ltcai 3.0.1 → 3.1.0

package/static/v3/js/core/api.22a41d42.js

@@ -0,0 +1,344 @@
+/* ============================================================================
+ * Lattice AI v3 — Integration adapter
+ *
+ * Every adapter call hits the real endpoint first (including /api/index/status,
+ * /api/graph, /api/search/hybrid, and /chat). If that endpoint is
+ * missing/unavailable, it returns an unavailable source with empty data so the
+ * UI can render a clear unavailable state without inventing counters or health.
+ *
+ * Return shape (never throws): { ok, status, data, source, error }
+ *   source: "live"        → returned by a real backend endpoint
+ *           "unavailable" → endpoint missing/down; no fake payload
+ * ========================================================================== */
+
+import { store } from "./store.34ebd5e6.js";
+
+const TIMEOUT_MS = 8000;
+const EMPTY_INDEX_STATUS = { generated_at: null, pipelines: {}, sources: [] };
+const EMPTY_GRAPH_STATS = { nodes: {}, edges: {}, total_nodes: 0, total_edges: 0 };
+const EMPTY_WORKSPACE_OS = { counts: {}, models: {} };
+const EMPTY_SYSINFO = { cpu_pct: null, ram_pct: null, gpu_mem_pct: null, gpu_mem_gb: null };
+const EMPTY_ADMIN = {
+  summary: { total_users: null, active_users: null, admin_users: null, total_messages: null },
+  users: [],
+  audit: { recent_events: [] },
+  security: {},
+  roles: { roles: [] },
+  policies: { policies: [] },
+  vpc: {},
+};
+
+async function raw(path, { method = "GET", body, headers } = {}) {
+  const ctrl = new AbortController();
+  const timer = setTimeout(() => ctrl.abort(), TIMEOUT_MS);
+  try {
+    const ws = store.get().workspaceId;
+    const res = await fetch(path, {
+      method,
+      credentials: "same-origin",
+      signal: ctrl.signal,
+      headers: {
+        "Accept": "application/json",
+        ...(body ? { "Content-Type": "application/json" } : {}),
+        ...(ws ? { "X-Workspace-Id": ws } : {}),
+        ...headers,
+      },
+      body: body ? JSON.stringify(body) : undefined,
+    });
+    let data = null;
+    const text = await res.text();
+    if (text) { try { data = JSON.parse(text); } catch { data = { raw: text }; } }
+    return { ok: res.ok, status: res.status, data };
+  } catch (err) {
+    return { ok: false, status: 0, data: null, error: err && err.name === "AbortError" ? "timeout" : String(err) };
+  } finally {
+    clearTimeout(timer);
+  }
+}
+
+function unavailableData(shape) {
+  const value = typeof shape === "function" ? shape() : shape;
+  if (Array.isArray(value)) return [];
+  if (value && typeof value === "object") return {};
+  return null;
+}
+
+/** Try the live endpoint; on any non-2xx/transport failure, return empty data. */
+async function withFallback(path, opts, shape) {
+  const res = await raw(path, opts);
+  if (res.ok && res.data && !res.data.raw) {
+    return { ...res, source: "live" };
+  }
+  return { ok: false, status: res.status, data: unavailableData(shape), source: "unavailable", error: res.error };
+}
+
+export const api = {
+  raw,
+
+  /** Generic GET with unavailable fallback. */
+  async get(path, shape = null) {
+    return withFallback(path, {}, shape);
+  },
+
+  /* ── Documented future surfaces ─────────────────────────────────────── */
+
+  /** GET /api/index/status — KG + Vector + Hybrid pipeline state. */
+  indexStatus() {
+    return withFallback("/api/index/status", {}, EMPTY_INDEX_STATUS);
+  },
+
+  /** POST /api/index/rebuild — rebuild the derived vector index (real run). */
+  rebuildIndex(opts = {}) {
+    return raw("/api/index/rebuild", { method: "POST", body: { full: false, include_nodes: true, include_chunks: true, ...opts } });
+  },
+
+  /** GET /api/graph — knowledge graph (nodes + edges). Falls back through the
+   *  current /knowledge-graph/graph route before reporting unavailable. */
+  async graph(params = {}) {
+    const qs = new URLSearchParams(params).toString();
