Replace the deprecated @react-email/components dependency (and its
deprecated sub-packages) with the unified react-email package per the
React Email 6 migration, updating all component and render imports across
the email templates from "@react-email/components" to "react-email".
This removes the npm deprecation warnings emitted by the old
@react-email/* packages while keeping the template build output
unchanged.
Signed-off-by: Sacha Al Himdani <sacha@probo.com>
Route document deletion through explicit confirmation dialogs so single and bulk delete actions require users to acknowledge the destructive operation before the mutation runs.
Signed-off-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: Sacha Al Himdani <SachaProbo@users.noreply.github.com>
The common tracker pattern "show" only printed attempt counters, so the
enrichment payload — run status, model, agent attribution, and per-field
outcomes — was invisible, and a partial run was indistinguishable from a
complete one. Render that payload the same way the common third party
"show" already does.
Surface the last enrichment attempt timestamp in both catalog list views
so an operator can spot stale or never-attempted rows at a glance.
Signed-off-by: Émile Ré <emile@probo.com>
The migration dropped the DEFAULT on
common_tracker_patterns.enrichment_attempts, so the column is now
NOT NULL with no default and inserts must set it explicitly. The e2e
seedCommonTrackerPattern helper still did a raw insert without it,
which broke TestThirdParty_ImportFromCommon and
TestTrackerPattern_CommonTrackerPatternID with a 23502 violation. Pass
0 for the seeded rows.
Signed-off-by: Émile Ré <emile@probo.com>
Three crash- and migration-recovery gaps in the unified enrichment
model left rows stuck or misclassified:
- Upsert re-armed a blank, newly-linked tracker pattern without
clearing its prior enrichment payload. A crash between the worker's
claim and persist then left the row with a stale payload, so the
stale-recovery sweep (which only catches rows with a null payload)
skipped it forever. Clear enrichment on re-arm so the row reads as
not-yet-completed again, and pin the behavior with a test.
- The migration added last_enrichment_attempt_at to
common_third_parties without seeding it. Rows with prior attempts
kept a NULL clock and could never satisfy the stale-reset predicate.
Backfill from updated_at, the historical claim-time proxy.
- The migration switched the tracker-pattern enriched-state source to
the enrichment payload without backfilling rows previously marked by
enriched_at, making already-enriched rows read as unenriched.
Seed a provenance sentinel for rows that carried the old done-flag.
Signed-off-by: Émile Ré <emile@probo.com>
Let an operator create or update a catalog third party and queue it for
the async enrichment worker in one step, so a minimal name/category row
gets its profile, compliance documents, owned domains, and logo filled
in without a separate reenrich call.
Enrichment is armed explicitly via RequestEnrichmentByIDs rather than
through the receiver: the coredata Upsert sets enrichment_requested_at
from the receiver only on insert and leaves it untouched on conflict,
so this is the only path that re-arms an existing row uniformly while
also resetting the attempt budget for a fresh run.
Signed-off-by: Émile Ré <emile@probo.com>
Make common_tracker_patterns and common_third_parties share one
enrichment-tracking model and fix the misleading proboctl status.
Both tables now carry the enrichment JSONB provenance payload, an
enrichment_attempts counter, and a last_enrichment_attempt_at clock.
On common_tracker_patterns the enriched_at done-flag is renamed to
last_enrichment_attempt_at and stamped at claim time, so it is truthful
to "attempt" rather than "success". A row is considered to have been
through the workflow when it carries an enrichment payload, not when a
timestamp is set, which lets stale recovery key off the payload being
absent with budget remaining, exactly like common_third_parties.
The claim path reads the attempt counter and timestamp back via
RETURNING so the in-memory receiver matches the database clock instead
of a separate app-side time.Now.
The enricher builds a per-field provenance payload (description and
third-party outcomes plus the mapping attribution) and persists it via
UpdateEnrichment, named to mirror the common-third-party sibling. The
common pattern enrichment worker gains a max-attempts ceiling so a
permanently failing row stops looping.
proboctl now shows "enriched" only when every field the last run
recorded an outcome for resolved a value, otherwise "partial (X/Y)",
replacing the misleading "enriched (no description)" label.
