* feat(review): persist and publish native code-review findings
Adds the backend core for reviewing a task's working diff before a PR
exists: task-scoped review runs and anchored, advisory findings that the
Changes/Review panel can render as inline comments.
- task_review_runs / task_review_findings schema, models, and repository,
with supersede-on-re-review and a restart sweep that cancels in-flight
runs rather than leaving them looking alive.
- ReviewService as the single write path, publishing bus events that the
WS gateway forwards to the client.
- internal/review: reviewer-identity resolution (agent profile > code-review
utility agent > user default, failing closed on passthrough-only profiles),
change-set collection, prompt batching, response parsing, anchor
extraction, and run orchestration.
- publish_review_findings_kandev MCP tool plus the task.review.* WS actions.
- Go djb2 that matches apps/web/lib/utils/hash.ts, asserted from shared
fixtures on both sides so stale detection agrees across the wire.
Spec: docs/specs/native-code-review/spec.md
Plan: docs/plans/native-code-review/plan.md
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat(review): add run_code_review workflow step and wire the review runner
Makes a review pass triggerable both ways and actually runnable end to end.
- run_code_review on_enter action across models, engine compile, and an
orchestrator callback that logs rather than propagates failures, so a task
never gets stuck in a step because no reviewer was configured.
- The step's optional agent_profile_id is what lets a step review with a
different model than it implemented with; export/import swap it for a
portable agent name/model/mode so it survives a cross-workspace round trip
instead of carrying a dangling id.
- Concrete adapters for reviewer identity, changed-file collection, prompt
rendering from the code-review utility agent's editable prompt, and
inference (session-bound agentctl, falling back to the host instance so a
review works on a task whose agent is not running).
- Boot sweep cancels review runs interrupted by a restart; the E2E reset
clears review state before deleting the tasks that own it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat(web): add review findings state, helpers, and live updates
Frontend plumbing for native code review, ahead of the UI that renders it.
- Review types, WS message map, and an API client that preserves the backend's
machine-readable failure code so the surface can show an actionable message.
- A review store slice that merges findings by id (a re-publish refreshes an
issue rather than listing it twice) and keeps a bounded run history.
- Findings helpers: composite file keys matching reviewFileKey, staleness via
the same per-file diff hash that drives review marks, and best-effort
re-anchoring from the stored anchor text so a moved finding follows the code
it was written about instead of pointing at whatever now occupies its line.
- Hooks for backfill, resolve/dismiss/reopen with optimistic rollback, and
sending a finding to the agent as follow-up context without changing its
status — handing a finding to an agent is not the human accepting it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat(web): render review findings inline in the Changes diff
Findings now appear where the code is, in the same annotation surface as
anchored review comments, with the run control that produces them.
- ReviewFindingCard: severity chip, category, Markdown body, and the three
advisory actions. A suggestion is labelled "not applied automatically"
because it is display-only in this iteration. Resolved and dismissed
findings collapse to one muted line with an Undo, so a reviewed diff does
not stay cluttered while the human can still change their mind.
- A review-finding annotation type threaded through the diff viewer, anchored
from the same per-file diff hash the backend stamped. A finding whose diff
moved is re-anchored from its stored text when that text survives, and
otherwise reported as unanchored rather than placed against unrelated code.
- Review changes control with idle / running / cancel / failed states, and an
inline Settings link for the one failure the user can actually fix.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat(web): make the review control reachable from the Review toolbar
Mounts the Review changes control next to the walkthrough action and lets an
inline finding wire its own actions.
The finding card container now reads the active task and session from the store
instead of taking four callbacks threaded through FileDiffViewer and DiffViewer;
an inline finding annotation only exists when the active task has findings, so
the store is always present and every other diff consumer is spared the
plumbing. Staleness likewise hashes the diff the viewer already holds, which is
already normalized, so it matches the hash the backend stamped without passing
it down.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat(web): surface stale review findings above their file's diff
A finding whose diff moved and whose anchor text is gone had nowhere to render,
so it was silently dropped. It now appears in a collapsed per-file banner above
the diff: the finding may still be valid, and placing it at its stored line
would point at whatever code now occupies that line instead.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* docs(review): mark the landed native code-review plan tasks done
Tasks 01-09 (backend persistence through inline diff rendering) are
implemented, tested, and committed. Tasks 10-13 — findings overview and step
editor, mobile parity, E2E, and public docs — remain pending.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* test(review): cover the on-demand review path end to end
Adds the mock-agent review responder and a Playwright spec, so both the happy
path and the no-reviewer path are verified against a real backend.
