AstrBot-Desktop/docs/reports/p02-dual-slots-handoff.md
YokiiroBW 34a563c65b feat(connection): validate auto-connect and guard operations switches
Add configuration-only auto-connect rules, runtime operations slot lock
while tasks are active, and duplicate server-instance warnings without
silent Profile merge. Surface policy status to the settings UI and cover
the new boundaries with Rust and Vitest checks.
2026-07-20 00:49:15 +08:00

80 行
4 KiB
Markdown

# P02 dual-slot checkpoint and handoff
Date: 2026-07-19
## Source state
- Branch: `feat/p02-dual-connections`
- Base merge: `885abd1553ed8f1428de7118a6f4b59f7e22a45c`
- Planning commit: `73693b9 docs(p02): start dual connection phase`
- Feature commit: `bcc08cf feat(connection): add persistent dual slot assignment`
- Remote transport: approved ordinary Forgejo HTTP push
- Force push, rebase, and automatic main merge: not used
## Completed scope
- SQLite schema v3 creates exactly two fixed rows: `companion` and `operations`.
- A foreign key clears slot assignment when a Profile is deleted.
- A unique Profile constraint prevents one Profile from occupying both slots.
- Rust exposes bounded list, assign, and clear commands.
- Slot requests reject unknown fields, malformed Profile IDs, missing Profiles, and duplicate cross-slot assignment.
- The frontend validates Rust payloads with strict Zod Schemas.
- The settings UI assigns or clears Profiles without activating REST/SSE or reading credentials.
- Stored `server_instance_id` is displayed as metadata only; the increment does not enroll or silently replace server identity.
## Security boundary
- No Token, password, API Key, Cookie, Authorization header, or NAS credential field was added to slot IPC.
- Slot assignment alone grants no local-tool permission.
- React does not receive stored credentials.
- Real Gateway compatibility, token refresh, identity enrollment, and certificate pinning remain explicitly unresolved in PD-0009 through PD-0012.
## Verification
- `npm.cmd --prefix H:\AstrBot-Desktop run check:frontend`: passed.
- Prettier: passed.
- ESLint: passed.
- TypeScript: passed.
- Vitest: 4 passed.
- Vite production build: passed.
- `npm.cmd --prefix H:\AstrBot-Desktop run test:e2e`: passed, Chromium 1/1.
- `cargo fmt --all --manifest-path H:\AstrBot-Desktop\Cargo.toml -- --check`: passed.
- `cargo clippy --workspace --all-targets --manifest-path H:\AstrBot-Desktop\Cargo.toml -- -D warnings`: passed.
- `cargo check --workspace --all-targets --manifest-path H:\AstrBot-Desktop\Cargo.toml`: passed.
- `cargo test --workspace --manifest-path H:\AstrBot-Desktop\Cargo.toml`: passed, 56 normal tests and 2 explicit manual tests ignored.
- Changed-file high-risk secret pattern scan: passed with no matches.
Not rerun at this incremental checkpoint:
- real Tauri development startup;
- Windows release or NSIS build;
- Forgejo Actions;
- real AstrBot Gateway compatibility;
- real UNC file transaction.
No success is claimed for the items above.
## Remaining P02 work
1. Add auto-connect validation for at most two Profiles and one Profile per slot.
2. Block Operations Profile switching while an operations task is active.
3. Add duplicate `server_instance_id` warnings without silent identity merging.
4. Resolve the credential enrollment and rotation lifecycle.
5. Resolve first-contact server identity confirmation.
6. Implement validated certificate fingerprint binding without disabling TLS verification.
7. Only then enable two independent configured REST/SSE runtimes and Mock-backed failure tests.
## Handoff status
The first controller status call failed before project lookup because the global dev-handoff config references missing files:
- identity file: `C:\Users\Administrator\Documents\Codex\2026-07-14\ke\work\ssh\astrbot_vm_ed25519`
- known-hosts file: `C:\Users\Administrator\Documents\Codex\2026-07-14\ke\work\ssh\known_hosts`
The controller also reported host-key verification failure. No handoff ID, VM workspace, branch, or HAPI session has been created or claimed.
After the source branch was clean and pushed, the formal `dev-handoff send --approve` call was also attempted with the verified test and secret-scan results. It failed before task creation with the same missing identity and host-key verification error.
## Next concrete step
Restore or reconfigure the authorized VM SSH identity and known-hosts paths, rerun `dev-handoff status`, then send this exact pushed checkpoint to `vm-codex`. If the controller stops at `ready_to_accept`, report that state honestly and wait for the VM worker.