feat(2.0): Phase 1 — Engine + Routing (Herzstueck)

Backend-Services: fit/caps/sources (portiert), discover (live HF + Fit + Caps + ranked recommendation), llama-swap write/register + groups (Ko- Residenz swap:false), LiteLLM-Gateway-Config + gateway-Service (model:auto + Fallbacks). Router: discover/fit/register/groups/routing; health zeigt gateway_reachable. Frontend: Modelle&Routing mit Caps-Chips, Fit-Badges, Discover-Tab (live), Routing-View. Lokal verifiziert: Backend-Smoke (alle Endpunkte) + Frontend-Build + Browser (Shell, Discover, Caps/Fit). Box-Verifikation offen. Docs: README + docs/STATUS.md (Phasen-Tracker + Resume-Guide). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-25 08:01:12 +02:00
parent 1b421e30f9
commit cff3f0b1a8
25 changed files with 1253 additions and 235 deletions
@@ -6,8 +6,8 @@
    <meta name="theme-color" content="#0d1117" />
    <link rel="manifest" href="/manifest.webmanifest" />
    <title>Mission Control 2.0</title>
-    <script type="module" crossorigin src="/assets/index-CP2tEA8E.js"></script>
-    <link rel="stylesheet" crossorigin href="/assets/index-CGVxdi48.css">
+    <script type="module" crossorigin src="/assets/index-BeAucPZJ.js"></script>
+    <link rel="stylesheet" crossorigin href="/assets/index-NuZDhqZy.css">
  </head>
  <body>
    <div id="root"></div>
@@ -3,6 +3,7 @@ import { Command as CommandIcon } from "lucide-react"
 import { NAV, type ViewId } from "@/nav"
 import { CommandPalette } from "@/components/CommandPalette"
 import { ModelsView } from "@/views/ModelsView"
+import { RoutingView } from "@/views/RoutingView"
 import { Placeholder } from "@/views/Placeholder"
 import { api, type Health } from "@/lib/api"
 import { cn } from "@/lib/utils"
@@ -88,7 +89,10 @@ export default function App() {

        <main className="flex-1 overflow-y-auto p-6">
          {view === "models" && <ModelsView />}
-          {view !== "models" && <Placeholder title={active.label} hint={active.hint} />}
+          {view === "routing" && <RoutingView />}
+          {view !== "models" && view !== "routing" && (
+            <Placeholder title={active.label} hint={active.hint} />
+          )}
        </main>
      </div>
    </div>
@@ -0,0 +1,27 @@
+import type { Capabilities } from "@/lib/api"
+
+function Chip({ children, tone = "muted" }: { children: React.ReactNode; tone?: "muted" | "primary" | "warn" }) {
+  const tones = {
+    muted: "bg-muted text-muted-foreground",
+    primary: "bg-primary/15 text-primary",
+    warn: "bg-amber-500/15 text-amber-500",
+  }
+  return (
+    <span className={`rounded px-1.5 py-0.5 text-[11px] font-medium ${tones[tone]}`}>{children}</span>
+  )
+}
+
+export function CapsChips({ caps }: { caps?: Capabilities }) {
+  if (!caps) return null
+  return (
+    <span className="inline-flex flex-wrap gap-1">
+      {caps.coder && <Chip>💻 Code</Chip>}
+      {caps.vision && <Chip>👁 Bild</Chip>}
+      {caps.reasoning && <Chip>🧠 Reason</Chip>}
+      {caps.moe && <Chip tone="primary">🧩 MoE{caps.active_b ? `·${caps.active_b}b` : ""}</Chip>}
+      {caps.tools === "yes" && <Chip tone="primary">🛠 Tools</Chip>}
+      {caps.tools === "likely" && <Chip tone="warn">🛠 Tools?</Chip>}
+      {caps.embedding && <Chip>🔢 Embed</Chip>}
+    </span>
+  )
+}
@@ -9,6 +9,26 @@ export async function api<T = unknown>(path: string, init?: RequestInit): Promis
  return res.json() as Promise<T>
 }

