14 KiB
Linux-MVP WebUI
React/TypeScript, no runtime demo data, no desktop/PTY/hypervisor integration.
Build assets: web/dist/; serve index.html and /assets/* from the TLS backend
on the same origin. Do not proxy authentication to a different origin. All API URLs
are relative; cookies use credentials: same-origin, mutations use X-CSRF-Token.
No token or password persistence, raw HTML rendering, automatic approval or retry.
Commands
From web/: npm ci, npm run typecheck, npm test, npm run build.
npm run dev binds loopback, but is not the authenticated TLS acceptance setup.
The current server is discovery.ease-technologies.de; management access uses the
configured SSH tunnel and verified TLS. Earlier local rmm-dev.internal:18443
examples below are historical. The notebook is development only, without an RMM
server or enrolled agent. No certificate-verification bypass is supported.
Script console (specification 1.20, source implementation)
run_script remains a durable plan/request, not a PTY or interactive terminal.
The planner accepts complete fixed POSIX/Bash or Windows PowerShell5.1 scripts.
The optional action-wide parameters.privileged_script_profile is exactly
full_maintenance_v1; missing preserves the existing bounded mode, while present
null/unknown values fail. Full maintenance requires root on Linux or SYSTEM on
Windows, plus scripts:execute and the separate root/system scope. SYSTEM scripts
require the explicit profile. The UI selection grants no permission: central/local
opt-in, target capabilities, independent helper checks and actual WebUI approval
remain necessary. Full authority includes filesystem, network and services, and
can alter the agent/PKI or cause a reboot. It has no sandbox/no-reboot promise.
The profile budgets are Linux 1–300s, Windows 1–120s, at most 32768 script bytes
and 65536 output bytes. RemoteSigned, when explicitly bound, remains limited to
the approved Windows job process.
Before submit/approval, src/script-console.ts projects every script from the
SHA256-verified stored plan bytes. It checks full payload structure, fixed
target/job/attempt IDs, script SHA256 and the job SHA256 reconstructed in Rust
declaration order. Every supplied payload/target alias must agree. It snapshots
the checked views, rejects later mutation and rechecks immediately before each
submit/approval action. The displayed script, arguments, identity, profiles,
budgets, deadlines and hashes come from that projection, never the draft editor.
Control characters, CR, BOM and bidi controls are made visible in the text view;
the full original JSON and stored bytes remain available. Existing pending-only
submission and interactivity/CSRF checks remain in force.
src/console-output.tsx displays the actual stored job status and final stdout/
stderr. Status polling is every four seconds; output arrives with the durable
result. There is no incremental output stream. Null result, empty output,
unknown exit code, unknown outcome, requested cancellation and truncation are
distinct. UTF-8 is a display attempt, not a claimed source encoding; invalid
UTF-8 is marked and original bytes can be downloaded as inert .bin files.
Terminal escapes and bidi controls are visible, never interpreted. Agent output
is React text and cannot create HTML, approval controls or follow-up jobs.
MCP already dispatches plan/submit/read-results to the same server application functions. Shared PlanRequest/Action schemas expose the profile; MCP has no approval tool and its bearer context cannot become an interactive approval session. No independent MCP execution or approval path was added.
Local typechecking and renderer tests are source evidence only. Native Linux/
Windows runs, deployment and actual UI/MCP acceptance remain separate gates.
Actually executed on 2026-11-14 for this source slice: npm ci --ignore-scripts --no-audit --no-fund, npm run typecheck, npm test (49 passed across seven
files), npm run build (TypeScript and Vite passed). Five script-console tests
read ../tests/script_console/codec-golden.json directly: the same three actual
Rust codec families cover legacy Linux agent, Linux root/full and Windows
SYSTEM/full with RemoteSigned. Checks cover full displayed script/target/attempt/
profile/budgets/hashes, all aliases, mutation during/after verification and
independent scopes. Two output tests cover hostile HTML/terminal/bidi bytes,
unknown/cancellation/truncation, invalid encodings and absent/invalid results.
No browser, server, VM job, approval or deployment was executed in these checks.
JSON adapter contract
src/api.ts is the sole transport/contract boundary. Responses missing required
fields fail explicitly. Unknown backend fields are displayed without interpretation.
Pages: {items,next_cursor,completeness,truncated,as_of}. Tenant items:
{tenant_id,name} (id accepted as ID alias). Session:
{principal:{id,display_name,simulated:boolean},tenants:[{tenant_id,name}],csrf_token}.
simulated is explicit server evidence, never inferred from a username.
