Add ON_COMPLETION to backend COMPOUND_TYPES and frontend rendering.
Completion handlers render as teal-tinted sections between the main
flow and error handlers, structurally parallel to onException.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
ELK renderer:
- Add EIP_WHEN, EIP_OTHERWISE, DO_CATCH, DO_FINALLY to COMPOUND_TYPES
so branch body processors nest inside their containers
- Rewrite node creation and result extraction as recursive methods
to support compound-inside-compound (CHOICE → WHEN → processors)
- Use fixed NODE_WIDTH=160 for leaf nodes instead of variable width
Frontend:
- Fix mousewheel crash: capture getBoundingClientRect() before
setState updater (React nulls currentTarget after handler returns)
- Anchor fitToView to top-left instead of centering
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
New interactive route diagram component with SVG rendering using
server-computed ELK layout coordinates. TIBCO BW5-inspired top-bar
card node style with zoom/pan, hover toolbars, config badges, and
error handler sections below the main flow.
Backend: add direction query parameter (LR/TB) to diagram render
endpoints, defaulting to left-to-right layout.
Frontend: 14-file ProcessDiagram component in ui/src/components/
with DiagramNode, CompoundNode, DiagramEdge, ConfigBadge, NodeToolbar,
ErrorSection, ZoomControls, and supporting hooks. Dev test page at
/dev/diagram for validation.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
ON_EXCEPTION and ERROR_HANDLER nodes are now treated as compound containers
in the ELK diagram renderer, nesting their children. The frontend
diagram-mapping builds separate FlowSegments for each error handler,
displayed as distinct sections in the RouteFlow component.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add attributes_text flattened field to OpenSearch indexing for both
execution and processor levels. Include in full-text search queries,
wildcard matching, and highlighting. Merge processor-level attributes
into ExecutionSummary. Add 'attribute' category to CommandPalette
(design-system 0.1.17) with per-key-value results in the search UI.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
POST /api/v1/search/executions is a read-only query using POST for the
request body. Skip it in AuditInterceptor to avoid flooding the audit
log with search operations.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
TreeReconstructionTest and PostgresExecutionStoreIT still passed the
removed diagramNodeId parameter. Missed by mvn compile (main only);
caught by mvn verify (test compilation).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Align with cameleer3-common rename: logForwardingLevel → applicationLogLevel
(root logger) and new agentLogLevel (com.cameleer3 logger). Both fields
are on ApplicationConfig, pushed via config-update. UI shows "App Log Level"
and "Agent Log Level" on AppConfig slide-in, AgentHealth config bar, and
AppConfigDetailPage.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Store application_name in route_diagrams at ingestion time (V7 migration),
resolve from agent registry same as ExecutionController. Move
findProcessorRouteMapping from ExecutionStore to DiagramStore using a
JSONB query that extracts node IDs directly from stored RouteGraph
definitions. This makes the mapping available as soon as diagrams are
sent, before any executions are recorded.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Agent now uses Camel processorId as RouteNode.id, eliminating the
nodeId mapping layer. Drop diagram_node_id column (V6 migration),
remove from ProcessorRecord/ProcessorNode/IngestionService/DetailService,
add /processor-routes endpoint for processorId→routeId lookup,
simplify frontend diagram-mapping and ExchangeDetail overlays,
replace N diagram fetches in AppConfigPage with single hook.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The frontend sends full ISO timestamps (e.g. 2026-03-19T17:55:29Z) but
the controller expected LocalDate (yyyy-MM-dd). This caused null parsing,
which threw NullPointerException in the repository WHERE clause. Changed
to accept Instant directly with sensible defaults (last 7 days).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds CompletableFuture-based request-reply mechanism for commands that
need synchronous results. CommandReply record in core, pendingReplies
map in AgentRegistryService, test-expression endpoint on config controller
with 5s timeout. CommandAckRequest extended with optional data field.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
DetailService deserializes attributes JSON from ExecutionRecord/ProcessorRecord and
passes them to ExecutionDetail and ProcessorNode constructors. ExecutionDocument and
ProcessorDoc carry attributes as a JSON string. SearchIndexer passes attributes when
building documents. OpenSearchIndex includes attributes in indexed maps and
deserializes them when constructing ExecutionSummary from search hits.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
IngestionService passes attributes (currently null, pending cameleer3-common update)
to ExecutionRecord and ProcessorRecord. PostgresExecutionStore includes the
attributes column in INSERT and ON CONFLICT UPDATE (with COALESCE), and reads
it back in both row mappers.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Add prefix query parameter to /admin/opensearch/indices endpoint so
the UI can fetch execution and log indices separately. OpenSearch admin
page now shows two card sections: Execution Indices and Log Indices,
each with doc count and size summary. Page restyled with CSS module
replacing inline styles. Delete endpoint also allows log index deletion.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add admin page at /admin/appconfig with a DataTable showing all
application configurations. Inline dropdowns allow editing log level,
engine level, payload capture mode, and metrics toggle directly from
the table. Changes push to agents via SSE immediately.
