Customer Support Agent

This pattern stores durable preferences, issue history, and resolutions without stuffing every previous exchange back into the prompt.

Pattern

Create one session per support conversation.
Store durable facts with structured tags such as kind:preference and kind:resolution.
Search with tag_filters before each answer.
Soft-delete or export data when the customer asks.

import asyncio

from remembr import RemembrClient, TagFilter


async def main() -> None:
    async with RemembrClient(api_key="rk_demo") as client:
        session = await client.create_session(metadata={"channel": "support"})
        await client.store(
            "Customer wants Friday billing summaries.",
            role="user",
            session_id=session.session_id,
            tags=["kind:preference", "topic:billing"],
        )
        results = await client.search(
            "How should we send billing updates?",
            session_id=session.session_id,
            tag_filters=[TagFilter(key="kind", value="preference")],
        )
        print(results.total)


asyncio.run(main())

Pattern​

Pattern