AI Agent News/stories/google-cloud-adds-object-enrichments-to-extract-agent-context-from-cloud-storage.md

Infrastructure

Google Cloud adds object enrichments to extract agent context from Cloud Storage

Google Cloud has launched object enrichments in preview, a managed Cloud Storage feature that runs AI models over stored images and PDFs and writes the results as metadata on each object. Google says enriching 10 million images or PDFs costs $700 at its published pricing.

Source: cloud.google.com/blog/products/storage-data-transfer/cost-effective-ai-context-extraction-from-unstructured-gcs-data

Object enrichments uses AI models to pull contextual metadata out of Cloud Storage objects and saves it in place as object context, a structured key-value store attached to each object. The feature gives customers access to the image models Google uses internally for services such as Google Photos, including object detection and image labeling, along with a PDF classification model. Google says its image annotations cost 75% less than similar services from other cloud providers.

Setup takes three steps: define the buckets and processing location, choose one or more of the five first-party models for images and PDFs, and grant the service IAM permissions to access the buckets. The service handles model inference, response parsing, retries and throttling, and writes results to each object's context. Object context moves with the object during copy, move and replication, and is removed when the object is deleted.

Developers can filter object listings by context key-value pairs, set IAM controls over who can read or change context, and flow context into BigQuery through Storage Insights datasets. Gemini-based processing for tasks such as summarization, transcription and translation is coming later, and Google asks interested customers to contact storage-ai@google.com for access. The Cloud Storage MCP server can give agents IAM-scoped access to objects and their context.

Why it matters if you build agents

Agent developers can precompute metadata across large object stores instead of sending raw images or files to a model at query time, which Google says avoids expensive token use and added delay. Because context stays attached to its bytes, teams no longer need a separate metadata store or cleanup jobs for copied or deleted objects. Since the feature is in preview, confirm pricing and access before relying on it in production.

Get stories like this every weekday

One short email with the day's agent tooling news.

Subscribe