Signed-off-by: Émile Ré <emile@probo.com>
A token without admin.directory.customer.readonly makes
Customers.Get return 403 forever. Treat that as terminal so
the source-name worker keeps the generic name and stops
retrying.
Signed-off-by: Bryan Frimin <bryan@probo.com>
Unbounded LoadAll* loaders materialised an entire result set in one
query with no ceiling. A table that is small in development can grow
without bound in production, so these loaders were a latent memory
and query-time hazard.
Remove the LoadAll* methods from pkg/coredata and walk the cursor-
paginated LoadBy* siblings instead through a shared page.LoadAll
helper. The helper advances a MaxCursorSize forward cursor until the
result set is exhausted and concatenates the pages. It caps a single
call at MaxLoadAllPages (20) batches of 500 rows and errors past that
rather than materialising an unbounded set, so a runaway caller fails
loudly instead of exhausting memory.
Callers that genuinely need every row now express that explicitly,
and the coredata load-naming rule and docs are updated to discourage
new unbounded loaders.
Signed-off-by: Sacha Al Himdani <sacha@probo.com>
The release guide was refactored from a single contrib/claude/release.md
into a contrib/claude/release/ directory with a README index and per-track
guides, but the AGENTS.md link was never updated and pointed at a missing
file. Repoint it to release/README.md. CLAUDE.md is a symlink to AGENTS.md,
so it inherits the fix.
Signed-off-by: Sacha Al Himdani <sacha@probo.com>
The MCP Profile schema declares additional_email_addresses as a
required, non-nullable array, but NewProfile passed the mail.Addrs
value through unchanged. A profile with no extra emails has a nil
slice, which marshals to JSON null and fails tool output validation
with "type: null, want array".
Default the nil slice to an empty mail.Addrs so the marshalled output
always honors the strict array schema.
Signed-off-by: Sacha Al Himdani <sacha@getprobo.com>
Long role strings in the access review table broke row layout when
drivers joined many roles into one comma-separated value. Expose
roles as a string array in GraphQL by splitting the stored role at
the API layer, and render the first three roles as badges with a
"+X more" popover for the rest.
Closes ENG-459.
Signed-off-by: Ludovic Vielle <ludovic@probo.com>
Employee document signatures recorded an acknowledgment with no
cryptographic proof, unlike document approvals which already create
and accept an electronic signature on every decision.
Mirror the approval flow on the sign path: generate the signed
document PDF, create-and-accept an esign record, and persist its id
on the document_version_signatures row through a new
electronic_signature_id column. Capture the signer IP and user agent
in the resolver, and re-check the published/archived preconditions
inside the transaction so the seal cannot race document state.
Make the consent wording a single backend source of truth shared by
the text that is sealed and the text shown in the UI. Define
DocumentSignatureConsentText and DocumentApprovalConsentText in the
probo service package and the NDA copy in the trust service, each
owned by the flow that uses it, and stop esign from appending the
generic clause to caller-provided consent text so approvals no
longer seal a duplicated sentence.
Expose the resolved consent text through GraphQL on
EmployeeDocumentVersion and DocumentVersionApprovalDecision, and have
the signing, approval, and NDA pages render it from the API instead
of hard-coded strings, mirroring how the NDA page already worked.
Align the wording with the actual interaction: the buttons read
"Review and sign" and "Review and approve", the clauses reference
those actions, and the inaccurate "typing my full name" phrasing is
removed everywhere.
Signed-off-by: Sacha Al Himdani <sacha@getprobo.com>
Denials were invisible in the audit trail and evaluator explainability
(policy_id, reason) was discarded before reaching logs. Emit a structured
authz decision line on every evaluation in evaluateMultiInTx — allow,
deny, no_match, and assumption errors — using the existing authorizer
logger with opaque IDs only.
Add decision_log.go with DecisionRecord and logDecision. Surface
PolicyID and Reason on EvaluationResult for logging. Audit log
behavior is unchanged (allow-only). Document the convention in
authorization.md.
Signed-off-by: Ludovic Vielle <ludovic@probo.com>
The campaign source node test queried entries before starting the
campaign, and the lifecycle test used the old accessEntry field name
instead of accessReviewEntry.
