Files
yvgude--lean-ctx/docs/contracts/personal-cloud-encryption-v1.md
T
wehub-resource-sync 26382a7ac6
CI / Clippy (push) Failing after 15m13s
CI / Test (ubuntu-latest) (push) Failing after 16m1s
CI / Test (macos-latest) (push) Has been cancelled
CI / Test (windows-latest) (push) Has been cancelled
CI / Build (no embeddings / no ORT) (push) Has been cancelled
CI / Format (push) Has been cancelled
CI / Cookbook (Node) (push) Has been cancelled
CI / Pi Extension (Node) (push) Has been cancelled
CI / Rust SDK (lean-ctx-client) (push) Has been cancelled
CI / Embed SDK (lean-ctx-sdk) (push) Has been cancelled
CI / Python SDK (leanctx) (push) Has been cancelled
CI / Hermes Plugin (Python) (push) Has been cancelled
CI / SDK Conformance Matrix (push) Has been cancelled
CI / Coverage (push) Has been cancelled
CI / cargo-deny (push) Has been cancelled
CI / Adversarial Safety (push) Has been cancelled
CI / Benchmarks (push) Has been cancelled
CI / Output-Quality Gate (eval A/B) (push) Has been cancelled
CI / Documentation (push) Has been cancelled
CI / CI Green (push) Has been cancelled
JetBrains Plugin / Actionlint (push) Has been cancelled
CodeQL / Analyze (actions) (push) Has been cancelled
CodeQL / Analyze (javascript-typescript) (push) Has been cancelled
CodeQL / Analyze (rust) (push) Has been cancelled
JetBrains Plugin / Validation (push) Has been cancelled
JetBrains Plugin / Build (push) Has been cancelled
JetBrains Plugin / Test (push) Has been cancelled
Security Check / Security Scan (push) Has been cancelled
chore: import upstream snapshot with attribution
2026-07-13 12:35:30 +08:00

62 lines
3.1 KiB
Markdown

# personal-cloud-encryption-v1 — Zero-Knowledge Vaults (Knowledge + Gotchas)
Status: **active** (GL #467) · Engine: `core/knowledge_vault.rs` ·
Server: `cloud_server/knowledge.rs` (`knowledge_blobs`),
`cloud_server/gotchas.rs` (`gotcha_blobs`)
## Claim
For E2E surfaces, the Personal Cloud backend stores **only ciphertext**. We
cannot read, search, sell or leak knowledge content — provably: the
decryption key is derived from the account API key, of which the server only
ever stores a SHA-256 hash.
## Construction
| Property | Value |
|---|---|
| Cipher | XChaCha20-Poly1305 (AEAD), 24-byte random nonce per seal |
| Key derivation | HKDF-SHA256(salt=`leanctx`, ikm=API key, info=`knowledge-vault-v1` \| `gotcha-vault-v1`) |
| Domain separation | distinct HKDF `info` per surface — the index-bundle key (`index-bundle-v1`), the knowledge-vault key and the gotcha-vault key can never open each other's blobs |
| Envelope | `{"v":1,"entries":[{category,key,value},…]}`, serialized then sealed; wire format `nonce ‖ ciphertext` |
| Consistency | whole-account snapshot, last-writer-wins (same model as `hosted-personal-index-v1`) |
The key is identical on every logged-in device (stable API key, not the
rotating OAuth token) — that is what makes cross-device pull work. Key
rotation = new API key + one re-push from any device that has the local store.
## Wire protocol (`/api/sync/knowledge`, `/api/sync/gotchas`)
Both routes speak the same dual wire format; each has its own blob table
(`knowledge_blobs` / `gotcha_blobs`) and purges its own legacy table
(`knowledge_entries` / `gotchas`).
| Request | Behaviour |
|---|---|
| `POST` `Content-Type: application/octet-stream` + `X-Entry-Count: N` | store vault blob (≤ 8 MB), then **delete the account's plaintext rows** — the built-in re-encryption migration |
| `POST` `Content-Type: application/json` | legacy plaintext upserts (deprecated; removed two releases after vault clients ship) |
| `GET` `Accept: application/octet-stream` | encrypted vault blob; `404` when the account has none yet |
| `GET` (anything else) | legacy plaintext listing |
`X-Entry-Count` is a client-declared display metadatum (dashboards show
counts and sizes); the server cannot verify it — by design.
Clients pull vault-first and fall back to the legacy listing on `404` *or*
when an older server ignores the `Accept` header (detected via the response
`Content-Type`).
## What is deliberately NOT E2E
| Surface | Why it stays aggregate/plaintext |
|---|---|
| Commands / CEP / Gain stats | numbers only (counts, token totals) — no content; they feed the savings dashboard and the opt-in leaderboard, which require server-side aggregation |
| Supporter / billing metadata | Stripe-owned, never includes code or knowledge |
| Index bundles | already E2E under `hosted-personal-index-v1` |
## Server obligations
- Zero-content logging: sizes, hashes, counts — never payloads.
- `knowledge_blobs` carries `sha256` over the ciphertext for drift detection.
- The legacy table stays queryable until the deprecation window closes, but
any vault push purges that account's plaintext rows immediately.