assistant: teach the agent about Search admin

Bring the LLM agent's prompts and tools current with the new /search
section and arcadia-search admin sidecar:

- New tools in admin-tools.ts:
  - list_search_corpora: enumerate tenants + corpora with build status,
    so the agent can pick a real corpus instead of guessing.
  - rebuild_search_corpus(tenant, corpus): isWrite=true, surfaces a
    confirm card. Use after uploads or when results look stale.
- search_kb description updated: names docs / operator-tools / files
  explicitly, and points at list_search_corpora when unsure.
- ARCADIA_KNOWLEDGE: adds search-corpus terminology, /search route,
  and a one-liner pointer to the three new tools.
- assistant.tsx UI_CONTROL_PREFACE: nav-search added, full Search
  page action catalog (search-refresh / -restart / -new-tenant /
  -new-corpus, corpora-search, per-row corpus-{t}-{c}-{rebuild,edit,
  delete,actions}, tenant-{id}-delete, dialog form fields). Recipe
  for the manual rebuild path, plus a note steering the agent to
  the rebuild_search_corpus tool by default.
- search.tsx publishes a "search" surface to admin-context with
  tenants + corpora summary, so the agent gets live state without
  needing a tool call when /search is mounted.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

app/lib/admin-tools.ts

@@ -30,6 +30,7 @@ import { listRoles } from "~/lib/arcadia/roles"
 import { revokeUserApiKey } from "~/lib/arcadia/api-keys"
 import { createRAGClient } from "@crema/lexical-rag-ui"
 import { BLOCK_INDEX, getBlockSchema } from "~/lib/block-schemas"
+import { searchAdmin, SearchAdminError } from "~/lib/search-admin"
 