Also adds a config bar on the AgentHealth page (/agents/:appId) for
per-application config management with the same 4 settings.
Backend: GET /api/v1/config list endpoint, findAll() on repository,
sensible defaults for logForwardingLevel/engineLevel/payloadCaptureMode.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The /routes/catalog endpoint now accepts optional from/to query
parameters instead of hardcoding a 24h window. The UI passes the
global filter time range so sidebar counts match what the user sees.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Defensive: use .keyword on the top-level exchangeId field too, in
case indices were created before the explicit keyword mapping was
added to the template.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Dynamically mapped string fields in OpenSearch are multi-field
(text + keyword). Term queries require the .keyword sub-field for
exact matching.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Existing log records only have exchangeId inside the mdc object, not
as a top-level indexed field. Use a bool should clause to match on
either exchangeId (new records) or mdc.camel.exchangeId (old records).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Index exchangeId from Camel MDC (camel.exchangeId) as a top-level
keyword field in OpenSearch log indices. Add exchangeId filter to
the log query API and frontend hook. Show a LogViewer on the
ExchangeDetail page filtered to that exchange's logs, with search
input and level filter pills.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add GET /api/v1/logs endpoint to query application logs stored in
OpenSearch with filters for application, agent, level, time range,
and text search. Wire up the AgentInstance LogViewer with real data
and an EventFeed-style toolbar (search input + level filter pills).
Fix agent events timeline autoscroll by reversing the DESC-ordered
events so newest entries appear at the bottom where EventFeed
autoscrolls to.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
LogIndexService in server-core imported LogEntry from cameleer3-common,
but the SNAPSHOT on the registry may not have it yet when the server CI
runs. Moved the dependency to server-app where both the controller and
OpenSearch implementation live.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Agents can now send application log entries in batches via POST /api/v1/data/logs.
Logs are indexed directly into OpenSearch daily indices (logs-{yyyy-MM-dd}) using
the bulk API. Index template defines explicit mappings for full-text search readiness.
New DTOs (LogEntry, LogBatch) added to cameleer3-common in the agent repo.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The findByApplication query only read config_val JSONB, ignoring the
version and updated_at SQL columns. The JSON blob contained version 0
from the original save, so agents saw no config and fell back to defaults.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add highlight field to ExecutionSummary record
- Request highlight fragments from OpenSearch when full-text search is active
- Pass matchContext to command palette for display
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add application_config table (V4 migration), repository, and REST
controller. GET /api/v1/config/{app} returns config, PUT saves and
pushes CONFIG_UPDATE to all LIVE agents via SSE. UI tracing toggle
now uses config API instead of direct SET_TRACED_PROCESSORS command.
Tracing store syncs with server config on load.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add logging to MetricsController: warn on parse failures, debug on
received metrics, buffer depth on 503
- Add GET /api/v1/admin/database/metrics-pipeline diagnostic endpoint
(buffer depth, row count, distinct agents/metrics, latest timestamp)
- Fix BackpressureIT test JSON to match actual MetricsSnapshot schema
(collectedAt/metricName/metricValue instead of timestamp/metrics)
- Upgrade cameleer3-common from 1.0-SNAPSHOT to 0.0.3 (adds engineLevel)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add application_name filter to OpenSearch query builder — sidebar
app selection now correctly filters the exchange list. The
application field was being resolved to agentIds in the controller
but never applied as a query filter in OpenSearch.
