This is a Long-Term Support (LTS) maintenance release in the 5.9.x line, focused on bug fixes, security patches, and stability improvements. Review the deployment guidelines before you upgrade: one Helm value for Prometheus relabeling is renamed.
Release Highlight
FlowX.AI 5.9.7 is a maintenance release for the 5.9.x LTS:
- Prometheus metrics: the
ServiceMonitorresources of the FlowX Helm charts select the management Service and scrape its metrics path, so Prometheus collects service metrics - Key fixes: reusable functions resolve when an overriding process runs as an embedded subprocess, and an error boundary event moves the token on during asynchronous advancing
- Audit index retention settings, more Kafka connection options in the
flowx-aiHelm chart, and security patches. Every change is listed in the full changelog
Upgrade notes
Upgrade compatibility: FlowX.AI 5.9.7 is compatible with 5.9.6. No database migration and no new Kafka topics.Zero-downtime upgrade: supported, after the steps below. See Upgrade within an LTS family.
- Prometheus relabeling: the Helm charts no longer read
metrics.serviceMonitor.relabellings. If you set it, move the rules tometrics.serviceMonitor.metricRelabelings, applied before ingestion, or tometrics.serviceMonitor.relabelings, applied before the scrape. See Prometheus ServiceMonitor. - API description endpoints: the services no longer serve the Swagger UI or the OpenAPI description at
/swagger-uiand/v3/api-docs. Update any tooling or ingress rules that read them. - Text Parsing: the Text Parsing strategy of the document parser no longer runs OCR by default. Scanned pages that have no text layer return no text: use an OCR or LLM strategy for them.
- SpiceDB CRD with Argo CD or several releases per cluster: the SpiceDB CRD moved to its own Helm subchart with a new
spicedb-operator-crds.enabledvalue. With Argo CD, upgrade with the value left attruebefore you set it tofalseanywhere. See SpiceDB CRD.
Full changelog
New
- Audit: set how long audit data is kept, and roll audit indices over when a primary shard reaches a set size, with two new environment variables. Both are off by default, so the Audit service deletes nothing until you set a retention period. See Configuring Elasticsearch and Audit index retention.
- flowx-ai Helm chart: Kafka connections support SASL
PLAIN, for example a Confluent Cloud API key over TLS. TLS withouttls.existingSecrettrusts the CA bundle of the image. OAuth can use an identity provider for Kafka only, Keycloak or Microsoft Entra ID, with its own scope and SASL extensions. The AI services receive the same settings. See How do I connect to Kafka with SASL?.
Changed
- Document parser: the Text Parsing strategy reads the text layer only and no longer runs OCR by default. Text parsing of DOCX files no longer fails while it looks for OCR language data.
- Helm charts: the SpiceDB CRD ships in its own
spicedb-operator-crdssubchart, which a newspicedb-operator-crds.enabledvalue turns off where another release or tool owns the CRD. Argo CD no longer prunes or deletes the CRD. See SpiceDB CRD. - Helm charts: Prometheus relabel rules are read from
metrics.serviceMonitor.relabelingsandmetrics.serviceMonitor.metricRelabelings. The misspelledmetrics.serviceMonitor.relabellingskey is no longer read. See Prometheus ServiceMonitor.
Fixed
- Helm charts: the
ServiceMonitorof each service selects the management Service and scrapes/manage/actuator/prometheuson themngtport, so Prometheus collects metrics instead of finding no target or getting a 404. - Process engine: a process that overrides a library process and runs as an embedded subprocess finds its reusable functions, and its nodes run in the overriding application’s build. Business rules no longer fail because
functionsis not defined. - Process engine: an error boundary event that catches an error during asynchronous advancing moves the token to its outgoing path, instead of leaving the token on the task.
- Media Library: importing a ZIP file into the Global Media Library from Windows no longer fails with a parsing error.
- Error responses: an unexpected server error returns a generic message and an
errorIdinstead of the exception message, and the service logs the full error with the sameerrorId. When a business rule, reusable function, or gateway condition that you test from FlowX Designer fails, the script error comes back with status 422, so Designer still shows it. - Stability fixes across project version merges, embedded subprocesses, the events gateway, the renderers, and the AI services.
Removed
- All services: the Swagger UI and the OpenAPI description,
/swagger-uiand/v3/api-docs, are no longer served.
Security
- All components: security patches and dependency upgrades to address known vulnerabilities.

