@mmmbuto/nexuscrew 0.8.44 → 0.8.46

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/CHANGELOG.md CHANGED
@@ -4,6 +4,57 @@ All notable changes to NexusCrew are tracked here.
4
4
 
5
5
  ## Unreleased
6
6
 
7
+ ## 0.8.46 — 2026-07-31 — "Native Kimi"
8
+
9
+ - The Fleet gains a native Kimi Code CLI engine (`kimi.native`) that launches
10
+ the official `@moonshot-ai/kimi-code` binary directly in the cell's working
11
+ directory. Authentication and providers stay with the CLI (device-code
12
+ login, `config.toml`); NexusCrew never reads, stores or injects Kimi
13
+ credentials, and existing installations receive the engine through an
14
+ idempotent, collision-safe backfill.
15
+ - The engine defaults to interactive (standard) permissions. The unsafe
16
+ policy maps to `--yolo`, which auto-approves regular tool calls while the
17
+ agent may still ask questions; the fully autonomous `--auto` mode is
18
+ deliberately not exposed.
19
+ - Because the CLI documents no interactive prompt flag (`kimi -p` is
20
+ non-interactive and skips the TUI), the cell prompt is never placed on the
21
+ command line: it is delivered with bracketed paste once the session is
22
+ ready, including on supervised restarts.
23
+ - The Claude Code "Kimi Code" provider (`claude.kimi-code`) is unchanged and
24
+ remains the managed K3 path with the 1M-context profile. The PWA presents
25
+ the two paths under distinct names ("Kimi Code CLI" vs "Kimi Code") with an
26
+ explanatory notice, and login-delegated engines no longer show a misleading
27
+ credential KEY section.
28
+
29
+ ## 0.8.45 — 2026-07-30 — "Resilient Directory"
30
+
31
+ - Fleet cell-directory reads no longer wait for external model catalog discovery.
32
+ A stalled optional engine binary therefore cannot make the MCP bridge present
33
+ a reachable local Fleet as unavailable.
34
+ - External model discovery has a bounded five-second budget and caches failures
35
+ as well as successful results, preventing repeated hung subprocess calls.
36
+ - The MCP cell directory now marks a local timeout explicitly, so a degraded
37
+ bridge can be distinguished from an unreachable remote node.
38
+ - Share activation now performs an authenticated, non-scanning preflight on
39
+ the hub-assigned reverse port. A listener owned by the same verified peer is
40
+ accepted; an unverified occupied port returns an actionable conflict without
41
+ restarting or publishing the private tunnel.
42
+ - Peer health reports a reverse listener that remains active while Share is
43
+ disabled, while `nexuscrew status` now states Share enabled or disabled for
44
+ inbound peers. These diagnostics never terminate a peer tunnel automatically.
45
+ - Shared peers can use an operator-preauthorized three-port reverse pool. The
46
+ product never writes `authorized_keys`; legacy or incompletely granted peers
47
+ remain private/usable and are diagnosed as not rotatable instead of guessing
48
+ another SSH port.
49
+ - A peer proposes a pool slot over its existing private channel and the hub
50
+ assigns a bounded lease and generation. Rotation runs only after all slots
51
+ have been independently verified, tries one candidate per ten-minute window,
52
+ and quarantines failures rather than killing unknown listeners.
53
+ - Slot ownership now uses a per-slot HMAC proof with a fresh challenge. The
54
+ hub never sends a peer bearer credential to a listener it is investigating;
55
+ a relayed proof from a different slot is rejected. Retired pool bases are
56
+ never reused automatically, protecting against stale `permitlisten` grants.
57
+
7
58
  ## 0.8.44 — 2026-07-28 — "Deliberate Switch"
8
59
 
9
60
  - The mobile cell quick rail now separates selection from switching: choose a
package/README.md CHANGED
@@ -66,7 +66,7 @@ install-script approval.
66
66
  | **Live terminals** | Attach to real tmux sessions through a real PTY, WebSocket and xterm.js. |
67
67
  | **Persistent workspaces** | Arrange sessions into decks with saved layouts, ordering, pins and per-cell drafts. |
68
68
  | **Multi-node Fleet** | See and control authorized cells across Linux, macOS and Android nodes. |
69
- | **AI-ready cells** | Launch Claude Code, Codex, Codex-VL, Pi, Agy or a trusted shell with explicit providers and policies. |
69
+ | **AI-ready cells** | Launch Claude Code, Codex, Codex-VL, Pi, Agy, Kimi Code CLI or a trusted shell with explicit providers and policies. |
70
70
  | **Mobile-native control** | Scroll tmux history, use terminal keys, dictate prompts and move files from a phone. |
71
71
  | **Operator alerts** | Receive visual, push, browser speech and opt-in node-native Audio Share TTS. |
72
72
 
@@ -101,6 +101,12 @@ Remote access is intentionally carried through SSH or a VPN you control:
101
101
  ssh -L 41820:127.0.0.1:41820 user@your-host
102
102
  ```
103
103
 
104
+ Shared mobile peers can use a small, pre-authorized reverse-port pool for
105
+ recovery from a real reverse-forward collision. NexusCrew never edits SSH
106
+ policy: the hub operator keeps `authorized_keys` authoritative, while the
107
+ product verifies a configured pool before it can rotate within it. See
108
+ [Connect nodes](docs/NODES.md#rotatable-reverse-port-pools).
109
+
104
110
  [Security model →](docs/SECURITY.md)
105
111
 
106
112
  ## Documentation