0 upstream modelLicence · Source | An option selected by the user. |
| Qwen3 0.6B text model | Apache-2.0 upstream model; quantized artifact separatelyLicence · Source | An option selected by the user. |
| Qwen3 1.7B text model | Apache-2.0 upstream model; quantized artifact separatelyLicence · Source | An option selected by the user. |
| Szpieg+ free edition | Free-edition terms and component rights pendingLicence · Source | Unavailable until rights and Pi port are accepted. |
In the first channel, the user installs official Raspberry Pi OS. The configurator checks its version and shows the provenance of additions. A future system variant with a L00P.AI splash screen will require a separate rights review for every component in the image. Raspberry Pi OS.
LM Studio permits personal and internal business use, but its current terms restrict redistribution of the app. We therefore plan an official vendor installation link. Claude Code and the ChatGPT app are also treated as separate vendor installations. Szpieg+’s stdlib approach reduces code dependencies; it is not itself a licence.
MEASUREMENT BEFORE A PROMISE
Local transcription.
Accuracy to be measured.
The Pi 500+ has 16 GB RAM, four 2.4 GHz Cortex-A76 cores and a 256 GB NVMe SSD as standard. That makes local ASR testing reasonable, but does not establish real-time speed or a particular error rate. US and EU describe keyboard-layout choices, not processor speed. Raspberry Pi 500+.
Whisper / whisper.cpp
Candidates: small, medium and large-v3-turbo. The runtime supports Raspberry Pi and quantisation. We will measure how much time improved accuracy costs on Polish recordings.
whisper.cpp →Parakeet TDT 0.6B v3
Candidate: an int8 export through sherpa-onnx. NVIDIA reports Polish WER of 7.31% on FLEURS and 7.28% on MLS. These are vendor results, not measurements on a Pi or our archive.
Model card →Qwen3-ASR 0.6B
Supports Polish; sherpa-onnx documents an ONNX export. It will be a third candidate for comparing accuracy and speed.
Model card →Initial text-model candidates: Qwen3 0.6B and 1.7B through llama.cpp / Ollama. We will choose quantization and the exact file after checking provenance, starting with short context and one task at a time. These are test candidates, not a performance guarantee or a claim of cloud-model quality. 0.6B · 1.7B.
How will we evaluate the below-4% WER target?
We will prepare at least two hours of human-verified reference material: studio, telephone, noise, overlapping speech, names and numbers. Test data will be kept separate from tuning data. We will record normalisation rules and raw results. Scribe will be another system to compare, not the ground truth.
WER = (S + D + I) / N — substitutions, deletions and insertions divided by reference words. We will also report name and number errors, results for each recording category, runtime, RAM and temperature. RTF below 1 means processing faster than the recording duration; it is a separate measure from WER.
We do not yet have a Pi 500+ measurement. Below 4% remains a target for a defined dataset, not a promise for any archive.
What must the first release include?
The first release will start from official Raspberry Pi OS installed by the user. We will check its version, free space and existing data. A later L00P.AI system variant will be built from a clean base and the same installer recipe, not cloned from a used disk. Raspberry Pi OS.
- Acceptance on a physical US-layout Pi 500+ and European keyboard setup tests.
- Exact versions, official sources, checksums, licences and the scope of system changes.
- Two paths: quick Hermes + Codex and advanced choice of Hermes/OpenClaw and Codex/Claude Code.
- Vendor sign-in without collecting passwords or copying tokens between devices.
- Read-only archive, transcription sample, playback from a word and separate outputs.
- Offline test of selected models, published ASR results and recovery after interruption.
The L00P.AI installer is intended to be free. Guided advanced setup and organisational adaptation may be paid services. Hardware and service accounts are separate. This is an independent integration without a claim of formal partnership with tool vendors.