eventhub/docs/runtime-event-processing.md

266 lines
8.2 KiB
Markdown

# Runtime event processing
Runtime processing is now centered on **processing executions**. An execution loads canonical EventHub runtime events from selected sources, optionally partitions the event set, then runs a configured processing plan made of common EPL and/or Java modules.
The preferred endpoint is:
```http
POST /api/eventhub/runtime-processing/executions
```
List available plans:
```http
GET /api/eventhub/runtime-processing/executions/plans
```
The older endpoint is still available for compatibility:
```http
POST /api/eventhub/runtime-processing/event-processing
```
but it should be treated as a legacy profile adapter. New clients should use `processingPlanKey`, not `profileKey`.
## First predefined plan
The first predefined plan is:
```text
driver-working-time-v1
```
Legacy alias:
```text
tachograph-driver-esper-v1
```
This plan runs the source-neutral driver working-time chain over canonical runtime events. Tachograph file/database data is only one possible source of those canonical events. Its modules include:
```text
event-to-activity-intervals
event-to-vehicle-usage-intervals
vehicle-evidence-attachment
support-evidence-normalization
driving-derived-projections
```
The important design point is that runtime processing is not tachograph-specific. Tachograph is now treated as a source/compatibility layer; the working-time plan consumes canonical driver activity, vehicle usage, and support-evidence events from any supported source.
## Example execution
```json
{
"processingPlanKey": "driver-working-time-v1",
"sourceSelection": {
"tenantKey": "default",
"driverKey": "12:12345678901234",
"occurredFrom": "2026-05-01T00:00:00Z",
"occurredTo": "2026-05-31T23:59:59Z",
"sourceInputs": [
{
"sourceFamily": "TACHOGRAPH_FILE_SESSION",
"eventBackend": "SOURCE_DB",
"sessionIds": [
"11111111-1111-1111-1111-111111111111",
"22222222-2222-2222-2222-222222222222"
]
},
{
"sourceFamily": "TACHOGRAPH_DB",
"eventBackend": "EVENTHUB_DB"
},
{
"sourceFamily": "YELLOWFOX_DB",
"eventBackend": "SOURCE_DB"
}
],
"expandVehicleEvents": true
},
"partitioning": {
"strategy": "DRIVER",
"includeAllPartitions": true,
"attachVehicleEvidence": true,
"vehicleEvidencePaddingMinutes": 15,
"includeDebug": true
},
"parameters": {
"significantDrivingMinutes": 3,
"minimumRestPeriodMinutes": 720,
"includePartitionDebug": true,
"includePartitionMetadata": true,
"includePartitionModuleResults": true,
"includeExecutionModuleResults": true,
"includeSupportEvidenceNormalization": true
}
}
```
`sourceFamilies + eventBackend` is still accepted as a legacy shorthand when every selected family uses the same backend. Use `sourceInputs` when one runtime request must mix direct-source and EventHub-backed inputs.
## Conceptual flow
```text
source selection
-> runtime event loaders
-> canonical EventHub events
-> partitioning
-> vehicle evidence attachment
-> support evidence normalization
-> event-to-interval EPL modules
-> driver working-time derived projection EPL/Java modules
-> partition result map
```
## Why processing plans instead of profiles
A plan describes **which modules should run over a runtime event set**. This matches the EventHub goal better than domain-oriented profiles.
Future plans can reuse the same event-loading infrastructure:
```text
vehicle-trip-detection-v1
vehicle-stop-detection-v1
driver-settlement-v1
mixed-driver-vehicle-correlation-v1
telematics-poi-clustering-v1
```
Each plan may use different partitioning and modules, but the source loading and canonical event model remain common.
