Iris concepts - #4003
Merged
Merged
Iris concepts#4003
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.
Contributor
Contributor
🧠 Redis MemoryFound 14 related items from repository history:
Memory updated at 1d84690 |
abrookins
reviewed
Sep 15, 2026
- Fix the dedup/extraction description: it's not a hit-count-based mechanism. Extraction weighs a new memory against existing ones using model judgment, not a simple similarity-threshold reject. - Note that the memory-types table is a simplified entry point, not the full picture — link to Memory types & extraction for the complete set of built-in (including automatic session summaries) and custom types.
dwdougherty
approved these changes
Sep 21, 2026
dwdougherty
left a comment
dwdougherty
left a comment
Collaborator
There was a problem hiding this comment.
Language LGTM, but I'm still a beginner when it comes to Iris. I'll put docs back on the review list.
Add a Mermaid diagram to each concepts page (cross-cutting architecture flow, Agent Memory's async extraction sequence, LangCache's threshold decision, Context Retriever's direct-access-vs-governed comparison) and to Data Integration's overview section, matching this repo's existing Mermaid conventions. Remove em dashes from the concepts pages' running prose per review feedback, replacing them with periods, colons, semicolons, or parentheses depending on what the sentence needs; colons specifically for description-list-style text (a term followed by its meaning).
Introduces a new render-hook + JS widget (context-map), following the existing decision-tree pattern, for a clickable flow diagram showing how a request draws on LangCache, Agent Memory, Context Retriever, and Data Integration as a layer rather than a fixed sequence. Includes scenario buttons that trace the cache-hit, memory, and business-data paths through the diagram. Also reorders the Agent Memory concepts page so its diagram appears before the memory-types table, and switches the LangCache concepts diagram to a horizontal layout to match its sibling pages. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
andy-stark-redis
left a comment
andy-stark-redis
left a comment
Contributor
There was a problem hiding this comment.
I think this is probably mergeable as it is, but I also think it might benefit from a bit of extra love, ideally.
Prose: rewrite marketing-toned headings and "this isn't X, it's Y" constructions into plain statements (Agent Memory), lead with what Context Retriever does instead of what it doesn't replace, lead with what semantic caching is before its failure mode (LangCache), reframe troubleshooting-style FAQ questions as conceptual ones, and split the context-map intro so the diagram instructions are clearly introduced as referring to the diagram. Also replace unqualified sentence-initial "it"/"this"/"that" with the noun they refer to across all four concepts pages, and align heading/table terminology (LangCache vs. semantic caching). Widget: shrink the context-map diagram and scenario buttons so they fit without horizontal scrolling, make the SVG responsive, and separate the shared detail panel into "Component" vs. "Scenario" variants with distinct styling. Add hover tooltips on nodes as a lightweight legend. Fix doc links pointing to the wrong URL when the site is deployed under a subpath (redis.io serves docs at /docs/latest/): the render hook now passes the site's actual base path to the JS via a data attribute, and the JS resolves every node/link URL against it instead of assuming the domain root. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
andy-stark-redis
approved these changes
Sep 23, 2026
andy-stark-redis
left a comment
andy-stark-redis
left a comment
Contributor
There was a problem hiding this comment.
