Agents are transient and should not be persisted in the users table.
The assignRoleToUser call caused a FK violation (user_roles → users),
resulting in HTTP 500 on registration. The AGENT role is already
embedded directly in the JWT claims.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Previously the refresh endpoint only returned a new accessToken, causing
agents to lose their refreshToken after the first refresh cycle and
forcing a full re-registration every ~2 hours.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Single JSONB key-value table replaces two singleton config tables, making
future config types trivial to add. Also fixes pre-existing IT failures:
Flyway URL not overridden by Testcontainers, threshold test ordering.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Database tables filtered to current_schema(), active queries to
current_database(), OpenSearch indices to configured index-prefix.
Delete endpoint rejects indices outside application scope.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- RoleAdminController.createRole: default null description to "" and null scope to "custom"
- RoleAdminController.updateRole: pass null audit details to avoid NPE when name is null
- GroupAdminController.updateGroup: pass null audit details to avoid NPE when name is null
- UserAdminController: add POST / createUser endpoint with default VIEWER role assignment
- UserAdminController: add PUT /{userId} updateUser endpoint for displayName/email updates
- Add V2 Flyway migration to create built-in Admins group (id: ...0010) with ADMIN role
- Add ADMINS_GROUP_ID constant to SystemRole
- Add user to Admins group on successful local login alongside role assignment
GroupAdminController with cycle detection, RoleAdminController
with system role protection, RbacStatsController for dashboard.
Rewrite UserAdminController to use RbacService.
New package com.cameleer3.server.core.rbac with SystemRole constants,
detail/summary records, GroupRepository, RoleRepository, RbacService.
Remove roles field from UserInfo. Implement PostgresGroupRepository,
PostgresRoleRepository, RbacServiceImpl with inheritance computation.
Update UiAuthController, OidcAuthController, AgentRegistrationController
to assign roles via user_roles table. JWT populated from effective system roles.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
When spring.flyway.url is set independently, Spring Boot does not
inherit credentials from spring.datasource. Add explicit user/password
to both application.yml and K8s deployment to prevent "no password"
failures on feature branch deployments.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Flyway needs public in the search_path to access TimescaleDB extension
functions (create_hypertable). The app datasource must NOT include public
to prevent accidental cross-schema reads from production data.
- spring.flyway.url: currentSchema=<branch>,public (extensions accessible)
- spring.datasource.url: currentSchema=<branch> (strict isolation)
- SPRING_FLYWAY_URL env var added to K8s base manifest
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
OIDC configuration is already fully database-backed (oidc_config table,
admin API, OidcConfigRepository). Remove the redundant env var binding
(SecurityProperties.Oidc), the env-to-DB seeder (oidcConfigSeeder), and
the OIDC section from application.yml.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Enable deploying feature branches into isolated environments on the same
k3s cluster. Each branch gets its own namespace (cam-<slug>), PostgreSQL
schema, and OpenSearch index prefix for data isolation while sharing the
underlying infrastructure.
- Make OpenSearch index prefix and DB schema configurable via env vars
(defaults preserve existing behavior)
- Restructure deploy/ into Kustomize base + overlays (main/feature)
- Extend CI to build Docker images for all branches, not just main
- Add deploy-feature job with namespace creation, secret copying,
Traefik Ingress routing (<slug>-api/ui.cameleer.siegeln.net)
- Add cleanup-branch job to remove namespace, PG schema, OS indices
on branch deletion
- Install required tools (git, jq, curl) in CI deploy containers
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Use singleton container pattern for PostgreSQL + OpenSearch testcontainers
(fixes container lifecycle issues with @TestInstance(PER_CLASS))
- Fix table name route_executions → executions in DetailControllerIT and
ExecutionControllerIT
- Serialize processor headers as JSON (ObjectMapper) instead of Map.toString()
for JSONB column compatibility
- Add nested mapping for processors field in OpenSearch index template
- Use .keyword sub-field for term queries on dynamically mapped text fields
- Add wildcard fallback queries for all text searches (substring matching)
- Isolate stats tests with unique route names to prevent data contamination
- Wait for OpenSearch indexing in SearchControllerIT with targeted Awaitility
- Reduce OpenSearch debounce to 100ms in test profile
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Delete all ClickHouse storage implementations and config
- Delete old core interfaces (ExecutionRepository, DiagramRepository, MetricsRepository, SearchEngine, RawExecutionRow)
- Delete ClickHouse SQL migration files
- Delete AbstractClickHouseIT
- Update controllers to use new store interfaces (DiagramStore, ExecutionStore)
- Fix IngestionService calls in controllers for new synchronous API
- Migrate all ITs from AbstractClickHouseIT to AbstractPostgresIT
- Fix count() syntax and remove ClickHouse-specific test assertions
- Update TreeReconstructionTest for new buildTree() method
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The empty servers list caused Swagger UI to construct request URLs
without the /api/v1 prefix, resulting in CORS/fetch failures.