+export interface Capabilities {
+  moe: boolean
+  active_b: number | null
+  tools: "yes" | "likely" | "no"
+  vision: boolean
+  coder: boolean
+  reasoning: boolean
+  embedding: boolean
+  ctx: number | null
+  params_b: number | null
+  arch: string | null
+}
+
+export interface Fit {
+  level: "perfect" | "marginal" | "too_tight"
+  text: string
+  req_gb: number
+  tps: number
+}
+
 export interface ModelInfo {
  name: string
  role: string | null
@@ -17,14 +37,52 @@ export interface ModelInfo {
  ctx: number | null
  ttl: number | null
  cmd: string
+  gguf_path: string
  filename: string
  quant: string
  size_bytes: number | null
  incomplete: boolean
+  capabilities: Capabilities
+}
+
+export interface DiscoverModel {
+  name: string
+  author: string
+  repo: string
+  role: string
+  params_b: number
+  quant: string
+  tags: string[]
+  downloads: number
+  fit: Fit
+  optimal_ctx: number
+  caps: Capabilities
+}
+
+export interface DiscoverCategory {
+  role: string
+  title: string
+  icon: string
+  models: DiscoverModel[]
+  recommended: string | null
+}
+
+export interface DiscoverResp {
+  updated: number
+  categories: DiscoverCategory[]
+  sys_ram_gb: number
+}
+
+export interface RoutingResp {
+  routes: { name: string; target: string; api_base: string | null }[]
+  fallbacks: Record<string, string[]>[]
+  context_window_fallbacks: Record<string, string[]>[]
+  gateway_reachable: boolean
 }

 export interface Health {
  status: string
  version: string
  engine_reachable: boolean
+  gateway_reachable: boolean
 }
@@ -1,18 +1,31 @@
 import { useEffect, useState } from "react"
-import { api, type ModelInfo } from "@/lib/api"
+import { api, type DiscoverResp, type Fit, type ModelInfo } from "@/lib/api"
+import { CapsChips } from "@/components/CapsChips"
+import { cn } from "@/lib/utils"

 function fmtSize(b: number | null) {
  if (!b) return "—"
  const gb = b / 1024 ** 3
  return gb >= 1 ? `${gb.toFixed(1)} GB` : `${(b / 1024 ** 2).toFixed(0)} MB`
 }
-
 function fmtCtx(c: number | null) {
-  if (!c) return "—"
-  return `${Math.round(c / 1024)}k`
+  return c ? `${Math.round(c / 1024)}k` : "—"
 }

-export function ModelsView() {
+function FitBadge({ fit }: { fit: Fit }) {
+  const tone = {
+    perfect: "bg-emerald-500/15 text-emerald-500",
+    marginal: "bg-amber-500/15 text-amber-500",
+    too_tight: "bg-red-500/15 text-red-500",
+  }[fit.level]
+  return (
+    <span className={cn("rounded px-1.5 py-0.5 text-[11px] font-medium", tone)}>
+      {fit.text} · ~{fit.req_gb} GB
+    </span>
+  )
+}
+
+function Installed() {
  const [models, setModels] = useState<ModelInfo[]>([])
  const [error, setError] = useState("")
  const [loading, setLoading] = useState(true)
@@ -24,71 +37,133 @@ export function ModelsView() {
      .finally(() => setLoading(false))
  }, [])