## Compatibility
Legacy profile endpoint:
```http
POST /api/eventhub/runtime-processing/event-processing
```
with:
```json
{
"profileKey": "tachograph-driver-esper-v1",
"scope": {},
"partitioning": {},
"parameters": {}
}
```
internally delegates to:
```text
processingPlanKey = driver-working-time-v1
```
## Source-neutral driver working-time modules
The former tachograph-named processing artifacts are now represented by source-neutral driver working-time names:
```text
runtime-driver-event-interval-preprocessor.epl
driver-working-time-derived-projections.epl
DriverWorkingTimeProcessingResultDto
DriverWorkingTimeProcessingCore
RuntimeDriverWorkingTimeScopeProcessingService
```
Legacy tachograph names are kept as compatibility adapters for existing file-session endpoints and Postman requests. New runtime code should use the `driver-working-time-*` classes/resources and the `driver-working-time-v1` processing plan.
## Module execution results
`/api/eventhub/runtime-processing/executions` now exposes module execution metadata explicitly.
Top-level execution response includes:
```text
moduleResults
```
This map contains one entry per executed module. The top-level entries are intentionally sanitized: they expose status, warnings, and metadata without duplicating the full partition output payload.
Each partition result also includes:
```text
partitionResults[*].moduleResults
```
For `driver-working-time-v1`, the partition-level module results currently expose:
```text
vehicle-evidence-attachment
support-evidence-normalization
driving-derived-projections
```
The first implementation uses wrapper modules around the existing validated driver-working-time scope service. Logical phase modules such as `event-to-activity-intervals`, `event-to-vehicle-usage-intervals`, and `vehicle-usage-merge` are registered and executed as delegated modules so they can later be split into true standalone EPL/Java modules without changing the external processing-plan contract.
## First-class EPL modules
Runtime processing now supports modules implemented directly with Esper EPL through the common `RuntimeEplModuleExecutor`.
The first source-neutral EPL modules are:
| Module key | Engine | EPL resource | Output statement |
|---|---|---|---|
| `event-to-activity-intervals` | `EPL` | `esper/runtime-driver-activity-intervals.epl` | `driverActivityIntervals` |
| `event-to-vehicle-usage-intervals` | `EPL` | `esper/runtime-driver-vehicle-usage-intervals.epl` | `driverVehicleUsageIntervals` |
These modules operate on canonical EventHub runtime events, not on tachograph-specific source rows. They are currently used as first-class phase modules in `driver-working-time-v1`; the final `driving-derived-projections` module remains a compatibility adapter over the validated working-time projection service until the remaining projection stages are split into direct EPL modules.
## Response shaping
`driver-working-time-v1` supports a small set of optional request parameters to trim diagnostic payloads from `/api/eventhub/runtime-processing/executions` and the legacy `/api/eventhub/runtime-processing/event-processing` adapter.
All of these default to `true` to preserve current behavior:
```text
includeExecutionModuleResults
includePartitionMetadata
includePartitionModuleResults
includeSupportEvidenceNormalization
includePartitionDebug
```
Meaning:
- `includeExecutionModuleResults`: include top-level `moduleResults` on `/executions`
- `includePartitionMetadata`: include `partitionResults[*].metadata`
- `includePartitionModuleResults`: include `partitionResults[*].moduleResults`
- `includeSupportEvidenceNormalization`: include `supportEvidenceNormalization` in each partition result payload
- `includePartitionDebug`: include `partitionDebug` in each partition result payload
Example slim response request:
```json
{
"processingPlanKey": "driver-working-time-v1",
"sourceSelection": {
"tenantKey": "default",
"driverKey": "12:12345678901234",
"occurredFrom": "2026-05-01T00:00:00Z",
"occurredTo": "2026-05-31T23:59:59Z",
"sourceInputs": [
{
"sourceFamily": "TACHOGRAPH_FILE_SESSION",
"eventBackend": "SOURCE_DB",
"sessionIds": [
"11111111-1111-1111-1111-111111111111"
]
}
]
},
"partitioning": {
"strategy": "DRIVER"
},
"parameters": {
"includeExecutionModuleResults": false,
"includePartitionMetadata": false,
"includePartitionModuleResults": false,
"includeSupportEvidenceNormalization": false,
"includePartitionDebug": false
}
}
```