+    const primary = await raw(`/api/graph${qs ? "?" + qs : ""}`);
+    if (primary.ok && primary.data && Array.isArray(primary.data.nodes)) {
+      return { ...primary, source: "live" };
+    }
+    const legacy = await raw("/knowledge-graph/graph");
+    if (legacy.ok && legacy.data && Array.isArray(legacy.data.nodes)) {
+      return { ...legacy, source: "live" };
+    }
+    return { ok: false, status: primary.status || legacy.status || 0, data: { nodes: [], edges: [] }, source: "unavailable", error: primary.error || legacy.error };
+  },
+
+  graphStats() {
+    return withFallback("/knowledge-graph/stats", {}, EMPTY_GRAPH_STATS);
+  },
+
+  /** POST /api/search/hybrid — fused KG + vector retrieval.
+   *  The backend returns `{ matches: [...] }` where each match carries
+   *  `source_scores: { keyword, vector, graph }`. Normalize that into the flat
+   *  result shape the view renders (title/path/snippet/score + per-signal). A
+   *  legacy `results` array is also accepted defensively. */
+  async hybridSearch(query, opts = {}) {
+    const res = await raw("/api/search/hybrid", { method: "POST", body: { query, ...opts } });
+    const live = res.ok && res.data
+      ? (Array.isArray(res.data.matches) ? res.data.matches
+        : Array.isArray(res.data.results) ? res.data.results
+        : null)
+      : null;
+    if (live) {
+      const items = live.map((m) => {
+        const ss = m.source_scores || {};
+        const meta = m.metadata || {};
+        return {
+          id: m.id || m.node_id,
+          title: m.title || m.id || "Untitled",
+          path: meta.path || meta.source || m.path || m.type || "",
+          snippet: m.snippet || m.summary || "",
+          score: typeof m.score === "number" ? m.score : 0,
+          vector: Number(ss.vector ?? m.vector) || 0,
+          lexical: Number(ss.keyword ?? m.lexical) || 0,
+          graph: Number(ss.graph ?? m.graph) || 0,
+        };
+      });
+      return { ok: true, status: res.status, data: items, source: "live", weights: res.data.weights || null };
+    }
+    return { ok: false, status: res.status, data: [], source: "unavailable", error: res.error };
+  },
+
+  /* ── Existing surfaces (used where helpful, all fallback-safe) ──────── */
+  workspaceOs() { return withFallback("/workspace/os", {}, EMPTY_WORKSPACE_OS); },
+  async models() {
+    const res = await raw("/models");
+    if (res.ok && res.data && !res.data.raw) {
+      const data = res.data;
+      const loadedIds = Array.isArray(data.loaded) ? data.loaded : [];
+      const recommended = Array.isArray(data.recommended) ? data.recommended.map((m) => ({
+        ...m,
+        name: m.name || m.display_name || m.id,
+        family: m.family || m.modality || "local",
+        state: loadedIds.includes(m.id) || data.current === m.id ? "loaded" : "available",
+      })) : [];
+      const loadedOnly = loadedIds
+        .filter((id) => !recommended.some((m) => m.id === id))
+        .map((id) => ({ id, name: id, family: "local", state: data.current === id ? "loaded" : "available" }));
+      return {
+        ok: true,
+        status: res.status,
+        source: "live",
+        data: { ...data, catalog: Array.isArray(data.catalog) ? data.catalog : [...recommended, ...loadedOnly] },
+      };
+    }
+    return { ok: false, status: res.status, data: { current: null, catalog: [] }, source: "unavailable", error: res.error };
+  },
+  loadModel(modelId, engine) {
+    return raw("/models/load", { method: "POST", body: { model_id: modelId, engine: engine || null } });
+  },
+  unloadModel(modelId) {
+    return raw(`/models/unload/${encodeURIComponent(modelId)}`, { method: "DELETE" });
+  },
+  sysinfo() { return withFallback("/local/sysinfo", {}, EMPTY_SYSINFO); },
+
+  adminSummary() { return withFallback("/admin/summary", {}, EMPTY_ADMIN.summary); },
+  adminUsers() { return withFallback("/admin/users", {}, EMPTY_ADMIN.users); },
