Background vector sync discovered newly provisioned users only on the
periodic user-manager poll (VECTOR_SYNC_USER_POLL_INTERVAL, default 60s),
delaying first indexing by up to a minute. Add a ProvisionSignal doorbell
that provisioning paths ring after storing a user's app password, waking
user_manager_task to re-poll and spawn the user's scanner at once. The
periodic poll remains the backstop (covers cross-replica provisioning).
- ProvisionSignal (stable reference, wait-and-re-arm) held on
VectorSyncState; closes the lost-wakeup window (no await between observing
the ring and re-arming; anyio.Event stickiness covers a mid-poll ring)
- user_manager_task races its poll timeout against the doorbell + shutdown
- notify_user_provisioned() rung from the three app-password provisioning
sites: Login Flow v2 web, MCP provisioning tool, management/BasicAuth API
Note: the pre-existing scanner_wake_event was never .set() and only wakes
existing scanners; a brand-new user has none, so the manager is what must
be nudged.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Round-3 review on PR #891 (no blockers):
- Add test_encode_dav_path_encodes_exactly_once pinning the documented
decoded-input precondition ("already%20encoded.pdf" -> "already%2520...").
- format_exception_group: proper singular/plural ("1 sub-exception" vs
"N sub-exceptions") instead of "(s)".
- oauth_sync: use `if doc_task is not None:` to match processor_task's guard.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two ingest-robustness fixes from card 309 (OHR-Bench smoke-test triage).
WebDAV paths flowed through the client already URL-decoded (unquote on the
PROPFIND/REPORT <d:href>, or raw MCP-tool input), so a '#' reached httpx as a
URL fragment and silently truncated the request -> spurious 404 on otherwise
valid files (e.g. law filenames with '#', commas, double/trailing spaces).
Route every caller-path builder through a new _webdav_path helper that
percent-encodes the path once (preserving separators); the MOVE/COPY
Destination header is encoded too.
Vector-sync runs inside anyio task groups, so a child-task failure surfaced as
a BaseExceptionGroup whose str() is the useless "unhandled errors in a
TaskGroup (N sub-exception)" -- hiding the real ConnectError operators need.
Add format_exception_group to flatten the group to its leaf exceptions and use
it at the broad catch/log sites in processor.py and oauth_sync.py.
Refs: Deck board 12 card 309 (AC #4 filename handling, AC #2 observability).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Nextcloud authenticates app passwords against the *loginName*, which differs
from the UID for OIDC-provisioned users (e.g. user_oidc makes the UID the
display name: UID "Ada Lovelace", loginName "ada@example.com"). The runtime
consumers of stored app passwords bound the UID as the BasicAuth username, so
every Notes/Files/Shares/CalDAV call returned HTTP 401.
PR #818 fixed only the provisioning endpoint; the consuming paths were missed.
Observed on a login_flow tenant (NC's own OIDC app as IdP): the background-sync
scan loop never started ("Credential validation failed ... HTTP 401") and
semantic search returned 0 results because the ACL shared_with_me lookup 401'd
and degraded to a self-only owner filter.
Root cause: NextcloudClient / CalendarClient conflated two identities — the
DAV/URL path identity (the user_id the whole system keys on = NC UID) and the
auth-credential username (the loginName). Decouple them:
- Thread a keyword-only auth_username through NextcloudClient -> CalendarClient
(defaults to username, so single-user / OAuth where UID == loginName is
unchanged).
- get_user_client_basic_auth (background sync + the /api/v1/vector-viz/search
endpoint) authenticates as the stored loginName, UID for paths.
- _get_client_from_login_flow (the get_client(ctx) MCP-tool path) does the same.
- cleanup_invalid_app_passwords validates with the loginName, so it no longer
401s and wrongly deletes a valid OIDC user's password.
The loginName is already persisted in app_passwords.username and returned by
get_app_password_with_scopes. Adds unit tests covering the UID != loginName
split for both client builders, the calendar credential/path split, and the
cleanup validation. Also genericises the example user in the #818 comment/test
(real name/email -> Ada Lovelace / ada@example.com).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
get_user_client_basic_auth built a fresh RefreshTokenStorage and ran
storage.initialize() — a full Alembic `upgrade` in a worker thread — on every
call. Once the /api/v1 search endpoints (unified_search, vector_search) and the
chunk-context endpoint were wired to use it, concurrent requests ran concurrent
Alembic upgrades, which race on Alembic's non-thread-safe module-global
EnvironmentContext proxy and intermittently raise `KeyError: 'script'` →
HTTP 500 (seen on multi-user-basic/nc31; nc32 got lucky).
Cache one process-wide, already-initialized storage instance behind a lazily
created anyio.Lock so the one-time migration runs exactly once and never races.
