Repository navigation
Architecture
Febin Francis edited this page Oct 13, 2026
·
1 revision
| Part | Path | Role |
|---|---|---|
| Android app | app/ |
Kotlin/Compose UI, wallet coordinator (MWA), ORE v3 instruction builder, operation ledger, Keystore-signed receipts |
| On-device LLM bridge | ondevice-llm/ |
Java-only module wrapping LiteRT-LM 0.8.0 so it never reaches the app's Kotlin compile classpath |
| Receipt checker | tools/receipt-checker/ |
Standalone JVM tool; verifies the integrity of signed receipt contents |
| ORE fork simulation |
tools/ore-fork-sim/, tools/ore-fixtures/
|
Mainnet-fork simulation and fixtures for the ORE Deploy instruction |
| Offline kit | tools/edgeore-kit/ |
Python tooling and tests |
| Storage gateway | services/storage-gateway/ |
Optional self-hosted gateway you run yourself |
Trust spine: exact-message review → MWA sign → returned-bytes check → one separate submit → Keystore-signed receipt. An unknown outcome is re-read from chain after restart and never resent.
The ORE v3 Deploy instruction (ix 6, 12 accounts) was checked against regolith-labs/ore at commit 48c203bd.
Built by CodesbyFebin · Devnet review candidate · Trust boundaries · Qualification
EdgeORE
- The EdgeORE handbook
- Read the evidence ledger
- Clone and pin the Android source
- Prepare the Android build environment
- Run and preserve qualification gates
- Install and identify the candidate APK
- Choose a runtime without tool shopping
- Authorize through Mobile Wallet Adapter
- Use the official mock wallet safely
- Test disconnect and reconnect after restart
- Read balance and obtain devnet funding
- Review the exact transaction message
- Submit once and observe confirmation
- Recover after an interrupted submission
- Understand budgets and spend reservations
- Read a receipt without overclaiming it
- Verify an export on the JVM
- Change signed lamports and observe rejection
- Separate integrity, provenance and completeness
- Use an Ollama-compatible owned host
- Understand endpoint and transport policy
- Evaluate the experimental on-device AI branch
- Evaluate AI quality and privacy claims
- Use the device-local encrypted vault
- Design an owned-node IPFS adapter
- Design backup and restore acceptance
- Pair an owned node with a pinned identity
- Observe health and revoke node access
- Interpret controls and device readings
- Threat model and production boundaries
- Collect diagnostics without leaking secrets
- Design customer-funded SOL contribution jobs
- Specify escrow before writing a program
- Settlement program status
- Use attestation for the claim it can support
- Evaluate federated learning and data contribution
- Bound mesh networking and agent hosting
- Write marketing copy that survives inspection
- Build the website and publish the GitBook handbook
- Contribute without merging every upstream repository