The mock agent answers any prompt carrying the review sentinel with fenced JSON,
deriving its anchors from the prompt's own diff rather than hardcoding them, so
findings land on files and lines the fixture really changed. It includes one
deliberately malformed entry to exercise the rejected-entry path.
Writing it caught an off-by-one shared by three diff walkers: splitting on "\n"
leaves an empty element after a trailing newline, which carries no diff marker
and was being counted as a context line, pushing new-side numbering one past the
end of a hunk. Interior blank lines still count, since a whitespace-stripped
blank context line legitimately appears as "".
Also adds a listInferenceAgents helper: default_utility_agent_id needs the
registered agent-type id, and an agent row UUID is rejected as not
inference-capable.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(review): close review-run races and stale client findings
Addresses the three blockers and three suggestions from the PR #1957 automated
reviews. Two independent reviewers found each blocker separately.
- Launch claimed the task slot only after inserting the run row, so two
concurrent launches both created a run and the loser returned a pending row
with no goroutine behind it — a review that never finished until the next
restart. The slot is now claimed before any DB write, and freed on every
early return.
- Cancelling only marked the row. The goroutine ran on the shared background
context, finished inference, and its completion overwrote the cancelled
status and published findings the user had declined. Runs now carry a cancel
handle, the WS action routes through the runner so the live context is
cancelled, CompleteRun/FailRun refuse to resurrect a terminal run, and a
cancelled pass neither publishes nor re-reports itself as failed.
- The backend deletes a superseded finding and inserts a replacement with a new
id, but the publish event carried only the new one, so every open browser
showed both at the same anchor until a reload. The event now names the
superseded ids and the slice drops them.
- Run() waited on the shared WaitGroup, blocking on every other task's review;
it now waits on its own pass only.
- createReviewSlice is typed over the `set` it actually uses, so call sites no
longer pass superfluous arguments (CodeQL js/superfluous-trailing-arguments).
- A GetAgentProfile fault no longer collapses into "profile not found", which
would have told the user to configure an agent they already have.
The cancel bug exposed a hole in the original tests: they asserted cancel could
not override a terminal run, but never the reverse. Both directions are covered
now, along with the concurrent-launch race, per-run waiting, and the supersede
path on both sides of the wire. Test fakes now mirror the service's terminal
guard rather than hiding it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Kandev Agent <agent@kandev.dev>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* feat(auth): add auth core — identities, sessions, PATs, invites, mode state machine
Phase 1 of opt-in authentication + multi-user segregation:
- internal/auth: argon2id password hashing (PHC), opaque session/PAT tokens
(SHA-256 at rest), login rate limiter, auth-mode state machine
(disabled/setup/enabled), setup-wizard promotion of the pre-auth
default-user row into the admin account, invites, admin user management
with last-admin guard.
- internal/auth/store: auth_identities (OIDC-ready), auth_sessions,
auth_api_tokens, auth_invites tables (SQLite+Postgres dialects).
- internal/auth/authn: request-identity plumbing (gin + plain contexts),
RequireAdmin middleware.
- users table: display_name/role/status columns via new MigrateLogger
wiring; AccountRepository surface (consumer-scoped, keeps existing
Repository fakes untouched).
- config: AuthConfig.Required/SessionTTLHours/CookieName
(KANDEV_AUTH_REQUIRED), FeaturesConfig.Auth; profiles.yaml + runtime
flag registry entry for KANDEV_FEATURES_AUTH.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(auth): enforce opt-in auth over HTTP — middleware, auth API, boot payload
Phase 2 of opt-in authentication + multi-user segregation:
- internal/auth/httpmw: global gin middleware after CORS. Disabled mode
injects the synthetic single-user admin identity (behavior unchanged);
enabled mode resolves session cookie then PAT bearer, with an explicit
allowlist (health, SPA bootstrap reads, credential endpoints,
self-authenticating webhooks), deferral for WS/proxy upgrades and
office agent JWTs, and SPA shell availability for the login page.