 // Lazy singleton — first tool call fetches /docs-index.json, subsequent
 // calls reuse the parsed MiniSearch instance.
@@ -732,7 +733,7 @@ const TOOLS: ToolDef[] = [
   {
     name: "search_kb",
     description:
-      "Lexical (BM25) search over the arcadia-search Tantivy backend. Use for the LARGER, server-hosted knowledge corpora — the same arcadia docs the browser RAG serves are indexed here as `corpus=docs` for parity, and additional corpora (uploaded files, runbooks, etc.) will land here as they're added. Returns chunks with snippets + chunk_ids that can be passed to `read_chunk` to expand. Prefer this over `search_docs` (browser) when you need richer hits or when the user is asking about content that wouldn't be in the bundled docs (e.g. uploaded files).",
+      "Lexical (BM25) search over the arcadia-search Tantivy backend. Returns chunks with snippets + chunk_ids that can be passed to `read_chunk` to expand. Prefer this over `search_docs` (browser) when you need richer hits or when the content wouldn't be in the bundled docs.\n\nKnown corpora on the platform-admin tenant:\n- `docs` — arcadia-app architecture/ops docs (same as the browser RAG, server-hosted for parity).\n- `operator-tools` — arcadia-search + arcadia-admin documentation (admin sidecar, deploy script, search admin UI, MULTI_TENANT, RAG, AI_FIRST, LIBS, LLM_PROXY_CONTRACT).\n- `files` — markdown/text files uploaded by tenant users via arcadia-app.\n\nIf you're not sure what's available, call `list_search_corpora` first. Operators can add new corpora via the `/search` route.",
     parameters: {
       type: "object",
       properties: {
@@ -740,7 +741,7 @@ const TOOLS: ToolDef[] = [
         corpus: {
           type: "string",
           description:
-            "Which indexed corpus to search. `docs` is the parity corpus (arcadia documentation). New corpora are added by the operator.",
+            "Which indexed corpus to search. See list_search_corpora for the live set; common values: `docs`, `operator-tools`, `files`.",
         },
         limit: {
           type: "integer",
@@ -796,6 +797,84 @@ const TOOLS: ToolDef[] = [
       return result
     },
   },
+  {
+    name: "list_search_corpora",
+    description:
+      "Enumerate the corpora currently configured on the arcadia-search admin sidecar. Returns each tenant's corpora with build status (indexed?, num_docs). Call this when you don't know what corpora exist before invoking `search_kb`, or when the user asks what knowledge is available. Requires the search admin token to be configured.",
+    parameters: {
+      type: "object",
+      properties: {},
+      additionalProperties: false,
+    },
+    isWrite: false,
+    run: async () => {
+      try {
+        const tenantsRes = await searchAdmin.listTenants()
+        const tenants = await Promise.all(
+          tenantsRes.tenants.map(async (t) => {
+            try {
+              const c = await searchAdmin.listCorpora(t.id)
+              return {
+                tenant: t.id,
+                corpora: c.corpora.map((cc) => ({
+                  corpus: cc.corpus,
+                  indexed: cc.indexed,
+                  num_docs: cc.num_docs,
+                })),
+              }
+            } catch {
+              return { tenant: t.id, corpora: [] }
+            }
+          }),
+        )
+        return { tenants }
+      } catch (err) {
+        if (err instanceof SearchAdminError) {
+          return {
+            error: `search-admin ${err.status}: ${err.message}`,
+            hint: "VITE_ARCADIA_SEARCH_ADMIN_TOKEN may be unset, or the sidecar (default :7801) may be down.",
+          }
+        }
+        throw err
+      }
+    },
+  },
+  {
+    name: "rebuild_search_corpus",
+    description:
+      "Trigger a synchronous rebuild of one corpus on arcadia-search. Use when the operator says the index is stale, after they've uploaded new files, or when search_kb returned suspiciously few/old hits. Returns chunk_count and built_at on success. The operator confirms before the rebuild runs (rebuilds can take seconds–minutes depending on corpus size).",
+    parameters: {
+      type: "object",
+      properties: {
+        tenant: {
+          type: "string",
+          description: "Search tenant id (e.g. `platform-admin`). See list_search_corpora for available tenants.",
+        },
+        corpus: {
+          type: "string",
+          description: "Corpus name within that tenant (e.g. `docs`, `operator-tools`, `files`).",
+        },
+      },
+      required: ["tenant", "corpus"],
+      additionalProperties: false,
+    },
+    isWrite: true,
+    run: async (args) => {
+      const tenant = typeof args.tenant === "string" ? args.tenant.trim() : ""
+      const corpus = typeof args.corpus === "string" ? args.corpus.trim() : ""
+      if (!tenant || !corpus) {
+        throw new Error("rebuild_search_corpus requires { tenant, corpus }")
+      }
+      try {
+        return await searchAdmin.rebuild(tenant, corpus)
+      } catch (err) {
+        if (err instanceof SearchAdminError) {
+          return { error: `search-admin ${err.status}: ${err.message}` }
+        }
+        throw err
+      }
+    },
+  },
   {
     name: "get_block_schema",
     description: `Fetch the full JSON schema + example for a rich-output block kind so you can emit it correctly in your reply. Call this the first time in a thread that you intend to render a particular kind. Available kinds: ${Object.entries(

app/lib/arcadia-knowledge.ts

@@ -16,6 +16,7 @@ Core entities and how they relate:
 - **Audit log entry** — append-only record of who did what. \`actor_type\` is one of: \`user\`, \`platform_admin\`, \`api_key\`, \`system\`. Per-tenant and platform-wide entries coexist.
 - **Feature flag** — boolean / variant gate. Platform-wide default + per-tenant override.
 - **Storage / billing config / SSO IdP / inbound webhook / API quota / data retention policy / approval workflow / announcement** — per-tenant or platform-level configurations the operator can manage.
+- **Search corpus** — a Tantivy index over a set of source documents, served by the arcadia-search service. Each corpus belongs to a search tenant (a separate id space from platform tenants — typically \`platform-admin\` for the operator's own knowledge). The operator manages corpora at \`/search\`: create/edit configuration JSON, rebuild on demand, restart the service. Built-ins on \`platform-admin\`: \`docs\` (arcadia architecture), \`operator-tools\` (arcadia-search + arcadia-admin docs), \`files\` (uploaded markdown/text files).
 