+  adminAudit() { return withFallback("/admin/audit", {}, EMPTY_ADMIN.audit); },
+  adminSecurity() { return withFallback("/admin/security/overview", {}, EMPTY_ADMIN.security); },
+  adminRoles() { return withFallback("/admin/roles", {}, EMPTY_ADMIN.roles); },
+  adminPolicies() { return withFallback("/admin/policies", {}, EMPTY_ADMIN.policies); },
+  vpcStatus() { return withFallback("/vpc/status", {}, EMPTY_ADMIN.vpc); },
+
+  /* ── Embeddings (real backend: /api/embeddings/*) ───────────────────── */
+  /** GET /api/embeddings/status — active provider, grade, dimensions, last index. */
+  async embeddingsStatus() {
+    const res = await raw("/api/embeddings/status");
+    if (res.ok && res.data && res.data.provider) {
+      return { ok: true, status: res.status, data: res.data, source: "live" };
+    }
+    // No backend → report unavailable honestly (never fabricate a provider).
+    return {
+      ok: false, status: res.status, source: "unavailable",
+      data: { provider: null, active_provider: null, model: null,
+        model_id: null, dimensions: null, grade: "unavailable",
+        state: "unavailable", fell_back: false, health: { status: "unavailable", detail: "backend unavailable" },
+        last_indexed_at: null },
+    };
+  },
+  embeddingsProviders() { return withFallback("/api/embeddings/providers", {}, { active: "hash", providers: [] }); },
+
+  /* ── Agents (real backend: AgentRuntime /agents/api/runtime/*) ───────── */
+  /** GET /agents/api/runtime/status — roles, roster, runs, health from the runtime. */
+  async agentRuntime() {
+    const res = await raw("/agents/api/runtime/status");
+    if (res.ok && res.data && res.data.runtime && Array.isArray(res.data.agents)) {
+      return { ok: true, status: res.status, data: res.data, source: "live" };
+    }
+    // Fallback: unavailable roster, no fabricated run ledger.
+    return {
+      ok: false, status: res.status, source: "unavailable",
+      data: { runtime: { ready: false, total_runs: 0, active_runs: 0 },
+        health: { status: "unknown", checks: {} }, roles: [],
+        agents: [], runs: [] },
+    };
+  },
+  /** POST /agents/api/run — execute the multi-agent pipeline for a goal. */
+  runAgent(goal, roles) { return raw("/agents/api/run", { method: "POST", body: { goal, roles: roles || [] } }); },
+
+  /* ── Local computer memory (real backend: /workspace/computer-memory) ── */
+  computerMemory() { return raw("/workspace/computer-memory"); },
+  setComputerMemory(enabled) {
+    return raw("/workspace/computer-memory", { method: "POST", body: { enabled, consent: { approved: !!enabled } } });
+  },
+
+  /* ── Organization workspaces (real backend: /workspace/orgs) ────────── */
+  createOrg(name) { return raw("/workspace/orgs", { method: "POST", body: { name } }); },
+
+  /* ── Chat (real backend: SSE /chat + /history/*) ────────────────────── */
+
+  /** GET /history/conversations — conversation list. */
+  async chatHistory() {
+    const res = await raw("/history/conversations");
+    const list = res.ok && Array.isArray(res.data) ? res.data
+      : res.ok && res.data && Array.isArray(res.data.conversations) ? res.data.conversations
+      : null;
+    if (list) return { ok: true, status: res.status, data: list, source: "live" };
+    return { ok: false, status: res.status, data: [], source: "unavailable" };
+  },
+
+  /** GET /history/conversations/{id} — messages for one conversation. */
+  async conversation(id) {
+    const res = await raw(`/history/conversations/${encodeURIComponent(id)}`);
+    if (res.ok && res.data && Array.isArray(res.data.messages)) {
+      return { ok: true, status: res.status, data: res.data.messages, source: "live" };
+    }
+    return { ok: false, status: res.status, data: [], source: "unavailable" };
+  },
+
+  deleteConversation(id) {
+    return raw(`/history/conversations/${encodeURIComponent(id)}`, { method: "DELETE" });
