v1.3.2

langflow-ai/langflowv1.3.2Jun 19, 2026by PawelPeczek-Roboflow

AI Summary

This release introduces a Switch Case flow-control block for more complex workflow branching, fixes runtime introspection issues on Jetson devices, and improves model loading capabilities. It also enhances error handling for video sources and adds usage persistence features.

Key Highlights

Add Switch Case flow-control block with per-case execution branches
Allow loading locally stored models in workflows
fix(runtime-introspection): fall back to L4T version for driver_version on Jetson
Set ERROR state and emit poison pill on video source failures
feat(usage): persist source_info into resource_details

New Features

Switch Case flow-control block
Local model loading in workflows
Enhanced error handling for video sources
Usage persistence feature
SAM3 interactive fast path restoration

Full Release Notes

## What's Changed
* Add Switch Case flow-control block with per-case execution branches by @patricknihranz in https://github.com/roboflow/inference/pull/2452
* fix(runtime-introspection): fall back to L4T version for driver_version on Jetson by @sberan in https://github.com/roboflow/inference/pull/2457
* Apply change to fix keypints visualisation after supervision 0.29.0 by @PawelPeczek-Roboflow in https://github.com/roboflow/inference/pull/2467
* [codex] Use model stat endpoint for inference-models auth by @hansent in https://github.com/roboflow/inference/pull/2414
* fix(docs): defer Kapa widget load to fix landing-page a11y audits by @Erol444 in https://github.com/roboflow/inference/pull/2469
* restore SAM3 interactive fast path for repeated clicks by @grzegorz-roboflow in https://github.com/roboflow/inference/pull/2480
* Allow loading locally stored models in workflows by @grzegorz-roboflow in https://github.com/roboflow/inference/pull/2479
* [codex] Update Gemini workflow model IDs by @hansent in https://github.com/roboflow/inference/pull/2468
* Set ERROR state and emit poison pill on video source failures so recovery fires by @alexnorell in https://github.com/roboflow/inference/pull/2482
* [codex] Make OWLv2 cache CPU offload configurable by @hansent in https://github.com/roboflow/inference/pull/2487
* ENT-1355 Split TRT compile into compile, install, and optional register by @NVergunst-ROBO in https://github.com/roboflow/inference/pull/2459
* fix(workflows): keep consensus masks boolean by @kounelisagis in https://github.com/roboflow/inference/pull/2451
* fix(ent-1352): cast GLM-OCR processor inputs to model dtype on Jetson by @NVergunst-ROBO in https://github.com/roboflow/inference/pull/2456
* Fix dependencies issues - 19.06.2026 by @PawelPeczek-Roboflow in https://github.com/roboflow/inference/pull/2499
* Fix GHSA-rv25-cq3f-x5f2 by @grzegorz-roboflow in https://github.com/roboflow/inference/pull/2500
* feat(usage): persist source_info into resource_details by @bigbitbus in https://github.com/roboflow/inference/pull/2483
* Feature/req 186 workflow logging expose parent block name by @rafel-roboflow in https://github.com/roboflow/inference/pull/2450
* version -> 1.3.2 by @grzegorz-roboflow in https://github.com/roboflow/inference/pull/2505
* [codex] Move SAM video remote errors to runtime by @hansent in https://github.com/roboflow/inference/pull/2506
* Codeflash RF-DETR instance segmentation optimizations by @dkosowski87 in https://github.com/roboflow/inference/pull/2464
* Release 1.3.2 by @dkosowski87 in https://github.com/roboflow/inference/pull/2510


**Full Changelog**: https://github.com/roboflow/inference/compare/v1.3.1...v1.3.2