Support Custom Metadata Labels on Vertex AI Submissions

[yu-iskw]

Checked

• I searched existing ideas and did not find a similar one
• I added a very descriptive title
• I've clearly described the feature request and motivation for it

Feature request

I’d like to be able to attach arbitrary metadata labels to each Vertex AI prediction call through the LangChainJS VertexAI wrapper. Currently, the VertexAI class in langchainjs does not expose a way to pass custom labels, which makes it hard to correlate usage, billing, or downstream processing back to specific application contexts.

Relevant links:

• Vertex AI custom labels guide: https://cloud.google.com/vertex-ai/generative-ai/docs/multimodal/add-labels-to-api-calls

Motivation

I’m always frustrated when I can’t distinguish or filter Vertex AI calls by their origin in my analytics or billing dashboards. For example:

• Multi-tenant apps need to tag calls by tenant ID.
• A/B testing scenarios need labels for experiment IDs.
• Cost allocation requires grouping usage by project, environment (dev/staging/prod), or feature flag.

Without first-class support for metadata labels in LangChainJS, I end up writing custom HTTP clients or patching the library at runtime, which is brittle and hard to maintain.

Proposal (If applicable)

No response

[yu-iskw]

The pull request was merged and released.

#8207