Skip to content

Pocket Node 0.1.0-rc2

Pre-release
Pre-release

Choose a tag to compare

View all tags
@GsxrGuy80s GsxrGuy80s released this 10 Jun 14:29
· 4 commits to main since this release
v0.1.0-rc2
1022444

Pocket Node 0.1.0-rc2

Release Candidate 2 — re-tagged from RC1 to resolve a release naming collision. Build is functionally identical to the RC1 build; only versionName (0.1.0-rc2) and versionCode (4) have changed.

Not a stable release. Core inference, Edge API, and local license validation are functional. Feedback welcome.

What This Is

Pocket Node runs large language models entirely on your Android device — no cloud, no subscription, no data leaving your phone. This RC2 build includes:

  • On-device LLM inference via llama.cpp (NDK/JNI, arm64-v8a)
  • Vulkan and OpenCL GPU acceleration where supported
  • Local license validation path included; public Pro key issuance is not yet open in RC2.
  • Edge API on port 11434 — OpenAI-compatible /v1/chat/completions endpoint for use on your local network
  • Model Hub with curated + community GGUF import
  • First-run device profiling (thread count, backend recommendation)

Tested Device

Device Result
Samsung Galaxy Z Fold 6 (SM-F956B) ✅ Functional

RC2 has been smoke-tested on the Z Fold 6 only. Other devices should work but are untested. Bug reports with device model and Android version are very helpful.

Install Instructions

This APK is sideloaded — it is not on the Play Store yet.

  1. Download PocketNode-0.1.0-rc2-signed.apk
  2. On your Android device: Settings → Apps → Special app access → Install unknown apps — enable for your browser or file manager
  3. Open the APK and tap Install
  4. On first launch, Pocket Node will scan for existing GGUF models or prompt you to download one

Minimum requirement: Android 9.0 (API 28) or higher, arm64-v8a device

Model Download

The built-in Operator model (recommended for first run) downloads on demand — it is not bundled in the APK.

  • Operator model size: ~1.68 GB — download over Wi-Fi recommended
  • Community GGUFs can be imported manually from local storage
  • Recommended Models list includes smaller options (135M–3B) for low-storage devices

Edge API

When Edge API is enabled in Settings, the device exposes:

POST http://<device-ip>:11434/v1/chat/completions   (OpenAI-compatible)
POST http://<device-ip>:11434/api/generate          (raw)
GET  http://<device-ip>:11434/                      (status)

Compatible with Continue.dev, Open WebUI, and any client that targets an OpenAI-compatible base URL.

Known Limitations (RC2)

  • No OTA updater — auto-update is disabled for RC2. Install the next release manually.
  • Tested on one device (Z Fold 6). Multi-device compatibility is unverified.
  • Vision models and RAG are not yet implemented.
  • Edge API has no authentication (LAN use only — do not expose to the internet).
  • minSdk is API 28 (Android 9.0). Older devices are not supported.
  • Public Pro key issuance is not open in RC2. Payment/keygen flow is still being finalized.

SHA-256 Verification

f1fe2887dd9f7ab0f9bd62021857bae0efdcb98c090b489a626b960154964126  PocketNode-0.1.0-rc2-signed.apk

Verify on Linux/macOS: sha256sum PocketNode-0.1.0-rc2-signed.apk
Verify on Windows: Get-FileHash PocketNode-0.1.0-rc2-signed.apk -Algorithm SHA256

Safety Note

This is a release candidate. The APK is signed with the Pocket Node release key (v2 scheme, verified).

  • Do not expose the Edge API to the internet. It has no authentication in RC2.
Assets 4
Loading