kpn load: laitteiston VRAM/RAM tarkistus, liian isot mallit merkitään

Hub: uusi GET /api/v1/hardware palauttaa natiivisolmun GPU/RAM-tiedot.
Frontend: kpn load hakee laitteistotiedon ja näyttää mallit joihin
laite riittää. Liian isot mallit näkyvät yliviivattuina + varoitus.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

network-poc/hub/src/main.rs

@@ -385,6 +385,7 @@ async fn main() {
         .route("/api/stats", get(api_stats))
         .route("/api/v1/chat/completions", axum::routing::post(api_chat_completions))
         .route("/api/v1/model", axum::routing::post(api_change_model))
+        .route("/api/v1/hardware", get(api_hardware))
         .route("/admin", get(admin_page))
         .nest_service("/", {
             let static_dir = std::env::var("STATIC_DIR").unwrap_or_else(|_| "../static".to_string());
@@ -959,6 +960,33 @@ struct ChatCompletionResponse {
     tokens_generated: u64,
 }
 
+async fn api_hardware(
+    axum::extract::State(state): axum::extract::State<Arc<AppState>>,
+) -> axum::response::Response {
+    // Etsitään natiivisolmun GPU-tiedot sessiosta
+    let sessions = state.db.get_sessions(50);
+    let native = sessions.iter().find(|s| {
+        s.get("node_type").and_then(|v| v.as_str()) == Some("native")
+    });
+
+    let (vram_mb, gpu_name, ram_mb) = if let Some(s) = native {
+        let gpus = s.get("gpus").and_then(|v| v.as_array());
+        let gpu = gpus.and_then(|g| g.first());
+        let vram = gpu.and_then(|g| g.get("vram_total_mb")).and_then(|v| v.as_u64()).unwrap_or(0);
+        let name = gpu.and_then(|g| g.get("name")).and_then(|v| v.as_str()).unwrap_or("?");
+        let ram = s.get("system").and_then(|v| v.get("ram_total_mb")).and_then(|v| v.as_u64()).unwrap_or(0);
+        (vram, name.to_string(), ram)
+    } else {
+        (0, "ei natiivisolmua".to_string(), 0)
+    };
+
+    axum::Json(serde_json::json!({
+        "gpu_name": gpu_name,
+        "vram_mb": vram_mb,
+        "ram_mb": ram_mb,
+    })).into_response()
+}
+
 async fn api_change_model(
     axum::extract::State(state): axum::extract::State<Arc<AppState>>,
     axum::Json(payload): axum::Json<serde_json::Value>,

network-poc/static/index.html

@@ -2373,19 +2373,41 @@ Files: ${Object.keys(generatedFiles).join(', ')}`;
             if (sub === 'load') {
                 const arg = parts[2];
                 const ollamaModels = [
-                    { id: '1', name: 'qwen2.5-coder:0.5b', size: '~400 MB', type: 'selain + Ollama' },
+                    { id: '1', name: 'qwen2.5-coder:0.5b', size: '~400 MB', vram_mb: 0, type: 'selain + Ollama' },
-                    { id: '2', name: 'qwen2.5-coder:1.5b', size: '~1 GB', type: 'Ollama GPU' },
+                    { id: '2', name: 'qwen2.5-coder:1.5b', size: '~1 GB', vram_mb: 1500, type: 'Ollama GPU' },
-                    { id: '3', name: 'qwen2.5-coder:7b', size: '~4.7 GB', type: 'Ollama GPU', default: true },
+                    { id: '3', name: 'qwen2.5-coder:7b', size: '~4.7 GB', vram_mb: 5500, type: 'Ollama GPU', default: true },
-                    { id: '4', name: 'qwen2.5-coder:14b', size: '~9 GB', type: 'Ollama GPU' },
+                    { id: '4', name: 'qwen2.5-coder:14b', size: '~9 GB', vram_mb: 10000, type: 'Ollama GPU' },
-                    { id: '5', name: 'qwen2.5-coder:32b', size: '~20 GB', type: 'Ollama GPU' },
+                    { id: '5', name: 'qwen2.5-coder:32b', size: '~20 GB', vram_mb: 21000, type: 'Ollama GPU' },
                 ];
                 if (!arg) {
-                    termLog('  Mallit:', '#c9d1d9');
+                    // Haetaan laitteistotiedot ja näytetään sopivat mallit
-                    for (const m of ollamaModels) {
+                    fetch('/api/v1/hardware').then(r => r.json()).then(hw => {
-                        const active = m.default ? ' <span style="color:#3fb950">← aktiivinen</span>' : '';
+                        const vram = hw.vram_mb || 0;
-                        termLog(`    <span style="color:#58a6ff">${m.id}</span>  ${m.name}  <span style="color:#8b949e">${m.size} | ${m.type}</span>${active}`);
+                        const ram = hw.ram_mb || 0;
-                    }
+                        const gpu = hw.gpu_name || '?';
-                    termLog('  Käyttö: kpn load &lt;numero&gt;', '#8b949e');
+                        const available = vram || ram; // CPU-fallback käyttää RAM:ia
+                        if (vram > 0) {
+                            termLog(`  <span style="color:#8b949e">GPU: ${gpu} | VRAM: ${Math.round(vram/1024)} GB | RAM: ${Math.round(ram/1024)} GB</span>`);
+                        } else if (ram > 0) {
+                            termLog(`  <span style="color:#8b949e">Ei GPU:ta | RAM: ${Math.round(ram/1024)} GB (CPU-moodi)</span>`);
+                        }
+                        termLog('  Mallit:', '#c9d1d9');
+                        for (const m of ollamaModels) {
+                            const fits = m.vram_mb === 0 || m.vram_mb < available;
+                            const active = m.default ? ' <span style="color:#3fb950">← aktiivinen</span>' : '';
+                            const icon = fits ? `<span style="color:#58a6ff">${m.id}</span>` : `<span style="color:#8b949e;text-decoration:line-through">${m.id}</span>`;
+                            const warn = !fits ? ' <span style="color:#f85149">⚠ ei mahdu</span>' : '';
+                            termLog(`    ${icon}  ${fits ? '' : '<span style="color:#8b949e">'}${m.name}  ${m.size} | ${m.type}${fits ? '' : '</span>'}${active}${warn}`);
+                        }
+                        termLog('  Käyttö: kpn load &lt;numero&gt;', '#8b949e');
+                    }).catch(() => {
+                        termLog('  Mallit:', '#c9d1d9');
+                        for (const m of ollamaModels) {
+                            const active = m.default ? ' <span style="color:#3fb950">← aktiivinen</span>' : '';
+                            termLog(`    <span style="color:#58a6ff">${m.id}</span>  ${m.name}  <span style="color:#8b949e">${m.size} | ${m.type}</span>${active}`);
+                        }
+                        termLog('  Käyttö: kpn load &lt;numero&gt;', '#8b949e');
+                    });
                     return;
                 }
                 const selected = ollamaModels.find(m => m.id === arg || m.name === arg);