Back to list

Cómo Empezar con Agent Memory

CERTIRSS · Documentation
How to Start with Agent Memory

Summary

Agent Memory requires access to Oracle AI Database 23ai or later, with database version 23.4 or later, and Python 3.10 to 3.14. SearchStrategy.HYBRID requires Oracle AI Database 23ai Release Update 23.6 or later; uses managed and DBMS_HYBRID_VECTOR.SEARCH. SearchStrategy.KEYWORD does not create local columns and vector indexes. SearchStrategy.VECTOR requires Oracle AI Vector Search, including data type VECTOR and vector indexes.

Install the SDK with pip install "oracleagentmemory==26.8.0". On compatible platforms, pip gets pre-compiled binary wheels.

To create or re-create a managed warehouse, the owner of the scheme requires system privileges CREATE TABLE y CREATE PROCEDURE. An application connection that you use schema_owner access an existing warehouse and have to use SchemaPolicy.REQUIRE_EXISTING; Do not create and update the managed objects of owner.

By default, Oracle AI Database's managed schemes do not set a retention period for messages and memories. Restraint is determined by MemoryRetentionConfig or lifetime adjustments by registration. If the database user lacks CREATE JOB, the program configuration ends with a warning and expired or orphaned fragments are not physically removed until a dedicated user believes purge work.

Updating
RSS

Search at CertiTips

12 available courses.