How an assistant is configured
A high-level walk-through of what an administrator does in the console. Every step is a form: no programming is needed.
- 1
Add your sources
Create each source you want the assistant to read. You can add as many of each kind as you need.
- Documents: create a document collection, upload files, and let the platform process them so they can be searched. New files are processed automatically.
- Databases: enter the connection details and choose which tables may be queried.
- Logs: point to your OpenSearch index.
- Web services: supply the address of the service's OpenAPI document and choose the operations to allow.
- 2
Group them into knowledge domains
A knowledge domain bundles related sources under a business name, such as "Retail banking". The same source can belong to several domains.
- 3
Decide what each agent may use
An agent can be limited to one knowledge domain or allowed to search across all of them. This is how you control which questions it can answer.
- 4
Assign people and roles
Users, roles and tenants are managed in the identity console. People sign in once and see only what their role allows.
- 5
Approve and go live
Changes are proposed by a maker and approved by a different checker before they take effect. The change log shows who changed what and when.
- 6
Test before you share
Each source has a test page. For example, the document test shows the passages retrieved for a question, with the file name, page and score, and no answer is generated.
See it for yourself
Open the console and follow along with the demo setup.