- internal/auth/httpapi: full auth surface — setup wizard, login/logout,
me, password change, sessions, PATs, invites (mint/preview/accept),
auth settings (enable/disable), admin user management.
- Boot payload: initialState.auth block always present; unauthenticated
visitors on auth-enabled instances get features+auth only (no data).
- External MCP: PAT required once auth is enabled (open while disabled).
- Admin guards on destructive system routes and runtime-flag overrides.
- Startup warning when bound non-loopback with auth disabled
(ServerConfig.NonLoopbackBinds fail-closed nudge).
- Wiring: auth store in Repositories, auth service in Services,
ClaimUnownedWorkspaces backfill on the task repository.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(auth): per-user workspace scoping across HTTP services
Phase 3 of opt-in authentication + multi-user segregation:
- Task service: workspaces, workflows, repositories, and tasks are scoped
to the request identity (owner-only visibility; pre-auth unowned rows
stay shared). Denials use *NotFound sentinels - no existence leak.
Internal callers and the disabled-mode synthetic identity remain
unscoped, preserving behavior exactly when auth is off.
- CreateWorkspace stamps the authenticated caller as owner.
- Session chokepoint: lifecycle Manager.GetOrEnsureExecution runs a
settable session-access check covering shell/files/ports/vscode/LSP.
- Secrets: per-user user_id column, scoped CRUD/list, ClaimUnowned wired
into the setup-wizard backfills.
- Boot payload inherits scoping automatically.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(auth): scope WebSocket gateway to authenticated users
Phase 4 of opt-in authentication + multi-user segregation:
- Connection auth: /ws, /terminal, /lsp, /vscode, /port-proxy routes are
guarded when auth is enabled - session cookie (same-origin upgrade),
resolved identity from the HTTP middleware, or ?token=<PAT> for
headerless programmatic clients; JSON 401 before any upgrade.
- Client identity: every WS client carries its authn.Identity; dispatched
RPC actions (workspace.list, task CRUD, ...) run under a context
carrying it, so WS RPC gets the same per-user scoping as HTTP.
- Subscription checks: task/session subscribe+focus verify workspace
visibility via the task service; user topic subscriptions are pinned to
the client's own user.
- Broadcast routing: new Hub.BroadcastToWorkspace routes workspace-scoped
events (task/workflow/repo/office/session-state) to the owner's clients
only, with safe fallbacks (unowned/unresolvable -> broadcast) so events
never vanish; instance-wide events (executors, environments, agent
profiles, system jobs) stay global by design.
- Replaced the long-standing discarded-token TODO in HandleConnection.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* docs(auth): add feature spec and backend architecture notes for opt-in auth
- docs/specs/auth/spec.md: product behavior (opt-in, setup wizard,
per-user workspaces, sessions/PATs, invites, public surface, v1 limits)
- apps/backend/AGENTS.md: internal/auth package layout + scoping rules
for new service entry points and WS broadcast call sites
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* docs(auth): public documentation for opt-in authentication
- New docs/public/authentication.md: enabling, users & invites, PATs,
public endpoints, limitations, disabling
- run-as-a-service.md: network-security note now points at the real
authentication boundary instead of reverse-proxy-only guidance
- meta.json: nav entry under Configure Kandev
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* docs(auth): ADR for opt-in authentication architecture
Records the load-bearing choices: mode state machine + synthetic identity
in disabled mode, opaque DB-backed credentials over JWTs, default-user
promotion at setup, service-layer scoping with no-admin-bypass, and the
explicit public allowlist.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(web): auth UI — login, setup wizard, invites, users & account settings
Adds the frontend half of opt-in authentication: a pre-auth gate in
main.tsx that renders the setup wizard or login/invite pages instead of
the app shell whenever the boot payload's auth block says so, a new
`auth` store slice + feature flag mirroring the backend's StatePayload,
credentialed fetch (`credentials: "include"`) with a 401 handler that
clears the session and redirects to /login, and a full auth-api.ts
client for the /api/v1/auth and /api/v1/users routes.
