Changelog

The Dashboard RefreshAsync method called AllPlugins.Clear() before

fetching the remote plugin registry, so when the API call timed out

the entire catch block skipped populating locally installed plugins,

system metrics, and workspace activation state — leaving the UI stuck

showing "0 installed · 0 available" with empty metric cards.

Restructured RefreshAsync to:

• Fetch local installed versions first (never fails)
• Wrap the remote registry fetch in its own try/catch so timeouts

are non-fatal — locally installed plugins still appear

• Always load system metrics regardless of registry availability

Also increased PrivStackApiClient timeout from 15s to 30s since the

plugin registry endpoint can be slow.

Phase 2 of BuildIndexAsync already calls WikiLinkParser.ExtractContentFromEntity

on each entity from ReadList. The "load individually" block re-fetched entities

without content via individual SdkAction.Read calls, but the individual Read

returns the same JSON shape — ExtractContentFromEntity produces the same empty

result. This caused 21+ unnecessary contact.Read SDK calls at startup.

Removing the block eliminates the N+1 query pattern with no behavioral change,

since entities that genuinely lack parseable text fields won't gain content from

a re-fetch.

Adds BacklinkService.PreBuildIndexAsync() and calls it during the

deferred background services block in App.axaml.cs. This eliminates

the ~700ms cold-start delay when the user first clicks an item with

the Info Panel open, since the cross-plugin backlink index is already

warm by the time they interact.

The existing SemaphoreSlim guard in EnsureIndexBuiltAsync ensures

concurrent calls (from a user click racing the startup pre-build)

are safe. If the pre-build fails, the index falls back to lazy

initialization on first query (existing behavior).

ParseActionBlocks now falls back to ExtractBareIntentJson when no

[ACTION] wrapper tags are found, recovering intent JSON that the AI

emits without the required tags. The fallback only triggers when the

primary parser found nothing, preventing double-extraction.

ActionFormatHeader now explicitly warns against JSON document objects

in content/description text slots.

Visibility changes: ParseActionBlocks, ExtractBareIntentJson,

ParsedAction, and FlattenSlotValue are now internal (testable via

InternalsVisibleTo).

6 new xUnit tests cover bare JSON recovery, surrounding prose

stripping, non-intent JSON rejection, multiple intents, and

malformed JSON handling.

When the user asks a question while viewing an entity, the RAG query is

now augmented with semantic keywords extracted from the entity's JSON

(job_title, company_name, department, tags, contexts, etc.). This causes

the embedding search to surface semantically related items across plugins

without any changes to the RAG service or Rust embedding layer.