Adding a relative server entry makes paths resolve correctly.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The active-count query scanned all wide rows on the base table, exceeding
the 3.6 GiB memory limit. Use PREWHERE status = 'RUNNING' so ClickHouse
reads only the status column first. Add SETTINGS max_memory_usage = 1 GiB
to all queries so concurrent requests degrade gracefully instead of crashing.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The JDBC driver sends java.sql.Timestamp with nanoseconds as a string
(e.g. '2026-03-15 10:13:58.105931162') which DateTime('UTC') rejects.
Add bucketTimestamp() helper that truncates to seconds for all rollup
query parameters.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
JDBC sends Timestamp params as strings, causing toStartOfFiveMinutes()
to fail with 'Illegal type String'. Floor to 5-minute boundaries in
Java instead and pass plain bucket >= ? comparisons.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
countMerge(total_count) in the avg expression resolved to the UInt64
alias 'total_count' instead of the AggregateFunction column. Rename
SELECT aliases (cnt, failed, avg_ms, p99_ms) to avoid shadowing.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
ClickHouse count() accepts no arguments, so the column type must be
AggregateFunction(count) not AggregateFunction(count, UInt64). The
latter causes countMerge() to fail with ILLEGAL_TYPE_OF_ARGUMENT.
Drop and recreate the table/MV to apply the corrected schema.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Pre-aggregate route execution stats into 5-minute buckets using a
materialized view with -State/-Merge combinators. Rewrite stats() and
timeseries() to query the rollup table instead of scanning the wide
base table. Active count remains a real-time query since RUNNING is
transient. Includes idempotent backfill migration.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
NODE_SPACING 40→60 gives edges more vertical room between nodes.
FIXED_H reverted to 40 to match backend NODE_HEIGHT.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Backend: Add group filtering to agent list, search, stats, and timeseries
endpoints. Add diagram lookup by group+routeId. Resolve application group
to agent IDs server-side for ClickHouse IN-clause queries.
Frontend: New route detail page at /apps/{group}/routes/{routeId} with
three tabs (Diagram, Performance, Processor Tree). SVG diagram rendering
with panzoom, execution overlay (glow effects, duration/sequence badges,
flow particles, minimap), and processor detail panel. uPlot charts for
performance tab replacing old SVG sparklines. Ctrl+Click from
ExecutionExplorer navigates to route diagram with overlay.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Sorting now applies to the entire result set via ClickHouse ORDER BY
instead of only sorting the current page client-side. Default sort
order is timestamp descending. Supported sort columns: startTime,
status, agentId, routeId, correlationId, durationMs.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
ReplacingMergeTree only deduplicates during background merges, so
every login was inserting a new row even when all fields were identical.
Now compares the existing record and skips the write if nothing changed.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace the static SCHEMA_FILES array with classpath pattern matching
(classpath:clickhouse/*.sql). Migration files are discovered and sorted
by filename, so adding a new numbered .sql file is all that's needed.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The 06-oidc-display-name-claim.sql migration was not registered in
ClickHouseConfig.SCHEMA_FILES, so the ALTER TABLE never ran on
existing deployments, causing startup failure when the repository
tried to SELECT the missing column.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
On re-login the upsert was inserting a new row with created_at=now(),
causing ClickHouse ReplacingMergeTree to accumulate rows until
background compaction. Now preserves the original created_at via
INSERT...SELECT from the existing record.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