 Tenant lifecycle (status field):
 
@@ -31,6 +32,7 @@ Things to keep in mind when assisting:
 - The operator can impersonate tenant users for debugging (POST /api/v1/admin/impersonate/:user_id) — surface this when they ask "why can't user X log in".
 - Quotas / rate cards / billing config errors usually surface as 402/403 from /api/v1 endpoints — diagnose by checking the tenant's billing-config and api-metering quotas.
 - The reference Phoenix app lives at \`reference/arcadia-app/\` in the workspace; its OpenAPI spec is at /api/openapi (sync via \`node ../lib-arcadia-client/scripts/sync-spec.mjs\`).
+- Search admin (arcadia-search) is a separate service. Manage tenants/corpora at \`/search\`. Use \`list_search_corpora\` if you don't know what's indexed; \`rebuild_search_corpus\` after uploads or when results look stale; \`search_kb\` / \`read_chunk\` to query.
 
 When the user asks something that maps to a tool, call it. When they ask about a concept, explain it from this primer in plain language. Write tools (suspend_tenant, activate_tenant) prompt the operator with an inline confirm card before they actually run — you do not need to ask in prose first; just call the tool and the user will see the confirmation UI. If the user denies a write, do not retry it; ask what they'd like to do differently.
 

app/routes/assistant.tsx

@@ -53,7 +53,7 @@ Rules:
 
 Known action ids across the app (use these even if not in "Available actions" — the page may not be mounted yet):
 