Signed-off-by: Bryan Frimin <bryan@probo.com>
Expose campaign sources as first-class nodes, paginate fetch attempts
instead of denormalized status fields, and bind entries to their
campaign snapshot. Update GraphQL, MCP, CLI, console, and e2e coverage
to match.
Signed-off-by: Bryan Frimin <bryan@probo.com>
Source category was never surfaced in the product and added noise to
snapshots and APIs. Remove the enum, columns, and service fields so
campaign sources track only the identity fields reviewers need.
Signed-off-by: Bryan Frimin <bryan@probo.com>
Fresh installs already had csv_data on campaign sources from the
initial migration; add a follow-up migration to backfill existing
rows from the live source snapshot.
Signed-off-by: Bryan Frimin <bryan@probo.com>
Bad API keys and expired OAuth tokens showed Connected because
probes ran only for OAuth2 and many providers had no ProbeURL.
Add a registry ProbeConnection dispatcher with static, dynamic,
and custom probes so all 41 providers are checked on demand.
Signed-off-by: Bryan Frimin <bryan@probo.com>
Cursor's Admin API exposes two independent removal signals: the isRemoved
boolean and a role value of "removed". They are not always consistent —
a member can carry role "removed" while isRemoved is still false, a known
gap documented on the Cursor community forum.
Previously Active was derived from isRemoved alone, so a member with
role "removed" but isRemoved=false was incorrectly reported as active.
Now either signal is sufficient to mark the account inactive.
Add a cassette entry and test case covering the inconsistent state
(role "removed", isRemoved false) to prevent regression.
Signed-off-by: Bryan Frimin <bryan@probo.com>
Decouple each campaign from the live access-review sources it was started
with by introducing a per-campaign source snapshot table
(access_review_campaign_sources). The snapshot captures the source name,
category, and connector at start time, so a review remains coherent even
after the underlying source is edited or deleted. Fetch tracking becomes
an append-only log (access_review_campaign_source_fetch_attempts) that
preserves every attempt with its own status and error rather than
overwriting a single row.
Rename the shared access-review tables and enums to use a consistent
access_review_ prefix throughout:
access_entries → access_review_entries
access_sources → access_review_sources
access_source_category → access_review_source_category
access_entry_* → access_review_entry_*
The same rename propagates to every coredata type, service, GraphQL
schema, MCP specification, CLI command, frontend component, and e2e test.
The accessreview package gains dedicated actions.go and policies.go files
for its own IAM policy set, mirroring the agentrun package pattern.
Signed-off-by: Bryan Frimin <bryan@probo.com>
The enrichment agent now reads full cookie-database and vendor pages
through the browser. The facts it gathers (setter, purpose, data
stored) are not protectable, but a source page's written description
is original expression. Instruct the agent to state the facts in its
own words and never copy a sentence verbatim, so stored descriptions
stay clear of third-party prose.
Signed-off-by: Émile Ré <emile@probo.com>
The tracker-mapping and common-pattern enrichment agents only had
web search, which returns title/url/snippet, so they could never
open a cookie-database or cookie-policy page to read which vendor
actually sets a tracker. This mis-attributed setters whose snippet
is misleading (e.g. _li_* read as LinkedIn rather than LiveIntent).
Wire the read-only headless-browser toolset into both agents, gated
on a configured Chrome endpoint, mirroring the common-third-party
enrichment worker: agent construction moves into the run path so each
run can carry a per-run browser that is closed when the run returns.
The prompts now direct the agent to open a promising result and read
the named setter from the full page text. Both agents stay unchanged
when no Chrome endpoint is configured.
Signed-off-by: Émile Ré <emile@probo.com>
Brand assets served at /api/files/v1/static had no cache headers.
Introduce brand.Assets to own the embedded filesystem, content-hash
ETags, and HTTP serving. Responses now carry Cache-Control and ETag
so clients can cache and revalidate; stable email URLs stay
revalidatable (max-age=3600, no immutable).
Replace hardcoded Default*Path constants with StaticPathPrefix,
logical filename constants, and StaticPath(). NewAssets validates
required assets at startup so a rename fails fast instead of 404ing
in sent emails. The files handler keeps routing and 404 rendering;
ServeAssets sets cache headers and serves the file.
Signed-off-by: Ludovic Vielle <ludovic@probo.com>