The backend session form {principal:string,tenant_id,testsimulation:boolean, csrf_token,roles,permissions} is also accepted, retaining the exact principal and
tenant ID as labels until the tenant list supplies its authoritative name. Explicit
testsimulation:false is a nonsimulated OIDC session, not a dev identity. Missing
simulation metadata fails closed; neither principal names nor roles imply simulation.
Dev login POST /auth/dev/login {username,password}, GET /auth/session, POST
/auth/logout {}. Session and CSRF stay in memory only. Backend must disable dev
login outside dev/test and audit simulation origin/testprincipal/plan binding.
Before offering login, the UI reads GET /auth/config using same-origin cookies,
no-store and verified browser TLS. Exact DTO:
{oidc_configured:boolean,oidc_login_url:'/auth/oidc/login'|null,dev_login_enabled:boolean}.
The only allowed OIDC link is the canonical relative /auth/oidc/login, and only
when configured. Issuer/external/absolute/return/query/fragment URLs are not used by
the frontend. OIDC callback codes/tokens stay with the backend; the normal reload
reads /auth/session. No local/session storage or auth query logging is used.
The password form requires explicit dev enablement and prominently identifies a
simulated dev/test principal. Production SSO-only configuration has no password
form or simulated-principal label. Configuration loading/error/absent modes never
fall back to dev login; errors and not_configured remain visible.
Executed UI auth integration: typecheck/build, 23 unit tests including SSO-only, config failure, redirect rejection and explicit nonsimulated session parsing; real TLS management/login-config checks plus all three pending/TLS checks passed. The current deployment announces OIDC not configured, dev login enabled. Its real SSO redirect/callback is therefore not exercised by these UI tests; the positive SSO presentation is unit-tested, not external-IdP or production acceptance.
All business endpoints are under /api/v1/tenants/{tenant}:
- GET
/devices[?cursor=],/devices/{asset}:asset_id, presence, health, last_contact_at, source_id, observed_at, received_at, quality, completeness, inventory/capabilities/other sections. Missing evidence staysunknown. - POST
/operation-plans: Spec 23.4 shape{tenant_id,action,targets:{asset_ids}, parameters,schedule:{not_before,start_before,offline_policy},rollout:{max_parallel, halt_after_failures},reason}. Times RFC3339 UTC; common parameters: execution_identity, timeout_seconds, max_output_bytes. Script parameters: variants with{os_family:'linux',interpreter,encoding:'utf-8',script_utf8}, arguments, expected_stdout. Other action fields are explicit JSON, never shell commands. - GET
/operation-plans/{id}:{plan_id,plan_hash,expires_at,resolved_targets:[], excluded_targets:[],plan_payload_bytes|plan_payload,approval_state?}. Preferredresolved_targetsentries are immutable previews{asset_id,job_payload_hash, payload,plan_target_id?}. Alternativelyresolved_targetslists asset IDs or{asset_id}records andpayloadssupplies the previews. Legacyjobsis accepted as the preview list only when neither preferred form is supplied. An incomplete preferred form is not silently replaced by legacy data. In legacy previews only, asset_id can also come from the exact payload. Every target requires a nonempty exact payload and payload hash; preview coverage must equal the fixed target set, duplicate asset IDs and contradictory payload asset IDs are rejected. Script plans require the reservedjob_id/attempt_idfrom the exact stored payload, as well as matching script/job hashes. These are fixed definitions, not proof of executable jobs or persisted attempts. Actual execution rows are created only after approval. Exact payload metadata and script/package hashes must be included. UI does not recompute the plan hash from JSON. Missing fixed targets/previews/payload/hash or excluded targets block submission and approval. All fields are shown, not just target count.approval_stateenables opening an existing pending request after reload. - POST
/operation-plans/{id}/submit {idempotency_key}→{operation_id, approval_state:'pending_approval'}(statealso accepted for submit status). No execution jobs before approval. The button then enables explicit review. - POST
/operation-plans/{id}/authorizations {plan_hash}→{operation_id}; only click handler after explicit checkbox. Backend enforces interactive session, pending state, exact immutable binding, permission, tenant, policy, expiry and deduplicates repeated clicks. No MCP/bearer fallback. - GET
/operations[?cursor=],/operations/{id}→ approval_state, execution_state, plan_id, planned_job_count/job_count, jobs (with job_id), counts. - GET
/jobs/{id}→ state, attempt, cancel_requested, pending_reason, output, results, verification, truncation/reboot flags; raw response shown completely. POST/jobs/{id}/cancel {reason}only requests cancellation.
