feat: deep-agent canvas, live observability, and multi-environment tooling

Self-hosted platform for building, testing, and shipping LangChain/LangGraph agents. Deep-agent sub-agents on the canvas, a live tracing/observability timeline, auto-provisioned built-in tools with import/export, per-environment tool variables, streamed evaluations, and per-user auth token forwarding.

apps/web/lib/models.ts

@@ -0,0 +1,46 @@
+"use client";
+/* Model picker data. The catalog (chat + embedding + reranker) is served by the backend
+   (GET /v1/models) from its canonical lists, so no model dropdown hardcodes options in the
+   frontend and the picker can only offer models the backend actually runs (and, for chat,
+   prices). See forge/model_catalog.py. */
+import { useEffect, useState } from "react";
+import { api, type EmbeddingModelInfo, type ModelCatalog, type ModelInfo, type RerankerModelInfo } from "./api";
+
+const EMPTY: ModelCatalog = { chat: [], embedding: [], reranker: [] };
+
+// Fetched once, shared across every picker. api.json() also de-dupes concurrent GETs, so even
+// a cold cache is a single round-trip.
+let _cache: ModelCatalog | null = null;
+
+function useModelCatalog(): ModelCatalog {
+  const [catalog, setCatalog] = useState<ModelCatalog>(() => _cache ?? EMPTY);
+  useEffect(() => {
+    if (_cache) return;
+    let alive = true;
+    api
+      .listModels()
+      .then((c) => {
+        _cache = c;
+        if (alive) setCatalog(c);
+      })
+      .catch(() => {
+        /* leave empty: selects still show the current value + any hardcoded default option */
+      });
+    return () => {
+      alive = false;
+    };
+  }, []);
+  return catalog;
+}
+
+export function useModels(): ModelInfo[] {
+  return useModelCatalog().chat;
+}
+
+export function useEmbeddingModels(): EmbeddingModelInfo[] {
+  return useModelCatalog().embedding;
+}
+
+export function useRerankerModels(): RerankerModelInfo[] {
+  return useModelCatalog().reranker;
+}