For AI agents: The documentation index is at https://docs.digitalocean.com/llms.txt. Markdown versions of pages use the same URL with index.html.md in place of the HTML page (for example, append index.html.md to the directory path instead of opening the HTML document).

DigitalOcean Agent Data Services is in private preview for invited design partners. During private preview, you use Agent Data Services through the API and the Control Panel.

To join the private preview, request access.

Agent Data Services includes:

• Behavioral data flows: Assemble your event streams, logs, and agent traces into redacted sessions stored as tables you can query. Look up any session by its identifier to return the full interaction.
• Multimodal content data flows: Turn documents, images, audio, and video into extracted text, chunks, transcripts, descriptions, categories, moderation flags, and embeddings.
• Built-in data flow functions: Chain built-in functions, such as transcription, chunking, embedding, redaction, field extraction, session assembly, and aggregation, in the order you define.
• Test and promote workflow: Test a data flow on sample data and inspect the results before promoting it to run on every new arrival.
• Object storage ingestion: Connect a Spaces bucket or other S3-compatible bucket and process new and existing objects without migrating them.
• Managed inference: Data flow steps that need a model can use any serverless model on DigitalOcean Inference, with no GPUs or separate inference service to manage.
• Unified catalog: Each workspace has one catalog that holds content, embeddings, structured attributes, and access rules together.
• Open table format: Structured and semi-structured data is stored in Apache Iceberg tables.
• Lineage: Trace each result back to the raw data it came from and each transformation applied to it.
• Unified query layer: Combine semantic search, keyword search, and attribute filters in one request, retrieve records with time-windowed aggregations, and run read-only analytical queries. Send queries from the API, the Control Panel, or a client that connects with the query credentials for a workspace.
• Redaction before embedding: Redact or mask sensitive fields before they are embedded.
• IAM access control: Control who can manage and query each workspace with team roles, custom token scopes, and workspace-scoped tokens. See Agent Data Services Access Control.
• Data isolation: Each customer’s data is isolated at rest and while it is served, encrypted in transit and at rest, and never used to train DigitalOcean models.
• Horizontal scaling: Scale a data flow to up to 40 replicas to process queued work in parallel.

For how these features fit together, see How Agent Data Services Works.