Thanks for the changes @mich-elle-luna - LGTM :-)
Diagram: move the "MCP" label from Context Retriever -> Data Integration onto Agent -> Context Retriever, where MCP actually applies. Point Data Integration's edge back at Context Retriever (the data layer it queries) instead of Agent. Reword the write-back loopback to "App writes session event" since it never connects to LangCache and both writes require application action, not automatic continuation. Agent Memory: describe the session summary as derived session state returned separately from long-term memories, not as its own long-term memory type. LangCache: state the actual default similarity threshold (0.85) and recommended starting range (0.8-0.9) instead of implying there's no default at all. CSS: stop scaling the context-map SVG down on narrow viewports (max-width: 100%), which shrank labels below readable size; let the container's existing horizontal scroll handle overflow instead. Concepts overview: clarify that the page's prose covers the three request-time services, while the diagram includes Data Integration for a complete picture. Convert all remaining Markdown-path internal links across the four new concepts pages to relref shortcodes, per repo convention. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Route top/bottom-routed loopback edges into the target's side instead of dropping straight into its top/bottom edge, so the write-back from Response to Agent Memory no longer cuts through the LangCache diamond stacked above it in the same column. Move the "Getting Started with Redis Iris" tutorial link into Further reading, and point Next steps directly at the three services' own concepts pages instead. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Renames agent-memory/concepts.md to overview.md, with matching title/linkTitle changes, and updates the landing-page tile and the cross-cutting concepts page's links/text to match. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Diagonal edges into process nodes clipped against an ellipse instead of the rect the node actually renders as, hiding the arrowhead under the node. The viewBox also left no room for the 2px hover/selected stroke on edge-of-grid nodes, so it got clipped by default overflow. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes using high effort and found 1 potential issue.
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit 92bfddc. Configure here.
| ' L ' + start.x + ' ' + dipY + | ||
| ' L ' + approachX + ' ' + dipY + | ||
| ' L ' + approachX + ' ' + to.cy + | ||
| ' L ' + entryX + ' ' + to.cy; |
There was a problem hiding this comment.
Loopback path overlaps diagram edges
Medium Severity
drawLoopbackEdge always drops in at to.cx + LOOPBACK_SIDE_CLEAR and enters the target from the right. On the Iris request-flow diagram that lands the e9 return path on top of the Agent Memory → Model call edge and across the LangCache → Cached response edge, so the dashed loopback reads as extra junctions that are not in the YAML.
Additional Locations (2)
Reviewed by Cursor Bugbot for commit 92bfddc. Configure here.
Reframe the top-level and Agent Memory concepts pages so they teach
concepts directly instead of assuming prior Redis experience, with
"if you know Redis" comparisons demoted to asides. Split the top-level
page into a section: the mental-model overview stays at concepts/,
with the interactive request-flow diagram moved to its own child page
so the main page isn't also carrying full per-service architecture.
Add "Key terms" sections to Agent Memory, Context Retriever, and
LangCache concepts pages so each names the real terminology (session
event fields, entity/tool vocabulary, cache attributes and search
strategies) instead of staying purely conceptual. Add a modeling
tradeoff section to Context Retriever, matching LangCache's existing
threshold tradeoff section.
Rewrite passages across all four pages for Google developer
documentation style: split multi-clause sentences, convert passive
constructions to active voice, remove em dashes and fronted
subordinate clauses, and fix one directional-language reference
("in the diagram below").
Restyle the context-map widget's node boxes to match the site's
Mermaid theme (white fill, red border) instead of gray, for visual
consistency between the two diagram types used across these pages.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Add four new concepts documents for Iris, RAM, Langcache and CR.
Note
Low Risk
Documentation and front-end diagram UX only; no auth, APIs, or runtime product behavior changes.
Overview
Adds a concepts layer for Redis Iris: new pages explain how Iris works end-to-end, how a single request flows through LangCache / Agent Memory / Context Retriever, plus dedicated concept guides for LangCache, Agent Memory (overview), and Context Retriever. Hub pages now use four-column card grids with links into those concepts, plus small copy tweaks (TTL/RAG/MCP expansions) and a Data Integration mermaid diagram.
Introduces an interactive
context-mapshortcode for docs: YAML in fenced blocks is hidden, parsed client-side into an SVG flow diagram with clickable nodes, scenario path buttons, and a detail panel. Wiring includes a Hugo render hook,context-map.js, Tailwind/CSS styling, conditional script load inbaseof.html, andstatic/schemas/context-map.json. The Iris request-flow page is the first consumer.Reviewed by Cursor Bugbot for commit 1d84690. Bugbot is set up for automated code reviews on this repo. Configure here.