+  },
+
+  /**
+   * POST /chat — streams the assistant reply over SSE.
+   * Parses `data: {chunk, model, trace}` events (terminator `[DONE]`), calling
+   * onChunk(delta, fullText) and onTrace(trace). If the endpoint is missing or
+ * not an event-stream, reports that chat is unavailable (no generated answer is
+ * invented). Resolves { source, text, trace, model, aborted }.
+   */
+  async streamChat(body, { onChunk, onTrace, signal } = {}) {
+    const ws = store.get().workspaceId;
+    let res;
+    try {
+      res = await fetch("/chat", {
+        method: "POST",
+        credentials: "same-origin",
+        signal,
+        headers: {
+          "Content-Type": "application/json",
+          "Accept": "text/event-stream",
+          ...(ws ? { "X-Workspace-Id": ws } : {}),
+        },
+        body: JSON.stringify({ stream: true, max_tokens: 2048, temperature: 0.2, ...body }),
+      });
+    } catch (err) {
+      if (err && err.name === "AbortError") return { source: "live", text: "", aborted: true };
+      return simulateChat(body, { onChunk, onTrace, signal });
+    }
+    const ctype = res.headers.get("content-type") || "";
+    if (!res.ok) {
+      let data = null;
+      try { data = await res.clone().json(); } catch {}
+      const detail = data && (data.detail || data.message || data.error);
+      const noModel = data && (data.error === "no_model_loaded" || /no .*model .*loaded/i.test(String(detail || "")));
+      if (noModel) {
+        return { source: "live", text: "", error: "no_model_loaded", errorMessage: String(detail || "No local model is loaded.") };
+      }
+      return simulateChat(body, { onChunk, onTrace, signal });
+    }
+    if (!res.body || !ctype.includes("text/event-stream")) {
+      return simulateChat(body, { onChunk, onTrace, signal });
+    }
+    const reader = res.body.getReader();
+    const decoder = new TextDecoder();
+    let buf = "", text = "", trace = null, model = null;
+    try {
+      for (;;) {
+        const { value, done } = await reader.read();
+        if (done) break;
+        buf += decoder.decode(value, { stream: true });
+        const parts = buf.split("\n\n");
+        buf = parts.pop();
+        for (const part of parts) {
+          const line = part.split("\n").find((l) => l.startsWith("data:"));
+          if (!line) continue;
+          const rawData = line.slice(5).trim();
+          if (rawData === "[DONE]") return { source: "live", text, trace, model };
+          let data; try { data = JSON.parse(rawData); } catch { continue; }
+          if (data.chunk) { text += data.chunk; onChunk && onChunk(data.chunk, text); }
+          if (data.model) model = data.model;
+          if (data.trace) { trace = data.trace; onTrace && onTrace(trace); }
+        }
+      }
+    } catch (err) {
+      if (err && err.name === "AbortError") return { source: "live", text, trace, model, aborted: true };
+      if (!text) return simulateChat(body, { onChunk, onTrace, signal });
+    }
+    return { source: "live", text, trace, model };
+  },
+};
+
+const sleep = (ms) => new Promise((r) => setTimeout(r, ms));
+
+/** Transparent unavailable stream — used only when no chat backend is available. */
+async function simulateChat(body, { onChunk, onTrace, signal } = {}) {
+  const q = (body && body.message) || "your question";
+  const reply =
+    `Chat is unavailable because the Lattice backend or active model is not reachable. ` +
+    `Start the server, load a model, and rebuild retrieval before sending “${q}”.`;
+  let text = "";
+  for (const word of reply.split(" ")) {
+    if (signal && signal.aborted) return { source: "unavailable", text, aborted: true };
+    const delta = (text ? " " : "") + word;
+    text += delta;
+    onChunk && onChunk(delta, text);
+    await sleep(16);
+  }
+  onTrace && onTrace(null);
+  return { source: "unavailable", text, trace: null };
+}

