breakglass: in-cluster emergency-recovery UI for the devvm
All checks were successful
ci/woodpecker/push/woodpecker Pipeline was successful

Viktor wanted a web UI on the claude service to act as his breakglass when
the devvm is down: open it, have Claude SSH in to diagnose/repair, and
power-cycle the VM via the Proxmox host if needed. This is the app half
(the infra stack + host bootstrap live in the infra repo).

New, ISOLATED ASGI app under app/breakglass/ (never imports app.main, so the
untrusted-input agents — recruiter-triage, nextcloud-todos — can't share a
process with the root-on-devvm / PVE-reset SSH key):
- pve.py: the LLM-independent power-verb path (status|forensics|reset|stop|
  start|cycle on VM 102), whitelist-validated client-side, executed over the
  forced-command SSH key (list argv, no shell).
- agent_session.py: multi-turn streamed chat — claude -p --session-id /
  --resume with --output-format stream-json, translated to a small SSE
  vocabulary (session/text/tool/result/error/done).
- auth.py: edge Authentik header OR bearer; fail-closed.
- server.py: FastAPI (session/chat-SSE/pve-verb routes) + serves the Svelte UI.
- Svelte SPA (frontend/, built into app/breakglass/static/ and committed — no
  in-cluster build, per ADR-0002): streamed chat + danger-styled manual VM
  controls with confirm-on-mutate.
- agents/breakglass.md: narrow tools (Bash/Read/Grep/Glob, no web), taught the
  ssh devvm / ssh pve aliases and cycle-vs-reset.
- docker-entrypoint-breakglass.sh: ssh-agent bootstrap from the mounted key +
  ssh aliases, then uvicorn app.breakglass.server. The breakglass Deployment
  overrides the image CMD with this; the existing service is untouched.

26 new tests (verb whitelist incl. injection attempts, stream-json→SSE
translation, auth gating, route behaviour); full suite 58 green.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
Viktor Barzin 2026-06-12 21:36:05 +00:00
parent 694530135d
commit 4f361d91eb
28 changed files with 3889 additions and 0 deletions

92
frontend/src/lib/api.js Normal file
View file

@ -0,0 +1,92 @@
// Same-origin API client. Auth is handled entirely by the edge proxy
// (Authentik / basic-auth / bearer) — this UI never sends or stores a token.
import { readEventStream } from './sse.js';
/** Open a fresh chat session. @returns {Promise<string>} session_id */
export async function openSession() {
const res = await fetch('/api/session', {
method: 'POST',
headers: { 'content-type': 'application/json' },
});
if (!res.ok) {
throw new Error(`could not open a session (HTTP ${res.status})`);
}
const body = await res.json();
if (!body || typeof body.session_id !== 'string') {
throw new Error('session response missing session_id');
}
return body.session_id;
}
/**
* Run one chat turn. Streams events to onEvent until the backend sends
* {kind:"done"} and the connection closes. Pass an AbortSignal to cancel.
*
* @param {{session_id: string, prompt: string, model?: string, signal?: AbortSignal}} opts
* @param {(event: object) => void} onEvent
*/
export async function streamChat({ session_id, prompt, model, signal }, onEvent) {
const payload = { session_id, prompt };
if (model) payload.model = model;
const res = await fetch('/api/chat', {
method: 'POST',
headers: {
'content-type': 'application/json',
accept: 'text/event-stream',
},
body: JSON.stringify(payload),
signal,
});
await readEventStream(res, onEvent);
}
/**
* List the PVE power verbs and which of them mutate VM state.
* @returns {Promise<{verbs: string[], mutating: string[]}>}
*/
export async function fetchVerbs() {
const res = await fetch('/api/pve/verbs');
if (!res.ok) {
throw new Error(`could not load VM controls (HTTP ${res.status})`);
}
const body = await res.json();
return {
verbs: Array.isArray(body.verbs) ? body.verbs : [],
mutating: Array.isArray(body.mutating) ? body.mutating : [],
};
}
/**
* Run a PVE power verb directly (no AI in the path). The backend returns 200
* on success and 502 when the verb's exit code is non-zero, but the JSON body
* carries {verb, exit_code, stdout, stderr, rejected} in BOTH cases so we
* read the body regardless of HTTP status and let the caller style on
* exit_code / rejected.
*
* @param {string} verb
* @returns {Promise<{verb: string, exit_code: number|null, stdout: string, stderr: string, rejected: boolean}>}
*/
export async function runVerb(verb) {
const res = await fetch(`/api/pve/${encodeURIComponent(verb)}`, {
method: 'POST',
headers: { 'content-type': 'application/json' },
});
// 400 = unknown verb (FastAPI HTTPException) — has {detail}, not the verb shape.
let body;
try {
body = await res.json();
} catch {
throw new Error(`VM control '${verb}' failed (HTTP ${res.status}, no body)`);
}
if (res.status === 400) {
throw new Error(body?.detail || `'${verb}' was rejected by the server`);
}
return {
verb: body.verb ?? verb,
exit_code: body.exit_code ?? null,
stdout: body.stdout ?? '',
stderr: body.stderr ?? '',
rejected: Boolean(body.rejected),
};
}