Knowledge Memory Provider

Overview

The KnowledgeMemoryProvider handles "Shared Knowledge" at the server (Guild) or Channel level. This is used for storing community-specific information such as: - Server Rules: Custom instructions for the bot within a specific guild. - Inside Jokes/Memes: Information that applies to everyone in a channel. - Local Facts: Information about a specific community or project.

Levels of Knowledge

The provider fetches knowledge in a hierarchical manner:

1. Guild Knowledge: Broad instructions or facts applicable to the entire server.
2. Channel Knowledge: Specific context applicable only to the current channel.

Implementation Details

Hierarchical Fetching

The get(guild_id, channel_id) method fetches both levels simultaneously. Channel-level knowledge usually overrides or supplements Guild-level knowledge in the final prompt.

Cache Management

• TTL Cache: Uses a standard time-to-live cache (default 5 minutes).
• Invalidation: Cache can be invalidated by administrative commands when knowledge is updated.

Usage in Prompting

Knowledge is typically injected early in the system prompt to set the "ground rules" for the conversation within that specific environment.

---

By separating User, Episodic, and Knowledge memory, the bot can distinguish between "what I know about you" vs "what I know about this place".