+  if (loading) return <div className="text-sm text-muted-foreground">Lade…</div>
+  if (error)
+    return (
+      <div className="rounded-md border border-border bg-card p-3 text-sm text-muted-foreground">
+        Engine nicht erreichbar oder keine Config gefunden ({error}).
+      </div>
+    )
+
+  return (
+    <div className="overflow-hidden rounded-xl border border-border bg-card">
+      <table className="w-full text-sm">
+        <thead>
+          <tr className="border-b border-border text-left text-xs text-muted-foreground">
+            <th className="px-4 py-2 font-medium">Modell</th>
+            <th className="px-4 py-2 font-medium">Rolle</th>
+            <th className="px-4 py-2 font-medium">Fähigkeiten</th>
+            <th className="px-4 py-2 font-medium">Kontext</th>
+            <th className="px-4 py-2 font-medium">Größe</th>
+          </tr>
+        </thead>
+        <tbody>
+          {models.length === 0 && (
+            <tr>
+              <td colSpan={5} className="px-4 py-8 text-center text-muted-foreground">
+                Keine Modelle konfiguriert.
+              </td>
+            </tr>
+          )}
+          {models.map((m) => (
+            <tr key={m.name} className="border-b border-border/50 last:border-0">
+              <td className="px-4 py-2.5 font-medium">{m.name}</td>
+              <td className="px-4 py-2.5">
+                {m.role ? (
+                  <span className="rounded-md bg-primary/15 px-2 py-0.5 text-xs text-primary">{m.role}</span>
+                ) : (
+                  <span className="text-muted-foreground">—</span>
+                )}
+              </td>
+              <td className="px-4 py-2.5">
+                <CapsChips caps={m.capabilities} />
+              </td>
+              <td className="px-4 py-2.5 text-muted-foreground">{fmtCtx(m.ctx)}</td>
+              <td className="px-4 py-2.5 text-muted-foreground">{fmtSize(m.size_bytes)}</td>
+            </tr>
+          ))}
+        </tbody>
+      </table>
+    </div>
+  )
+}
+
+function Discover() {
+  const [data, setData] = useState<DiscoverResp | null>(null)
+  const [error, setError] = useState("")
+  const [loading, setLoading] = useState(true)
+
+  useEffect(() => {
+    api<DiscoverResp>("/api/discover")
+      .then(setData)
+      .catch((e) => setError(String(e)))
+      .finally(() => setLoading(false))
+  }, [])
+
+  if (loading) return <div className="text-sm text-muted-foreground">Suche aktuelle Modelle…</div>
+  if (error || !data)
+    return (
+      <div className="rounded-md border border-border bg-card p-3 text-sm text-muted-foreground">
+        Modell-Quellen gerade nicht erreichbar ({error}).
+      </div>
+    )
+
+  return (
+    <div className="space-y-6">
+      <div className="text-xs text-muted-foreground">
+        Live von HuggingFace · Hardware-Fit für ~{data.sys_ram_gb} GB · ⭐ = beste Wahl je Kategorie
+      </div>
+      {data.categories.map((cat) => (
+        <div key={cat.role} className="space-y-2">
+          <h3 className="text-sm font-semibold">{cat.title}</h3>
+          <div className="grid gap-2 sm:grid-cols-2">
+            {cat.models.map((m) => (
+              <div key={m.repo} className="rounded-lg border border-border bg-card p-3">
+                <div className="flex items-center gap-2">
+                  {cat.recommended === m.repo && <span title="beste Wahl">⭐</span>}
+                  <span className="truncate text-sm font-medium">{m.name}</span>
+                </div>
+                <div className="mt-1 text-xs text-muted-foreground">{m.author}</div>
+                <div className="mt-2 flex flex-wrap items-center gap-1">
+                  <CapsChips caps={m.caps} />
+                  <FitBadge fit={m.fit} />
+                </div>
+              </div>
+            ))}
+          </div>
+        </div>
+      ))}
+    </div>
+  )
+}
+
+export function ModelsView() {
+  const [tab, setTab] = useState<"installed" | "discover">("installed")
  return (
    <div className="space-y-4">
      <div>
        <h1 className="text-xl font-semibold">Modelle &amp; Routing</h1>
        <p className="text-sm text-muted-foreground">
-          In llama-swap konfigurierte Modelle (Phase 0: read-only). Installieren, Gruppen &amp;
-          Auto-Routing folgen in Phase 1.
+          Installierte Modelle (mit Fähigkeiten) und aktuell beste Modelle für deine Hardware.
        </p>
      </div>

-      {loading && <div className="text-sm text-muted-foreground">Lade…</div>}
-      {error && (
-        <div className="rounded-md border border-border bg-card p-3 text-sm text-muted-foreground">
-          Engine nicht erreichbar oder keine Config gefunden ({error}).
-        </div>
-      )}
+      <div className="inline-flex rounded-lg border border-border bg-card p-0.5 text-sm">
+        {(["installed", "discover"] as const).map((t) => (
+          <button
+            key={t}
+            onClick={() => setTab(t)}
+            className={cn(
+              "rounded-md px-3 py-1.5 transition-colors",
+              tab === t ? "bg-primary/15 text-primary" : "text-muted-foreground hover:text-foreground",
+            )}
+          >
+            {t === "installed" ? "Installiert" : "Modelle finden"}
+          </button>
+        ))}
+      </div>