package/static/v3/js/core/api.js

@@ -2,21 +2,31 @@
  * Lattice AI v3 — Integration adapter
  *
  * Every adapter call hits the real endpoint first (including /api/index/status,
- * /api/graph, /api/search/hybrid, and /chat). If that
- * endpoint is missing/unavailable, it degrades to a clearly-labeled SAMPLE
- * payload from fixtures.js and reports `source: "placeholder"` so the UI can
- * badge it. No backend logic is implemented here — only transport + graceful
- * fallback, which keeps the v3 frontend resilient during local setup.
+ * /api/graph, /api/search/hybrid, and /chat). If that endpoint is
+ * missing/unavailable, it returns an unavailable source with empty data so the
+ * UI can render a clear unavailable state without inventing counters or health.
  *
  * Return shape (never throws): { ok, status, data, source, error }
  *   source: "live"        → returned by a real backend endpoint
- *           "placeholder" → fixture fallback (backend not yet available)
+ *           "unavailable" → endpoint missing/down; no fake payload
  * ========================================================================== */
 
 import { store } from "./store.js";
-import * as fx from "./fixtures.js";
 
 const TIMEOUT_MS = 8000;
+const EMPTY_INDEX_STATUS = { generated_at: null, pipelines: {}, sources: [] };
+const EMPTY_GRAPH_STATS = { nodes: {}, edges: {}, total_nodes: 0, total_edges: 0 };
+const EMPTY_WORKSPACE_OS = { counts: {}, models: {} };
+const EMPTY_SYSINFO = { cpu_pct: null, ram_pct: null, gpu_mem_pct: null, gpu_mem_gb: null };
+const EMPTY_ADMIN = {
+  summary: { total_users: null, active_users: null, admin_users: null, total_messages: null },
+  users: [],
+  audit: { recent_events: [] },
+  security: {},
+  roles: { roles: [] },
+  policies: { policies: [] },
+  vpc: {},
+};
 
 async function raw(path, { method = "GET", body, headers } = {}) {
   const ctrl = new AbortController();
@@ -46,28 +56,35 @@ async function raw(path, { method = "GET", body, headers } = {}) {
   }
 }
 
-/** Try the live endpoint; on any non-2xx/transport failure, use the fixture. */
-async function withFallback(path, opts, fixture) {
+function unavailableData(shape) {
+  const value = typeof shape === "function" ? shape() : shape;
+  if (Array.isArray(value)) return [];
+  if (value && typeof value === "object") return {};
+  return null;
+}
+
+/** Try the live endpoint; on any non-2xx/transport failure, return empty data. */
+async function withFallback(path, opts, shape) {
   const res = await raw(path, opts);
   if (res.ok && res.data && !res.data.raw) {
     return { ...res, source: "live" };
   }
-  return { ok: true, status: res.status, data: typeof fixture === "function" ? fixture() : fixture, source: "placeholder", error: res.error };
+  return { ok: false, status: res.status, data: unavailableData(shape), source: "unavailable", error: res.error };
 }
 
 export const api = {
   raw,
 
-  /** Generic GET with fixture fallback. */
-  async get(path, fixture = null) {
-    return withFallback(path, {}, fixture);
+  /** Generic GET with unavailable fallback. */
+  async get(path, shape = null) {
+    return withFallback(path, {}, shape);
   },
 
   /* ── Documented future surfaces ─────────────────────────────────────── */
 
   /** GET /api/index/status — KG + Vector + Hybrid pipeline state. */
   indexStatus() {
-    return withFallback("/api/index/status", {}, fx.INDEX_STATUS);
+    return withFallback("/api/index/status", {}, EMPTY_INDEX_STATUS);
   },
 
   /** POST /api/index/rebuild — rebuild the derived vector index (real run). */
@@ -76,7 +93,7 @@ export const api = {
   },
 
   /** GET /api/graph — knowledge graph (nodes + edges). Falls back through the
-   *  current /knowledge-graph/graph route before the fixture. */
+   *  current /knowledge-graph/graph route before reporting unavailable. */
   async graph(params = {}) {
     const qs = new URLSearchParams(params).toString();
     const primary = await raw(`/api/graph${qs ? "?" + qs : ""}`);
@@ -87,11 +104,11 @@ export const api = {
     if (legacy.ok && legacy.data && Array.isArray(legacy.data.nodes)) {
       return { ...legacy, source: "live" };
     }
-    return { ok: true, status: 200, data: fx.GRAPH, source: "placeholder" };
+    return { ok: false, status: primary.status || legacy.status || 0, data: { nodes: [], edges: [] }, source: "unavailable", error: primary.error || legacy.error };
   },
 
   graphStats() {
-    return withFallback("/knowledge-graph/stats", {}, fx.GRAPH_STATS);
+    return withFallback("/knowledge-graph/stats", {}, EMPTY_GRAPH_STATS);
   },
 
