Iris concepts - #4003
Iris concepts#4003mich-elle-luna wants to merge 2 commits into
Conversation
Adds a cross-cutting concepts page plus one per service (Agent Memory, LangCache, Context Retriever), each covering how a pattern developers already know from Redis needs to be rethought: fuzzy semantic-cache hits instead of exact-key equality, memory that writes itself in the background instead of app-driven session storage, and governed tool-calling instead of direct database access. Link the new pages from each service's landing page and from the top-level context-engine index. Cross-references use the render-hook plain-path link style introduced by DOC-7059, matching the rest of this tree rather than relref shortcodes.
Adds the cross-cutting concepts page as a fourth tile alongside the three services. Also fixes icon collisions: the per-service Concepts tiles reused an icon already assigned to another tile on the same page (ai-brain.svg). All Concepts tiles now use a distinct, consistent icon (ai-model.svg) instead.
🧠 Redis MemoryFound 5 related items from repository history:
Memory updated at d037049 |
| **What happens if I write directly to long-term memory instead of letting extraction do it?** | ||
| Both paths are supported. Use direct writes for bulk imports or external knowledge sources — anything that didn't originate in a conversation. Automatic extraction is for facts that emerge from session events themselves. | ||
|
|
||
| **Why does dedup sometimes skip a memory I expected to be created?** |
There was a problem hiding this comment.
I might hold this for a troubleshooting section for operators rather than concepts. Up to you though!
RAM doesn't track hit counts on similar docs, so if you keep this it should probably read something like:
Automatic extraction considers the conversation and existing memories when deciding what to create or update. If a new memory looks like it may be a duplicate, we won't store it. If your application needs to ensure a memory exists, you can create it directly rather than use extraction.
There was a problem hiding this comment.
This is a little more accurate because our extraction process is fairly sophisticated. An LLM tries to determine if very similar memories are actually dissimilar and should be tracked, rather than cleanly rejecting new paragraphs that are very similar to existing ones. Some of the details are encoded here: https://github.com/redislabsdev/iris/blob/38063355c6d2272ac76629945d75cab25fd7b700/memory/dataplane/internal/usecases/worker/promote_instruct_strategy.go#L145
|
|
||
| Redis Agent Memory splits memory the same way people do, and each kind maps to a specific tier in the service: | ||
|
|
||
| | Human memory | What it holds | Redis Agent Memory tier | |
There was a problem hiding this comment.
I'm wondering how best to explain the difference in LTM on this page. For example, I don't think we document it, but you can extract thread summaries as long-term memories. In that case, we keep an LTM that's an always-updated summary of a specific thread. That means there are semantic, episodic, summary, and custom memory types possible, which is complicated enough that we might want to go high level here and then expand more on those in an LTM page. I defer to your judgment!
2 participants
Add four new concepts documents for Iris, RAM, Langcache and CR.
Note
Low Risk
Documentation-only changes to Hugo markdown; no application code, APIs, or infrastructure behavior.
Overview
Adds four new “Concepts” docs for Redis Iris: a cross-service Redis Iris concepts page plus service-specific pages for Agent Memory, LangCache, and Context Retriever. They explain mental-model shifts (context as a budget, async memory extraction/dedup, semantic cache false positives and thresholds, governed tools vs raw SQL) with FAQs and links to quickstarts.
The context-engine hub and each service landing page now promote those pages: the top image-card grid expands from three to four columns with a Concepts card, and the Iris overview adds a short “read concepts before you build” callout for Redis-familiar readers.
Reviewed by Cursor Bugbot for commit d037049. Bugbot is set up for automated code reviews on this repo. Configure here.