OSS Agent

更新时间:
复制 MD 格式

OSS Agent lets you manage buckets, diagnose issues, run health checks, analyze costs, and search multimodal data in your buckets through natural language.

Important

OSS Agent is available by invitation only. Request access.

Overview

Built on the Tongyi large model, OSS Agent interprets natural language, executes multi-step API call chains, and visualizes results. It provides two core capabilities:

  • Intelligent O&M: Configure buckets, diagnose access issues, run health checks, and analyze costs with natural language.

  • Talk to Bucket: Powered by OSS Metaquery, search and query multimodal data (videos, images, audio, documents) in your buckets using natural language — no file paths or query syntax required.

Operations are tiered by risk. Low-risk reads run directly. Medium-risk writes require confirmation. High-risk operations such as deletions and lifecycle rule changes are not executed — the agent provides instructions for you to run them manually.

Prerequisites

  • You have activated OSS.

  • If you use a RAM user, ensure it has the required permissions on the target bucket.

Use OSS Agent

  1. Log in to the OSS console and go to the OSS Agent chat page.

  2. On first visit, read and agree to the User Notice.

  3. Enter a request in the input box, or select a question from the Intelligent O&M Expert, Q&A Assistant, or Talk to Multimodal Data cards.

The interface has three areas:

  • Left sidebar: Create chats, select a region and bucket, view history, and manage skills.

  • Chat area: Enter requests to interact with OSS Agent. Use @ to reference a specific bucket.

  • Workspace: A collaborative area with built-in tools such as Create Bucket, Resource Plan Recommendation, and Error Code Diagnosis.

Intelligent O&M

Multi-step operations

OSS Agent breaks multi-step requests into an ordered API call chain and displays each result in real time. Steps involving costs or risks require confirmation.

Example: Create a bucket named sample-bucket in the China (Shanghai) region, enable server-side encryption and versioning, and configure a policy to allow only internal network access.

Agent execution chain: ① Create a bucket in the China (Shanghai) region → ② Enable server-side encryption (AES256) → ③ Enable versioning

(Enabled) → ④ Configure a bucket policy to restrict access to internal networks (VPC) only.

Note

For parameter-sensitive operations such as creating a bucket, the agent populates the console form in the workspace for review. The operation runs only after you confirm.

Access anomaly diagnosis

Provide an inaccessible object link or error message. The agent runs a multi-step diagnosis, identifies the root cause, and recommends a fix.

Example: I can't access this image. Can you tell me why? https://examplebucket.oss-cn-hangzhou.aliyuncs.com/exampleobject?x-oss-additional-headers=host&x-oss-credential=LTAI********************%2F20241203%2Fcn-hangzhou%2Foss%2Faliyun_v4_request&x-oss-date=20241203T034420Z&x-oss-expires=3600&x-oss-signature=70c542eaf652ac291c0c343d63ac24ede41c0526661d9d4c63c0906a**********&x-oss-signature-version=OSS4-HMAC-SHA256

The agent identifies the root cause (for example, "This V4 presigned URL had a 3,600-second validity period. The signature has expired.") and provides resolution steps.

Bucket health check

Run a health check on a bucket. The agent evaluates security, data protection, cost management, and basic information, then reports each item with optimization suggestions. After you confirm, the agent applies the corrections.

Example: Run a health check on my bucket @samplebucketname.

The agent runs multiple API calls to cover all four dimensions.

Dimension

Check item

Security Check

Bucket ACL, server-side encryption, block public access, hotlink protection, CORS

Data Protection

versioning, delete markers, multipart upload fragment cleanup

Cost Management

lifecycle rule, access logging, access monitoring

Basic Information

Bucket overview (region, storage class, redundancy strategy, object count, storage usage)

Cost analysis

The agent retrieves billing data, analyzes storage, traffic, and request costs, and generates a per-bucket cost report.

Example: Analyze my costs.

The agent breaks down costs by storage, traffic, and requests.

Talk to Bucket

How it works

Talk to Bucket uses OSS Metaquery data indexing, combining AI content awareness and vector search in three layers:

  1. Content awareness layer: A multimodal model analyzes videos, images, audio, and documents to produce semantic descriptions and structured tags.

  2. Vector indexing layer: Converts the parsed content into high-dimensional vectors to build a vector index.

  3. Semantic matching layer: Converts your query into an embedding, performs multi-path search and re-ranking, and returns the top-K results.

OSS Agent guides you through enabling data indexing, vectorizing data, and running natural-language searches.

Data indexing

Toggle the Talk to Bucket switch in the floating panel at the lower left of the input box. When enabled, messages automatically trigger multimodal data search and Q&A for the selected bucket. The agent checks whether Metaquery data indexing is enabled and guides you through setup if needed.

Important

Enabling Metaquery data indexing incurs fees for AI content awareness and semantic vector search (Data indexing fees). The agent displays a cost notice before proceeding. Confirm before continuing.

Vectorizing existing data takes time. Searches are unavailable until indexing completes. The agent shows the current indexing status.

Use cases

Image semantic search

Search images by describing their content instead of relying on tags or file names.

Example: @sample-bucket Find images of a volcanic eruption.

The agent returns matching images, each with its file path, size, resolution, tags, and a description.

Chat with multimodal data

Locate video clips by describing an event instead of watching each file.

Example: Check who started the fire. If you find the video, please send it to me.

The agent returns matching videos with path, duration, AI-generated scene description, object detection results, and a 15-minute presigned download URL.

Note

AI content awareness results are for reference only. Accuracy depends on factors such as video clarity and does not replace manual review.

Document semantic search

Search documents by content rather than file names. The agent returns matching documents and access links based on semantic matching.

Example: Find me documents about the internet platform project.

Quotas and limitations

Item

Description

Concurrent sessions

Up to two concurrent sessions per UID on a single device.

Session lifecycle

Sessions end automatically after one hour of inactivity or eight hours of total conversation time.

High-risk operation restrictions

The agent does not execute high-risk operations such as deletions (buckets, objects, configurations), lifecycle rule management, WORM, data replication, overwrites, and static website hosting. It provides console links for manual execution.

Billing

OSS Agent is in a free, invitation-only preview. Operations that use other OSS features, such as Metaquery data indexing, are billed at standard OSS rates (Billing overview). The agent notifies you before running any billable operation.