Settings additions: a System > Users page (list/create/disable/promote
users, invite-link dialog) and System > Authentication page (enable/
disable with confirm + env-locked banner), gated on the `auth` feature
flag and admin role. A new Account sidebar group (Profile & Password,
API Tokens) is shown to signed-in users once auth is enabled.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* test(auth): E2E lifecycle suite + fix workspace event WS leak it caught
New dedicated Playwright auth project (backend restarts with
KANDEV_AUTH_REQUIRED=true, isolated from the tokenless default suite):
setup-mode boot gating, admin setup wizard, login/logout + bad
credentials, PATs, admin-surface authorization, anonymous WS rejection,
UI login flow, and the acceptance test - two-user segregation across
HTTP, boot payload, and WebSocket.
The segregation spec immediately caught a real leak: workspace.* event
payloads carry their ID as 'id' (the payload IS the workspace DTO), not
'workspace_id', so BroadcastToWorkspace fell back to broadcast-to-all
for workspace create/update/delete. Fixed with action-aware extraction
plus a unit regression test.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(auth): close cross-user access gaps found in security audit
Two independent security audits of the auth branch found several
authorization gaps where session/workspace-scoped surfaces were not wired
to the new per-user chokepoint. All fixed:
- IDOR on reverse-proxy/terminal routes (vscode, port-proxy, terminal):
these resolve executions via GetExecutionBySessionID /
GetOrEnsureExecutionForEnvironment / EnsurePassthroughExecution, which
skipped the session-access check that only GetOrEnsureExecution ran. Added
the check inside the env/passthrough resolvers (before the cache
short-circuit) and an explicit CheckSessionAccess gate in the two bare-lookup
proxy handlers. An authenticated non-owner can no longer reach another
user's IDE, forwarded ports, or shell.
- Cross-user agent-conversation read: ListMessages / ListMessagesPaginated /
SearchMessages now authorize the session.
- Cross-user repository-script read/inject: script CRUD now authorizes the
parent repository's workspace.
- Office HTTP surface: a group middleware enforces workspace ownership on
:wsId routes for real users (agent-JWT callers keep their claim scoping).
- Login rate-limit bypass: SetTrustedProxies(nil) so a spoofed
X-Forwarded-For can't rotate the limiter key on a directly-reachable
backend.
- Invite double-provisioning race: MarkInviteUsed is now the atomic
serialization point (consumed before user creation).
- Setup wizard: serialized with a mutex so concurrent first-visitors can't
corrupt the promoted admin profile.
- Log sanitizer: explicit kandev_pat_ redaction rule.
Regression tests added for message/script scoping and the
execution-access-before-cache invariant. Auth E2E (incl. two-user
segregation) still green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(auth): scope repository active-session-count endpoint
GET /api/v1/repositories/:id/active-session-count went through
CountActiveSessionsByRepository, which was not identity-scoped - a
cross-user info leak of another owner's active session counts. Added the
same authorizeRepositoryID gate the script routes use.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(auth): add json tags to auth store models — omit digests, fix token UI
The auth store structs (APIToken, Invite, LoginIdentity, Session) had db:
tags but no json: tags, so the handlers that serialize them directly sent
Go-cased keys (CreatedAt, Name) AND leaked the SHA-256/password digests to
clients. Two consequences:
- The API-tokens UI showed 'Invalid Date' and an empty name (frontend reads
created_at/name).
- token_sha256 / password_hash were serialized in list responses.
Added json tags with snake_case keys the frontend reads and json:"-" on
every credential digest. Regression test asserts digests never serialize and
the snake_case timestamp/name keys are present. Also adds the
auth-screenshots capture spec (artifacts gitignored).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(web): move auth enable into feature toggles; show current user in sidebar
Remove the standalone Authentication settings page and its dedicated
settings/get-auth-settings API surface now that enabling auth lives in the
Feature Toggles system, and drop env_required now that /auth/me no longer
returns it. Add a current-user chip with a logout action to the sidebar
footer, shown only in enabled auth mode.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* refactor(auth): make features.auth the single on/off control
Per review feedback: authentication configuration now lives entirely in
the existing Feature Toggles system rather than a separate Authentication
settings section.