Callers passing an explicit `storage` are unaffected. Also removes a redundant
per-request migration from the search hot path.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds the seven §10.2 hook-point modules + five env vars so Astrolabe Cloud can
offload document processing to the external document-processor / embedding
gateway. Purely additive: with every setting unset the server behaves exactly
as today, so self-hosters are unaffected (Deck #92).
Hook points (all default to current monolith behavior):
- config: EMBEDDING_PROVIDER, INGEST_MODE, STATUS_BACKEND,
COLLECTION_METADATA_SOURCE, FACT_EVENT_EMITTER (+ supporting settings),
validated in Settings.__post_init__ (fail-fast STATUS_BACKEND=local with
INGEST_MODE=external); shared canonical.py.
- vector/payload_keys.py + acl_hash.py: cross-impl NAMESPACE/point_id (§2.2)
and BLAKE2b-128 ACL hash (§11), pinned by fixtures shared with the
document-processor repo.
- embedding/gateway_client.py: OpenAI-compatible GatewayProvider authenticating
via M2M OIDC client-credentials (separate realm); manual-only registry entry.
- vector/collection_metadata.py: sentinel-point / API metadata source with env
fallback.
- vector/queue/: hexagonal ingest producer ports + memory/NATS adapters
(Postgres seam); INGEST_MODE=external publishes mcp.ingest.requested.{tenant}
instead of the in-memory stream and skips the in-process processor pool. The
lifespan becomes a composition root across both deployment branches.
- vector/queue/status.py: STATUS_BACKEND=bus subscriber feeding a StatusStore
the vector-sync status endpoint reads.
- admin/payload_backfill.py: POST /api/v1/admin/payload-backfill (admin scope);
processor writes the new payload keys; query-side ACL pre-filter gated behind
ACL_PREFILTER_ENABLED (default off).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Follow-up to #787/#789 (ADR-022 cleanup). After
`oauth_enabled ↔ enable_login_flow` became an invariant, the
`use_basic_auth=False` branch in `vector/oauth_sync.py` — and the
parameter wiring that fed it — was no longer reachable from any
supported deployment mode. This commit removes the dead code.
- nextcloud_mcp_server/vector/oauth_sync.py:
- Deleted `get_user_client_oauth` (the OAuth-token refresh helper) and
its `VECTOR_SYNC_SCOPES` constant.
- Deleted the `get_user_client` dispatcher. Internal callers now call
`get_user_client_basic_auth` directly.
- Dropped the `use_basic_auth: bool` parameter from `user_scanner_task`,
`multi_user_processor_task`, `_run_user_scanner_with_scope`, and
`user_manager_task`.
- Dropped the `token_broker` parameter from the same four functions —
they no longer need it now that the OAuth-refresh path is gone. The
`TokenBrokerService` constructed in `app.py` is still used by the
management API revoke endpoint, just not by background sync.
- Simplified the user-list query in `user_manager_task` to always read
from the `app_passwords` table.
- Replaced all `mode_label = "BasicAuth" if use_basic_auth else "OAuth"`
with a literal `[BasicAuth]` log prefix (keeps existing log filters
working).
- Updated the module docstring to describe the post-cleanup shape.
- Dropped the now-unused `TYPE_CHECKING` import of `TokenBrokerService`.
- nextcloud_mcp_server/app.py: dropped the `use_basic_auth = True` block
and the now-stale `token_broker if not use_basic_auth else None` /
`use_basic_auth` positional args from the two `tg.start(...)` calls in
the multi-user vector-sync lifespan. Token broker construction stays —
still consumed by the management API revoke endpoint via
`app.state.oauth_context["token_broker"]`.
- tests/integration/test_app_password_provisioning.py: deleted four tests
that exercised the now-removed OAuth-refresh path
(`test_oauth_mode_uses_refresh_token_only`,
`test_oauth_mode_raises_error_without_token`,
`test_get_user_client_oauth_function`,
`test_oauth_mode_requires_token_broker`) plus the
`test_get_user_client_dispatches_to_basic_auth` test for the deleted
dispatcher. Updated the module docstring + imports accordingly. The
BasicAuth-mode tests (`test_basic_auth_mode_uses_local_storage`,
`test_multiple_users_basic_auth_mode`, etc.) all remain.
No runtime-behaviour change in any supported deployment mode — the deleted
branches were already unreachable post-PR #787. 3 files changed,
+59 / -301; 1010 unit tests pass; integration jobs for
`mcp-login-flow` and `mcp-multi-user-basic` are the critical regression
gates before merge.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Same pattern as the ENABLE_LOGIN_FLOW removal in the previous commit:
the deployment mode (MCP_DEPLOYMENT_MODE) is the single source of truth
for selecting an auth flow. The ENABLE_MULTI_USER_BASIC_AUTH env-var
alias is redundant with `MCP_DEPLOYMENT_MODE=multi_user_basic`.