   /** POST /api/search/hybrid — fused KG + vector retrieval.
@@ -123,11 +140,11 @@ export const api = {
       });
       return { ok: true, status: res.status, data: items, source: "live", weights: res.data.weights || null };
     }
-    return { ok: true, status: res.status, data: fx.hybridResults(query), source: "placeholder", error: res.error };
+    return { ok: false, status: res.status, data: [], source: "unavailable", error: res.error };
   },
 
   /* ── Existing surfaces (used where helpful, all fallback-safe) ──────── */
-  workspaceOs() { return withFallback("/workspace/os", {}, fx.WORKSPACE_OS); },
+  workspaceOs() { return withFallback("/workspace/os", {}, EMPTY_WORKSPACE_OS); },
   async models() {
     const res = await raw("/models");
     if (res.ok && res.data && !res.data.raw) {
@@ -149,17 +166,23 @@ export const api = {
         data: { ...data, catalog: Array.isArray(data.catalog) ? data.catalog : [...recommended, ...loadedOnly] },
       };
     }
-    return { ok: true, status: res.status, data: fx.MODELS, source: "placeholder", error: res.error };
+    return { ok: false, status: res.status, data: { current: null, catalog: [] }, source: "unavailable", error: res.error };
   },
-  sysinfo() { return withFallback("/local/sysinfo", {}, fx.SYSINFO); },
+  loadModel(modelId, engine) {
+    return raw("/models/load", { method: "POST", body: { model_id: modelId, engine: engine || null } });
+  },
+  unloadModel(modelId) {
+    return raw(`/models/unload/${encodeURIComponent(modelId)}`, { method: "DELETE" });
+  },
+  sysinfo() { return withFallback("/local/sysinfo", {}, EMPTY_SYSINFO); },
 
-  adminSummary() { return withFallback("/admin/summary", {}, fx.ADMIN.summary); },
-  adminUsers() { return withFallback("/admin/users", {}, fx.ADMIN.users); },
-  adminAudit() { return withFallback("/admin/audit", {}, { recent_events: fx.ADMIN.audit }); },
-  adminSecurity() { return withFallback("/admin/security/overview", {}, fx.ADMIN.security); },
-  adminRoles() { return withFallback("/admin/roles", {}, { roles: fx.ADMIN.roles }); },
-  adminPolicies() { return withFallback("/admin/policies", {}, { policies: fx.ADMIN.policies }); },
-  vpcStatus() { return withFallback("/vpc/status", {}, fx.ADMIN.vpc); },
+  adminSummary() { return withFallback("/admin/summary", {}, EMPTY_ADMIN.summary); },
+  adminUsers() { return withFallback("/admin/users", {}, EMPTY_ADMIN.users); },
+  adminAudit() { return withFallback("/admin/audit", {}, EMPTY_ADMIN.audit); },
+  adminSecurity() { return withFallback("/admin/security/overview", {}, EMPTY_ADMIN.security); },
+  adminRoles() { return withFallback("/admin/roles", {}, EMPTY_ADMIN.roles); },
+  adminPolicies() { return withFallback("/admin/policies", {}, EMPTY_ADMIN.policies); },
+  vpcStatus() { return withFallback("/vpc/status", {}, EMPTY_ADMIN.vpc); },
 