- The auth mode is derived from the features.auth runtime flag
(KANDEV_FEATURES_AUTH): off -> disabled, on+no-admin -> setup,
on+admin -> enabled. Removed the parallel auth.mode system setting,
the SetEnabled service method, GET/PATCH /api/v1/auth/settings, the
AuthConfig.Required field, KANDEV_AUTH_REQUIRED, and the auth service's
Settings-store dependency and EnvRequired/env_required.
- The features.auth registry entry is re-described as the actual enable
switch (high-risk, restart-required) and defaults off in every profile
(dev was previously true — that only revealed UI before; now it would
enforce, so it must be an explicit opt-in).
- Setup no longer persists a mode; completing the wizard creates the admin
and the mode auto-derives to enabled.
- Startup 'exposed without auth' warning + all docs (public guide,
run-as-a-service, spec, ADR, AGENTS.md) point at the feature toggle.
- E2E: auth specs use KANDEV_FEATURES_AUTH; screenshot spec captures the
Feature Toggles page instead of the removed Authentication page; the
member-restriction assertion targets an admin invite route.
- Tests updated across service/httpmw/httpapi/profiles for the derived
mode; full suite green (only pre-existing sandbox-FS failures remain).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(auth): complete cross-user isolation — integrations, automation, plan/walkthrough, session reads
A follow-up isolation audit found that per-user scoping had been wired into
the task service, secrets, office, and the WS gateway — but several
workspace/task/session-scoped surfaces still resolved a caller-supplied ID
with no ownership gate. Closed all of them; cross-user access now returns
404 across the board.
- Integrations (jira/linear/sentry/slack/azure-devops/gitlab/github/
workflow-sync): a global middleware authorizes the workspace_id (query, or
gitlab's /workspaces/:id/ path) on those route groups. Previously a member
who knew another user's workspace ID could read/overwrite their integration
connection config and watches.
- Automation: workspace-authorizer seam gates get/list/create/update/delete/
enable/disable, triggers, run history, and — critically — the webhook
secret. Was fully cross-tenant over WS before.
- Task plans (PlanService) and walkthroughs (WalkthroughService): task-
authorizer seam on every read/write. These bypassed the task service.
- Session reads that hit the repo directly — session turns, task context,
git snapshots, file reviews — now call AuthorizeSession/TaskAccess.
Internal callers (pollers, schedulers, MCP) carry no identity and stay
unscoped; disabled mode is unaffected. Regression tests added
(plan/walkthrough scoping, automation authorizer matrix, integration path
extraction) and the two-user E2E now asserts integration routes are shielded.
Docs (spec, ADR, public guide) updated to state what is isolated and that
filesystem/agent-credential isolation remains an OS-level limitation.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(web): current-user chip uses border hover, not accent fill
The sidebar current-user chip used hover:bg-accent (a purpleish fill).
Switch to the border-hover idiom used elsewhere (border border-transparent
-> hover:border-border) with the neutral hover:bg-muted/50 the adjacent
workspace picker uses, so the hover reads as a subtle outline instead of a
colored background. Screenshot spec now captures the chip hover state.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(web): right-align current-user chip in the sidebar footer
Push the user chip to the far right of the expanded footer (ml-auto) so it
sits opposite the left-aligned icon buttons; collapsed mode keeps the
centered vertical stack unchanged.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(auth): reject synthetic identity on management routes and scope office run broadcasts
Address code-review findings on the opt-in authentication PR:
- P1: while auth is disabled the middleware injects a synthetic admin
(RoleAdmin), and RequireAdmin alone let it through — so anyone hitting a
not-yet-enabled instance could mint a PAT or plant an admin that survives
enablement and hijacks first-run setup. Add authn.RequireRealIdentity and
guard the credential/session/user-management routes with it (public
bootstrap routes stay reachable). Responds 404 while disabled.
- P1: office run events carry task_id but no workspace_id, so they fell back
to a global broadcast and leaked across users. Resolve the workspace from
the task and route via the new Hub.BroadcastToWorkspaceOrDrop, which fails
closed (drops rather than global-broadcasts) when the workspace is unknown
and auth is enforced.
- nit: type createAuthSlice as a single-arg creator and call it with just set.
Also register the authentication docs page and its settings routes in
docs/public/coverage.json so the public-docs validator passes.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Kandev Agent <agent@kandev.dev>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Carlos Florêncio