-Sidebar / nav: nav-overview, nav-tenants, nav-resources, nav-activity, nav-assistant, nav-library, nav-settings, sidebar-toggle, mobile-nav-toggle
+Sidebar / nav: nav-overview, nav-tenants, nav-resources, nav-activity, nav-search, nav-assistant, nav-library, nav-settings, sidebar-toggle, mobile-nav-toggle
 Appbar: appbar-search (input), appbar-scripts, appbar-font-size, appbar-surface, appbar-background, theme-toggle, appbar-notifications, appbar-avatar
 Account menu (after click appbar-avatar): avatar-profile (→ /profile), avatar-settings, avatar-help, avatar-signout
 Profile page: profile-avatar-upload, profile-avatar-remove, profile-name, profile-email, profile-title, profile-bio, profile-signature, profile-default-agent, profile-save, profile-revert, profile-reset
@@ -66,6 +66,7 @@ Assistant page: assistant-model, assistant-agent, assistant-thread, assistant-th
 Library page: library-search, library-open-<id>, library-copy-<id>, library-download-<id>, library-delete-<id>
 Resources page: resources-search, resources-new-name, resources-create, resources-status-<id>, resources-delete-<id>
 Tenants page: tenants-refresh, tenants-search (input), tenants-create. Per-row (use the tenant's slug — see the "tenants" surface in Admin context for available slugs): tenant-<slug>-actions (open the kebab first), tenant-<slug>-suspend, tenant-<slug>-activate, tenant-<slug>-deactivate. Recipe to suspend a tenant: click nav-tenants, wait_for tenants-refresh, click tenant-<slug>-actions, wait_for tenant-<slug>-suspend, click tenant-<slug>-suspend.
+Search page (/search — manage arcadia-search tenants and corpora): search-refresh, search-restart (with confirm), search-new-tenant, search-new-corpus, corpora-search (input). Per-tenant chip: tenant-<id>-delete. Per-corpus row (id is "<tenant>-<corpus>"): corpus-<tenant>-<corpus>-actions (kebab), corpus-<tenant>-<corpus>-rebuild, corpus-<tenant>-<corpus>-edit, corpus-<tenant>-<corpus>-delete. New-tenant dialog: tenant-form-id (input), tenant-form-cancel, tenant-form-save. New/edit-corpus dialog: corpus-form-tenant (select, only when creating), corpus-form-config (textarea, JSON), corpus-form-cancel, corpus-form-save. Recipe to rebuild a corpus: click nav-search, wait_for search-refresh, click corpus-<tenant>-<corpus>-actions, wait_for corpus-<tenant>-<corpus>-rebuild, click corpus-<tenant>-<corpus>-rebuild. (NOTE: prefer the \`rebuild_search_corpus\` tool over UI-driving for rebuilds — it's atomic and gives a structured result; UI-drive only when the user explicitly wants to see it happen.)
 Login page: login-email, login-password, login-submit
 Notifications popover: appbar-notifications (open), notif-mark-all-read, notif-clear, notif-open-<id>, notif-dismiss-<id>
 Create a notification (hidden bridge — always available, even when not visible): fill the four hidden inputs, then click the submit button. Recipe:
@@ -113,7 +114,7 @@ function buildAdminPreface(activeAgent: Agent | undefined, uiControl: boolean):
   const ctx = formatAdminContextForPrompt()
   const parts = [
     "You are the operator's assistant inside Arcadia Admin. Be precise and direct. You have native function tools attached to this conversation — call them whenever the user asks about live platform state (counts, statuses, listings, lookups). Never invent tenant slugs, user counts, or statuses; if you need data, call a tool.",
-    "Two retrieval surfaces exist for documentation/knowledge: `search_docs` (browser-side, BM25 over the bundled arcadia docs — fast, always available, small corpus) and `search_kb` (server-side, BM25 over arcadia-search — same docs as `corpus=docs` for parity, plus larger and additional corpora as the operator adds them). For questions about the bundled arcadia docs either is fine; prefer `search_kb` when you want richer hits or when the user is asking about content that wouldn't be in the bundled docs (uploaded files, tenant-specific knowledge). When `search_kb` returns a chunk_id you want to expand, call `read_chunk(chunk_id, corpus)`.",
+    "Two retrieval surfaces exist for documentation/knowledge: `search_docs` (browser-side, BM25 over the bundled arcadia docs — fast, always available, small corpus) and `search_kb` (server-side, BM25 over arcadia-search — `docs` (arcadia parity), `operator-tools` (arcadia-search + arcadia-admin admin docs), `files` (uploaded files), plus any custom corpora the operator adds via /search). For questions about the bundled arcadia docs either is fine; prefer `search_kb` for richer hits or for content outside the bundled docs (uploaded files, the admin tooling itself, tenant-specific knowledge). If unsure what corpora exist, call `list_search_corpora`. When `search_kb` returns a chunk_id you want to expand, call `read_chunk(chunk_id, corpus)`. When the operator says results look stale or after they've uploaded new files, call `rebuild_search_corpus(tenant, corpus)`.",
     RICH_OUTPUT_PREFACE,
     ARCADIA_KNOWLEDGE,
     persona,

app/routes/search.tsx

@@ -60,6 +60,7 @@ import {
 } from "~/lib/search-admin"
 import { pageTitle } from "~/lib/page-meta"
 import { useSession } from "~/lib/session"
+import { useRegisterAdminContext } from "~/lib/admin-context"
 
 export const meta = () => pageTitle("Search")
 
@@ -140,6 +141,24 @@ export default function SearchRoute() {
     return { indexed, docs }
   }, [corpora])
 
+  // Publish a snapshot to the assistant's admin context so the agent
+  // can answer "what corpora exist?" / "is the docs corpus indexed?"
+  // without having to call list_search_corpora.
+  const adminSurface = useMemo(
+    () => ({
+      endpoint: searchAdmin.baseUrl,
+      tenants: tenants.map((t) => ({ id: t.id, corpus_count: t.corpus_count })),
+      corpora: corpora.map((c) => ({
+        tenant: c.tenant,
+        corpus: c.corpus,
+        indexed: c.indexed,
+        num_docs: c.num_docs,
+      })),
+    }),
+    [tenants, corpora],
+  )
+  useRegisterAdminContext("search", adminSurface)
+
   const rebuild = useCallback(
     async (tenant: string, corpus: string) => {
       const id = `${tenant}/${corpus}`