Unlike the ENABLE_LOGIN_FLOW removal — where silent removal was safe
because Login Flow v2 is the auto-detection default — silent removal
here would be a surprise: a user with only ENABLE_MULTI_USER_BASIC_AUTH=true
in their .env would auto-detect into LOGIN_FLOW after upgrade (wrong
runtime mode). Mitigation: detect_auth_mode now reads os.environ
directly for both legacy aliases and raises ValueError with a one-line
migration message if either is set. Applied retroactively to
ENABLE_LOGIN_FLOW as well — loud is better than silent.
- nextcloud_mcp_server/config.py:
- Drop the dynaconf env-var alias entry for ENABLE_MULTI_USER_BASIC_AUTH.
- Update the `enable_multi_user_basic_auth` field docstring to mark it
as derived / not user-settable.
- `_is_multi_user_mode()` (early-config helper, runs before Settings
is built) switched to checking MCP_DEPLOYMENT_MODE directly. Now
consistent with the canonical detection in detect_auth_mode.
- nextcloud_mcp_server/config_validators.py:
- Drop the auto-detection branch (`if settings.enable_multi_user_basic_auth`).
Selection of MULTI_USER_BASIC is now exclusively via the explicit
MCP_DEPLOYMENT_MODE branch.
- Add `enable_multi_user_basic_auth` to `_sync_derived_flags` alongside
`enable_login_flow` — both flags are now derived from the resolved mode.
- Drop `enable_multi_user_basic_auth` from
`MODE_REQUIREMENTS[MULTI_USER_BASIC].required` and from the
`forbidden` lists of SINGLE_USER_BASIC and LOGIN_FLOW (no longer
user input → no meaningful forbidden check).
- Add loud-deprecation `ValueError` block at the top of detect_auth_mode
that errors with a clear migration message when ENABLE_MULTI_USER_BASIC_AUTH
or ENABLE_LOGIN_FLOW is found in os.environ.
- tests/unit/test_config_validators.py:
- Switch ~10 fixtures from `enable_multi_user_basic_auth=True` to
`deployment_mode="multi_user_basic"` (mirrors `enable_login_flow`
treatment from the previous commit).
- Switch two `patch.dict(os.environ, {"ENABLE_MULTI_USER_BASIC_AUTH": "true"})`
blocks to use MCP_DEPLOYMENT_MODE.
- Rename `test_forbidden_multi_user_basic_auth` to
`test_forbidden_multi_user_basic_when_credentials_present` — the
scenario is now an explicit-mode + credentials conflict, not an
env-var-flag conflict.
- Add `test_legacy_enable_multi_user_basic_auth_env_var_errors` and
`test_legacy_enable_login_flow_env_var_errors` to exercise the new
loud-deprecation ValueError path.
- docker-compose.yml: mcp-multi-user-basic profile switched to
`MCP_DEPLOYMENT_MODE=multi_user_basic`.
- env.sample: replaced `#ENABLE_MULTI_USER_BASIC_AUTH=true` example with
`#MCP_DEPLOYMENT_MODE=multi_user_basic`.
- docs/authentication.md, configuration.md, troubleshooting.md,
auth-flows.md, webhook-management-guide.md,
configuration-migration-v2.md, ADR-025: replaced env-var examples
with the canonical MCP_DEPLOYMENT_MODE form.
- docs/ADR-020: marked partly superseded by ADR-022.
- CLAUDE.md: Multi-User BasicAuth section updated to set
MCP_DEPLOYMENT_MODE.
- nextcloud_mcp_server/vector/oauth_sync.py: module docstring updated.
BREAKING CHANGE: ENABLE_MULTI_USER_BASIC_AUTH is no longer read from
the environment, and setting it now raises a startup ValueError with
a migration message. Replace `ENABLE_MULTI_USER_BASIC_AUTH=true` with
`MCP_DEPLOYMENT_MODE=multi_user_basic`. The same loud-deprecation
check is also applied to the recently-removed ENABLE_LOGIN_FLOW —
replace with `MCP_DEPLOYMENT_MODE=login_flow` (or drop both;
`login_flow` is the auto-detect default when no other auth env vars
are set).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
caldav 3.x lists niquests as a mandatory dependency and prefers it over
httpx. Passing httpx.BasicAuth via the auth= argument breaks under the
niquests backend with "Unexpected non-callable authentication" — see #731.
Switch CalendarClient.__init__ from auth=Auth|None to keyword-only
password/token, and forward them to AsyncDAVClient as password= plus an
explicit auth_type ("basic" or "bearer"). caldav then builds whichever
auth object its active backend needs (niquests.auth.HTTPBasicAuth or
httpx.BasicAuth), so we stay backend-agnostic.