   /* ── Embeddings (real backend: /api/embeddings/*) ───────────────────── */
   /** GET /api/embeddings/status — active provider, grade, dimensions, last index. */
@@ -170,10 +193,10 @@ export const api = {
     }
     // No backend → report unavailable honestly (never fabricate a provider).
     return {
-      ok: true, status: res.status, source: "placeholder",
-      data: { provider: "hash", active_provider: "hash", model: "lattice-local-hash-v1",
-        model_id: "lattice-local-hash-v1:384", dimensions: 384, grade: "fallback",
-        state: "fallback", fell_back: false, health: { status: "unknown", detail: "backend unavailable" },
+      ok: false, status: res.status, source: "unavailable",
+      data: { provider: null, active_provider: null, model: null,
+        model_id: null, dimensions: null, grade: "unavailable",
+        state: "unavailable", fell_back: false, health: { status: "unavailable", detail: "backend unavailable" },
         last_indexed_at: null },
     };
   },
@@ -186,12 +209,12 @@ export const api = {
     if (res.ok && res.data && res.data.runtime && Array.isArray(res.data.agents)) {
       return { ok: true, status: res.status, data: res.data, source: "live" };
     }
-    // Fallback: clearly-badged sample roster, no fabricated run ledger.
+    // Fallback: unavailable roster, no fabricated run ledger.
     return {
-      ok: true, status: res.status, source: "placeholder",
+      ok: false, status: res.status, source: "unavailable",
       data: { runtime: { ready: false, total_runs: 0, active_runs: 0 },
         health: { status: "unknown", checks: {} }, roles: [],
-        agents: fx.AGENTS.map((a) => ({ ...a, last_status: null, last_at: null })), runs: [] },
+        agents: [], runs: [] },
     };
   },
   /** POST /agents/api/run — execute the multi-agent pipeline for a goal. */
@@ -215,7 +238,7 @@ export const api = {
       : res.ok && res.data && Array.isArray(res.data.conversations) ? res.data.conversations
       : null;
     if (list) return { ok: true, status: res.status, data: list, source: "live" };
-    return { ok: true, status: res.status, data: fx.CHAT.conversations, source: "placeholder" };
+    return { ok: false, status: res.status, data: [], source: "unavailable" };
   },
 
   /** GET /history/conversations/{id} — messages for one conversation. */
@@ -224,8 +247,7 @@ export const api = {
     if (res.ok && res.data && Array.isArray(res.data.messages)) {
       return { ok: true, status: res.status, data: res.data.messages, source: "live" };
     }
-    const sample = (fx.CHAT.conversations.find((c) => c.id === id) || {}).messages || [];
-    return { ok: true, status: res.status, data: sample, source: "placeholder" };
+    return { ok: false, status: res.status, data: [], source: "unavailable" };
   },
 
   deleteConversation(id) {
@@ -236,8 +258,8 @@ export const api = {
    * POST /chat — streams the assistant reply over SSE.
    * Parses `data: {chunk, model, trace}` events (terminator `[DONE]`), calling
    * onChunk(delta, fullText) and onTrace(trace). If the endpoint is missing or
-   * not an event-stream, degrades to a clearly-labeled SAMPLE stream (no real
-   * generation is invented). Resolves { source, text, trace, model, aborted }.
+ * not an event-stream, reports that chat is unavailable (no generated answer is
+ * invented). Resolves { source, text, trace, model, aborted }.
    */
   async streamChat(body, { onChunk, onTrace, signal } = {}) {
     const ws = store.get().workspaceId;
@@ -303,25 +325,20 @@ export const api = {
 
 const sleep = (ms) => new Promise((r) => setTimeout(r, ms));
 
-/** Transparent SAMPLE stream — used only when no chat backend is available. */
+/** Transparent unavailable stream — used only when no chat backend is available. */
 async function simulateChat(body, { onChunk, onTrace, signal } = {}) {
   const q = (body && body.message) || "your question";
   const reply =
-    `This is a sample response. Connect a local model and the v3 retrieval backend ` +
-    `(/api/search/hybrid, /api/graph) to ground answers to “${q}” in your workspace. ` +
-    `The context panel shows the knowledge-graph entities, vector matches, and indexed ` +
-    `files that would be used to answer — this preview does not run a model.`;
+    `Chat is unavailable because the Lattice backend or active model is not reachable. ` +
+    `Start the server, load a model, and rebuild retrieval before sending “${q}”.`;
   let text = "";
   for (const word of reply.split(" ")) {
-    if (signal && signal.aborted) return { source: "placeholder", text, aborted: true };
+    if (signal && signal.aborted) return { source: "unavailable", text, aborted: true };
     const delta = (text ? " " : "") + word;
     text += delta;
     onChunk && onChunk(delta, text);
     await sleep(16);
   }
-  const trace = fx.sampleTrace(q);
-  onTrace && onTrace(trace);
-  return { source: "placeholder", text, trace };
+  onTrace && onTrace(null);
+  return { source: "unavailable", text, trace: null };
 }
-
-export { fx };