feat: add vision events SDK and CLI support

[rvirani1]

Description

• Adds SDK support for the vision events API: write single/batch events, query with pagination, list use cases, and upload images
• Adds roboflow vision-events CLI command group with subcommands: write, write-batch, query, use-cases, upload-image
• Thin-client design with no client-side validation, so new event types and fields work without an SDK release
• 21 unit tests covering all endpoints, pagination, passthrough integrity, filter forwarding, and error handling

Fixes ENT-1082

Type of change

• New feature (non-breaking change which adds functionality)
• This change requires a documentation update

How has this change been tested, please provide a testcase or example of how you tested the change?

Locally. 21 new unit tests + full suite (396 tests), ruff, mypy all pass.

Will the change affect Universe? If so was this change tested in universe?

N/A

Any specific deployment considerations

None

Docs

• Docs updated? What were the changes:

CLI commands and SDK methods need to be documented in roboflow-product-docs (docs.roboflow.com).

[renzon]

Manual E2E Test Results — Vision Events SDK & CLI

Tested against local dev server (localapi.roboflow.one) with workspace renzo-sandbox and a manually created use case (jEYfHQY5Ib1EjW4imyBT).

SDK Results (all pass ✅)

Method	Result
write_vision_event	Created event, created: true, no warnings
write_vision_events_batch	Created 5 events in one call (minor eventData warnings on non-quality_check types)
upload_vision_event_image	Got sourceId + GCS URL; with --name and --metadata
Write event with uploaded image	Created event referencing sourceId from upload
query_vision_events	Found 7 events across 4 types
query_vision_events (filtered)	Correctly filtered to 1 safety_alert event
get_vision_event_metadata_schema	Discovered 3 custom fields: camera_id, line, shift
list_vision_event_use_cases	Listed the use case with correct id, name, status

CLI Results (all pass ✅)

Command	Exit Code	Result
vision-events write '{...}'	0	Created event
vision-events write-batch '[...]'	0	Created 2 events
vision-events query <use-case> (text mode)	0	Listed all 10 events with types
vision-events query <use-case> -t quality_check -l 2 --json	0	Filtered + paginated correctly (hasMore: true, nextCursor present)
vision-events use-cases --json	0	Listed the use case
vision-events metadata-schema <use-case> --json	0	Showed 4 discovered fields
vision-events upload-image <path> --json	0	Uploaded real JPEG, got sourceId

Result on UI:

Caveat: Use Case (Solution) must be created manually

To run these tests, I had to create the use case (Solution) through the Vision Events UI (/renzo-sandbox/vision-events → "+ Create Use Case"). The SDK and CLI don't expose Solution CRUD — the create/update/archive/unarchive endpoints exist on the server (/query/vision-events/{workspace}/solutions) and are used by the frontend, but they're not included in this PR.

Question: Is this intentional — i.e., use cases are meant to be managed only through the UI? Or should we also add Solution CRUD to the SDK/CLI (create-use-case, update-use-case, archive-use-case, etc.) so users can fully manage the lifecycle programmatically?