Threaded raw credentials through NextcloudClient — added keyword-only
password/token to its __init__, and updated from_env, from_token, and
the four call sites that build NextcloudClient (context.py basic-auth
and Login Flow paths, auth/userinfo_routes.py, vector/oauth_sync.py).
Four new unit tests pin the construction wiring so the niquests
regression can't recur silently — basic, bearer, no-creds, and
password-precedence cases.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Many identity providers (AWS Cognito, Okta, Azure AD) reject or mishandle
colons in OAuth scope names. This migrates all custom scopes from
`resource:action` to `resource.action` format (e.g., `notes:read` →
`notes.read`), which is universally accepted and aligns with industry
conventions (Microsoft, Google).
Includes Alembic migration 004 for stored scope strings and ADR-024
documenting the rationale and RFC references.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Login Flow v2 is a deployment-wide mode where all users authenticate
with app passwords (not OAuth refresh tokens). Set use_basic_auth=True
when enable_login_flow is true so the background sync user manager
queries the app_passwords table and scanners use app password
authentication for Nextcloud API calls.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
When enable_login_flow is true, also check the app_passwords table
for provisioned users. Previously, OAuth mode only queried the
refresh_tokens table, missing users who were provisioned via
Login Flow v2 (which stores app passwords, not refresh tokens).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Move httpx import to top-level and use anyio task group for concurrent
validation in cleanup_invalid_app_passwords (storage.py)
- Respect Retry-After header for 429 responses, capped at 300s (oauth_sync.py)
- Soften pre-validation exceptions so transient failures don't crash the
background sync task (oauth_sync.py)
- Replace f-string SQL with blanket DELETE and add returncode checks (conftest.py)
- Extract clear_stale_test_state() helper to deduplicate cleanup logic
in astrolabe background sync tests
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The revoke test failed because it only completed Step 2 (app password) but
not Step 1 (OAuth authorization). In hybrid mode, Astrolabe requires both
steps for $isFullyConfigured=true, which gates the "Revoke Access" button.
Changes:
- Use complete_astrolabe_authorization() in revoke test for full two-step flow
- Add stale state cleanup (app passwords, bruteforce entries, Astrolabe prefs)
to both enablement and revoke tests
- Add startup cleanup of invalid app passwords in BasicAuth mode
- Pre-validate credentials before entering scanner loop to fail fast
- Handle 401/403/429 in scanner with proper backoff and circuit breaking
- Clean up app passwords in test_users_setup fixture teardown
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Previously, the multi-user BasicAuth mode attempted to retrieve app passwords
via OAuth client_credentials grant, which Nextcloud OIDC doesn't support.
This fix implements local storage for app passwords:
- Add app_passwords table via Alembic migration (002)
- Add store/get/delete methods to RefreshTokenStorage
- Add management API endpoints for app password provisioning:
- POST /api/v1/users/{user_id}/app-password
- GET /api/v1/users/{user_id}/app-password
- DELETE /api/v1/users/{user_id}/app-password
- Update oauth_sync.py to read from local storage
- Update Astrolabe to send app passwords to MCP server after validation
- Add app-hook to configure mcp_server_url in Nextcloud
The flow is now:
1. User creates app password in Nextcloud Security settings
2. User enters it in Astrolabe Personal Settings
3. Astrolabe validates against Nextcloud, then sends to MCP server
4. MCP server stores encrypted app password locally
5. Background sync uses locally stored password
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Completely separates multi-user BasicAuth mode from OAuth mode with no
fallback between them. These are now mutually exclusive authentication
strategies based on deployment configuration.
Changes:
- Create separate functions: get_user_client_basic_auth() and
get_user_client_oauth() with clear separation of concerns
- Update get_user_client() to dispatch based on use_basic_auth parameter
- Pass use_basic_auth through all background sync tasks
- Update app.py to determine auth mode at startup
- Rewrite integration tests to verify no OAuth fallback in BasicAuth mode
- Fix test assertions for response field names and duplicate title handling
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Implement multi-user BasicAuth pass-through mode (ADR-020) where each
request includes BasicAuth credentials that are forwarded to Nextcloud
APIs without persistent storage.
Changes:
- Add _get_client_from_basic_auth() in context.py to extract credentials
from Authorization header (set by BasicAuthMiddleware)
- Add AstrolabeClient for app password provisioning via Astrolabe API
- Update oauth_sync.py with dual credential support (app passwords first,
then refresh tokens as fallback)
- Simplify oauth_tools.py provisioning logic
- Add integration tests for app password provisioning and multi-user BasicAuth
Features:
- Stateless multi-user mode: credentials passed per-request
- Optional background sync via app passwords (stored in Astrolabe)
- Falls back to refresh tokens if app password not available
- Test coverage for provisioning flow and pass-through mode
Related: ADR-019 (Multi-user BasicAuth), ADR-020 (Deployment Modes)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
