Please select the area the issue is related to
AI Workspace
Please select the aspect the issue is related to
Aspect/API (API backends, definitions, contracts, interfaces, OpenAPI), Aspect/UI (Frontend layouts, components, styling)
Description
On an App LLM Proxy's Guardrails & Policies tab, clicking Add under Global Guardrails & Policies opens a catalog panel with a Categories filter (defaulting to AI). Under that default filter, Word Count Guardrail never appears — not by scrolling, and not by typing its name into the search box.
Root cause: the panel fetches the catalog with a hardcoded limit=40. The AI category currently has 43 policies, returned alphabetically, so the last 3 — Token Based Ratelimit, URL Guardrail, Word Count Guardrail — fall past the cutoff and are silently dropped. The search box only filters the already-fetched 40 items client-side, so it can never surface one that didn't make the batch, no matter what's typed.
Suggested fix:
- Wire the search box to the catalog API's own
search param (server-side), instead of filtering the already-fetched page client-side.
- Add pagination: load the next page via the API's existing
offset support as the user scrolls to the bottom of the list (infinite scroll), instead of a one-shot capped fetch — so no category is silently truncated regardless of how large it grows.
Steps to Reproduce
- On any App LLM Proxy, go to Guardrails & Policies → Add (under Global Guardrails & Policies).
- Leave Categories on the default AI filter.
- Search
Word Count, or scroll the full list — no result; the list ends at "Time-Based Model Routing".
- Switch Categories to Guardrails instead — Word Count Guardrail now appears.
Severity Level of the Issue
Severity/Critical (Core functionality is broken but there is a workaround. Need urgent attention)
Environment Details (with versions)
- AI Workspace console at
ai-workspace.bijira.dev
- Observed 2026-09-09
Please select the area the issue is related to
AI Workspace
Please select the aspect the issue is related to
Aspect/API (API backends, definitions, contracts, interfaces, OpenAPI), Aspect/UI (Frontend layouts, components, styling)
Description
On an App LLM Proxy's Guardrails & Policies tab, clicking Add under Global Guardrails & Policies opens a catalog panel with a Categories filter (defaulting to AI). Under that default filter, Word Count Guardrail never appears — not by scrolling, and not by typing its name into the search box.
Root cause: the panel fetches the catalog with a hardcoded
limit=40. The AI category currently has 43 policies, returned alphabetically, so the last 3 — Token Based Ratelimit, URL Guardrail, Word Count Guardrail — fall past the cutoff and are silently dropped. The search box only filters the already-fetched 40 items client-side, so it can never surface one that didn't make the batch, no matter what's typed.Suggested fix:
searchparam (server-side), instead of filtering the already-fetched page client-side.offsetsupport as the user scrolls to the bottom of the list (infinite scroll), instead of a one-shot capped fetch — so no category is silently truncated regardless of how large it grows.Steps to Reproduce
Word Count, or scroll the full list — no result; the list ends at "Time-Based Model Routing".Severity Level of the Issue
Severity/Critical (Core functionality is broken but there is a workaround. Need urgent attention)
Environment Details (with versions)
ai-workspace.bijira.dev