Also restore snake_case sort column mapping since the OpenSearch
toMap() serializer uses snake_case field names (start_time, route_id,
etc.), not camelCase.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add engine_level, input_body, output_body, input_headers, output_headers
to the executions INSERT/SELECT/UPSERT and row mapper. Required for
REGULAR mode where route-level payloads exist but no processor records.
Note: requires ALTER TABLE migration to add the new columns.
- Add @RequestBody(required=false) CommandAckRequest to ack endpoint for
receiving agent command results (backward compat with old agents)
- Record command results in agent event log via AgentEventService
- Add set-traced-processors to mapCommandType switch
- Inject AgentEventService dependency
Complete the group→application terminology rename in the agent
registry subsystem:
- AgentInfo: field group → application, all wither methods updated
- AgentRegistryService: findByGroup → findByApplication
- AgentInstanceResponse: field group → application (API response)
- AgentRegistrationRequest: field group → application (API request)
- JwtServiceImpl: parameter names group → application (JWT claim
string "group" preserved for token backward compatibility)
- All controllers, lifecycle monitor, command controller updated
- Integration tests: JSON request bodies "group" → "application"
- Frontend: schema.d.ts, openapi.json, agent queries, AgentHealth
RBAC group references (groups table, GroupAdminController, etc.)
are NOT affected — they are a separate domain concept.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The stats endpoint was calling statsForRoute(null, agentIds) when
only application was set — this filtered by route_id=null, returning
zero results. Now correctly routes to statsForApp/timeseriesForApp
which queries the stats_1m_app continuous aggregate by application_name.
Also reverts the group parameter alias workaround — the deployed
backend correctly accepts 'application'.
Three code paths now:
- No filters → stats_1m_all (global)
- application only → stats_1m_app (per-app)
- routeId (±application) → stats_1m_route (per-route)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The backend was renamed from group→application but Docker build cache
may serve old code. Accept 'group' as a fallback alias so the UI works
with both old and new backends. Applies to GET /search/executions,
/search/stats, and /search/stats/timeseries.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
PostgreSQL JDBC driver can't infer SQL type for java.time.Instant.
Convert from/to parameters to java.sql.Timestamp before binding.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
TimescaleDB add_continuous_aggregate_policy and add_compression_policy
cannot run inside a transaction block. Move all policy calls to V2
with flyway:executeInTransaction=false directive.
Also fix stats_1m_processor_detail: add WITH NO DATA and
materialized_only = false.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Consolidate V1-V7 Flyway migrations into single V1__init.sql with
all columns renamed from group_name to application_name. Requires
fresh database (wipe flyway_schema_history, all data).
- DB columns: executions.group_name → application_name,
processor_executions.group_name → application_name
- Continuous aggregates: all views updated to use application_name
- OpenSearch field: group_name → application_name in index/query
- All Java SQL strings updated to match new column names
- Delete V2-V7 migration files (folded into V1)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The execution-related "group" concept actually represents the
application name. Rename all Java fields, API parameters, and frontend
types from groupName→applicationName and group→application for clarity.
- Java records: ExecutionSummary, ExecutionDetail, ExecutionDocument,
ExecutionRecord, ProcessorRecord
- API params: SearchRequest.group→application, SearchController
@RequestParam group→application
- Services: IngestionService, DetailService, SearchIndexer, StatsStore
- Frontend: schema.d.ts, Dashboard, ExchangeDetail, RouteDetail,
executions query hooks
Database column names (group_name) and OpenSearch field names are
unchanged — only the API-facing Java/TS field names are renamed.
RBAC group references (groups table, GroupRepository, GroupsTab) are
a separate domain concept and are NOT affected by this change.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Added groupName field to ExecutionSummary Java record and OpenSearch mapper
- Dashboard stat cards use locale-formatted numbers (en-US)
- Added inspect column (↗) linking directly to exchange detail page
- Fixed duplicate React key warning from two columns sharing executionId key
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