-      {!loading && !error && (
-        <div className="overflow-hidden rounded-xl border border-border bg-card">
-          <table className="w-full text-sm">
-            <thead>
-              <tr className="border-b border-border text-left text-xs text-muted-foreground">
-                <th className="px-4 py-2 font-medium">Modell</th>
-                <th className="px-4 py-2 font-medium">Rolle</th>
-                <th className="px-4 py-2 font-medium">Kontext</th>
-                <th className="px-4 py-2 font-medium">Quant</th>
-                <th className="px-4 py-2 font-medium">Größe</th>
-              </tr>
-            </thead>
-            <tbody>
-              {models.length === 0 && (
-                <tr>
-                  <td colSpan={5} className="px-4 py-8 text-center text-muted-foreground">
-                    Keine Modelle konfiguriert.
-                  </td>
-                </tr>
-              )}
-              {models.map((m) => (
-                <tr key={m.name} className="border-b border-border/50 last:border-0">
-                  <td className="px-4 py-2.5 font-medium">
-                    {m.name}
-                    {m.incomplete && (
-                      <span className="ml-2 rounded bg-muted px-1.5 py-0.5 text-xs text-muted-foreground">
-                        unvollständig
-                      </span>
-                    )}
-                  </td>
-                  <td className="px-4 py-2.5">
-                    {m.role ? (
-                      <span className="rounded-md bg-primary/15 px-2 py-0.5 text-xs text-primary">
-                        {m.role}
-                      </span>
-                    ) : (
-                      <span className="text-muted-foreground">—</span>
-                    )}
-                  </td>
-                  <td className="px-4 py-2.5 text-muted-foreground">{fmtCtx(m.ctx)}</td>
-                  <td className="px-4 py-2.5 text-muted-foreground">{m.quant || "—"}</td>
-                  <td className="px-4 py-2.5 text-muted-foreground">{fmtSize(m.size_bytes)}</td>
-                </tr>
-              ))}
-            </tbody>
-          </table>
-        </div>
-      )}
+      {tab === "installed" ? <Installed /> : <Discover />}
    </div>
  )
 }
@@ -0,0 +1,79 @@
+import { useEffect, useState } from "react"
+import { api, type RoutingResp } from "@/lib/api"
+import { cn } from "@/lib/utils"
+
+export function RoutingView() {
+  const [data, setData] = useState<RoutingResp | null>(null)
+  const [error, setError] = useState("")
+
+  useEffect(() => {
+    api<RoutingResp>("/api/routing").then(setData).catch((e) => setError(String(e)))
+  }, [])
+
+  return (
+    <div className="space-y-4">
+      <div>
+        <h1 className="text-xl font-semibold">Routing</h1>
+        <p className="text-sm text-muted-foreground">
+          Der LiteLLM-Gateway bündelt alle Modelle zu einem Endpunkt. <code>auto</code> = schnell im
+          Alltag, eskaliert bei Bedarf auf <code>heavy</code>. Gilt für Hermes <em>und</em> Vibe Coding.
+        </p>
+      </div>
+
+      {error && (
+        <div className="rounded-md border border-border bg-card p-3 text-sm text-muted-foreground">
+          Gateway-Config nicht lesbar ({error}).
+        </div>
+      )}
+
+      {data && (
+        <>
+          <div className="flex items-center gap-2 text-sm">
+            <span
+              className={cn(
+                "h-2 w-2 rounded-full",
+                data.gateway_reachable ? "bg-emerald-500" : "bg-amber-500",
+              )}
+            />
+            Gateway {data.gateway_reachable ? "online" : "offline (Config wird trotzdem angezeigt)"}
+          </div>
+
+          <div className="overflow-hidden rounded-xl border border-border bg-card">
+            <table className="w-full text-sm">
+              <thead>
+                <tr className="border-b border-border text-left text-xs text-muted-foreground">
+                  <th className="px-4 py-2 font-medium">Gateway-Modell</th>
+                  <th className="px-4 py-2 font-medium">→ Backend (llama-swap)</th>
+                </tr>
+              </thead>
+              <tbody>
+                {data.routes.map((r) => (
+                  <tr key={r.name} className="border-b border-border/50 last:border-0">
+                    <td className="px-4 py-2.5 font-medium">{r.name}</td>
+                    <td className="px-4 py-2.5 font-mono text-xs text-muted-foreground">{r.target}</td>
+                  </tr>
+                ))}
+              </tbody>
+            </table>
+          </div>
+
+          {data.fallbacks.length > 0 && (
+            <div className="rounded-xl border border-border bg-card p-4 text-sm">
+              <div className="mb-2 font-medium">Eskalation (Fallbacks)</div>
+              <ul className="space-y-1 text-muted-foreground">
+                {data.fallbacks.map((f, i) => {
+                  const [k, v] = Object.entries(f)[0]
+                  return (
+                    <li key={i}>
+                      <code>{k}</code> → <code>{v.join(", ")}</code>
+                    </li>
+                  )
+                })}
+              </ul>
+            </div>
+          )}
+        </>
+      )}
+    </div>
+  )
+}