Errors: {code,message,retryable,correlation_id,details}. Non-JSON errors also remain
visible; failed polling retains the old response alongside an error, never a green
success message. Tenant changes unmount all selection/plans/jobs and abort requests.
List pagination uses server cursors, without pretending a truncated page is complete.
Browser checks and separate acceptance gates
tests/browser/approval.spec.ts uses actual UI interactions only. It never uses an
API client to authorize, never writes storageState, and disables traces/screenshots/
video. Its restricted reporter suppresses raw step logs/errors/stdout/stderr because
Playwright failure call logs can contain password-fill arguments. Do not override
the reporter or enable Playwright DEBUG/PWDEBUG/trace logging for secret-bearing
tests. Failure artifacts remain in gitignored web/node_modules/.cache/browser-results;
inspect them only locally as potentially confidential and never publish them.
It is opt-in via RMM_E2E_PROFILE=dev/test and RMM_E2E_MODE=execute; without
both values no positive authorization/execution test runs. A separate
pending.spec.ts uses RMM_E2E_MODE=pending-only, never sends authorization
requests and verifies login/cookie, device selection, exact preview, pending with
zero executable jobs, session/plan reload, CSRF denial on logout and normal logout.
Set RMM_E2E_ASSET_ID to the explicitly selected Linux test asset and
RMM_E2E_LOGIN_FILE to an owned mode-0600 JSON file {username,password} in the
gitignored credentials directory. Never pass secret values in CLI arguments, log
them, bundle them or use production accounts. The test only submits Hello World;
it never reboots or updates the VM.
The checked configuration isolates browser HOME to
web/node_modules/.cache/browser-home, with a private NSS trust database there;
only the public project dev CA is imported, never CA/server private keys. Browser
launch resolves the two fixed dev test DNS names to PC loopback via Chromium's
host-resolver rules; it does not change hosts files or global trust/configuration.
TLS name/chain verification stays enabled. Node's CA setting alone does not
establish browser trust. No ignore-HTTPS option, insecure flag, direct API
authorization or production simulation is allowed.
Reproduce the pending-only checks with python3 deploy/browser/run_pending.py
from repository root after installing the browser/local NSS tool as documented in
deploy/browser/README.md. The runner imports the public CA, enrolls an explicitly
synthetic local mTLS test device, advertises only the capability required for plan
validation, runs real same-origin browser UI checks, and revokes that fixture in
cleanup. No executor, simulated result, VM access or approval is involved. Pending
requests remain durable and the synthetic device is visibly revoked afterwards.
Actually executed on 2026-11-12: real TLS Chromium login, secure HttpOnly SameSite
cookie, tenant/device list, plan preview, pending submission with zero jobs/null
execution, reload/review, CSRF-denied logout, valid logout, rejection of the
wrong certificate DNS name and rejection of the server chain without the isolated
CA. All three real browser tests passed. A server-side count guard confirmed no new
jobs/outbox/grants. Sanitized local evidence:
web/node_modules/.cache/browser-evidence/pending.json. No positive browser
authorization, Hello-World result or VM feature acceptance is claimed here.
Verified-TLS fetches also confirmed the served same-origin index and both built
asset hashes match the current web/dist build.
Build/unit checks are local development evidence only. Real backend/agent execution, VM feature evidence, browser acceptance and production approval remain separate gates.
Gebundene Configure-only-Wiederherstellung (Schema4)
Frontend prüft native_recovery mit ursprünglichem Grant/Versuch/Partial-Proof,
Transaktionsschema2 und ausschließlich configure_only/Fixture1.0→1.0. Normale
Schemas2/3 erlauben keine Operationserweiterung; Nichtnative-, Altversions-,
Hash- und Bindungsmanipulationen bleiben verweigert. Elf gezielte Safetytests,
Typecheck und Produktionsbuild bestanden. Tatsächliche separate Browser-
Vorbereitung/0-Jobgate/Genehmigung mit exakten Bytes ist belegt; der erste Plan
lief ohne Offer/Grant an der alten Journalgrenze ab. Paketabnahme bleibt separat:
../docs/implementation-status.md. Keine API-/DB-Freigabeabkürzung.