Repository navigation
DOC-7126: Add a scripted LangCache interactive demo - #4159
Conversation
A browser-only demo for the Iris launch, in the same style and with the same food delivery sample data as the Context Retriever and Agent Memory demos. Readers send a help assistant's questions in the order they choose and watch hits, misses, and stored entries; replay new questions at a similarity threshold they set, to see the trade-off between hits and wrong answers; and compare a restaurant attribute against no attributes, including deleting a restaurant's entries by attribute. Request and response shapes, status codes, and defaults follow the LangCache OpenAPI spec and the create-service page. Similarity scores, answers, and timings are illustrative. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
🧠 Redis MemoryFound 7 related items from repository history (2 new this commit):
Memory updated at c8e82ae |
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 7f7cc41. Configure here.
The hit-ratio wrap-up only appeared when the step was drawn with no animation phase, but the last animation frame always has one, so readers saw it only after leaving the tab and coming back. Every in-flight frame returns earlier in questionNotes, so the wrap-up no longer checks the phase. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Adds upstream's browser-only LangCache interactive demo page. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

A browser-only demo for the Iris launch, in the same style and with the same food delivery sample data as the Context Retriever and Agent Memory demos. Readers send a help assistant's questions in the order they choose and watch hits, misses, and stored entries; replay new questions at a similarity threshold they set, to see the trade-off between hits and wrong answers; and compare a restaurant attribute against no attributes, including deleting a restaurant's entries by attribute.
Request and response shapes, status codes, and defaults follow the LangCache OpenAPI spec and the create-service page. Similarity scores, answers, and timings are illustrative.
Note
Low Risk
Documentation and static client-side assets only; no production services, auth, or data paths change.
Overview
Adds a browser-only LangCache interactive demo for the Iris docs, aligned with the Context Retriever demo (shared
context-retriever-demo.css, same food-delivery sample customers/restaurants).A new
interactive-demo.mdpage embeds the{{< langcache-demo >}}shortcode, which loads version-hashed CSS/JS. The scriptedlangcache-demo.jswalks readers through three tabs: choosing help-assistant questions and watching cache hits/misses and API-shaped search/store flows; replaying new prompts with an adjustable similarity threshold; and a side-by-side restaurant attribute lane (scoped search/delete vs cross-restaurant wrong answers and stale menu data). Request/response shapes mirror the LangCache API; scores and timings are illustrative only.The LangCache
_index.mdgains a link to the new demo page.Reviewed by Cursor Bugbot for commit c8e82ae. Bugbot is set up for automated code reviews on this repo. Configure here.