fix(google): avoid double-counting cached tokens in input_token_details.text

[guoyangzhen (guoyangzhen)]

Problem

When using @langchain/google with Gemini models that support implicit caching, input_token_details.text includes all text tokens (including cached ones), while input_token_details.cache_read also counts the cached tokens. This causes LangSmith to double-count cached tokens when calculating costs.

LangSmith calculates cost as:

input_cost = (text × regular_price) + (cache_read × discounted_price)

Since text includes cached tokens, those tokens are charged at both the regular rate and the discounted cache rate.

Root Cause

In convertGeminiGenerateContentResponseToUsageMetadata, Gemini's promptTokensDetails[TEXT] modality count equals the total text tokens (including cached). The code sets both text and cache_read without adjusting for the overlap.

Fix

Subtract cache_read from text modality count so the values don't overlap, matching how cost is calculated in LangSmith.

Fixes #10339

[changeset-bot]

⚠️ No Changeset found

Latest commit: e0b7aaa

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

This PR includes no changesets

When changesets are added to this PR, you'll see the packages that this PR includes changesets for and the associated semver types

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR