feat: add SettingsPage with HuggingFace model search and fix flexbox layout

- Add SettingsPage component with macOS/Linear-style settings UI (sp-* classes)
- Fix HuggingFace results list squeeze: sp-card-row--full now has align-items: stretch
  overriding the inherited flex-start, ensuring children fill 100% horizontal width
- Fix sp-hf-file row layout: width 100%, grid columns for name/quant/size/download
- Fix sp-hf-name truncation: explicit overflow/ellipsis/white-space in CSS
- Add white-space: nowrap to sp-hf-size and sp-hf-quant-badge to protect secondary columns
- Add runtime-deps crate, Tailwind/PostCSS config, runtime-defaults, MessageBubble test
- Update Cargo workspace, inference-server, tauri runtime types and process handling

crates/inference-core/src/profile.rs

@@ -11,29 +11,34 @@ pub enum RuntimeProfile {
 
 impl RuntimeProfile {
     /// Returns sensible defaults for each profile.
+    ///
+    /// `n_ctx` sets the KV cache / context window size. Larger values allow longer
+    /// conversations but consume more VRAM. Override at runtime with `LOCALLLM_CTX`.
+    /// `n_batch` controls the prefill chunk size — kept independent of `n_ctx` to
+    /// avoid pre-allocating the full window as a single batch buffer (OOM risk).
     pub fn defaults(&self) -> ProfileDefaults {
         match self {
             Self::Turbo4B => ProfileDefaults {
                 n_gpu_layers: 999,
-                n_ctx: 4096,
-                n_batch: 512,
-                n_ubatch: 128,
+                n_ctx: 131_072, // 128 K — covers very long conversations; needs ~4 GB KV VRAM at FP16
+                n_batch: 2048,  // prefill chunk size; independent of n_ctx
+                n_ubatch: 512,
                 flash_attn: true,
                 n_threads: 8,
             },
             Self::Balanced9B => ProfileDefaults {
                 n_gpu_layers: 999,
-                n_ctx: 2048,
-                n_batch: 256,
-                n_ubatch: 64,
+                n_ctx: 65_536, // 64 K
+                n_batch: 2048,
+                n_ubatch: 256,
                 flash_attn: true,
                 n_threads: 8,
             },
             Self::CpuFallback => ProfileDefaults {
                 n_gpu_layers: 0,
-                n_ctx: 2048,
-                n_batch: 128,
-                n_ubatch: 32,
+                n_ctx: 16_384, // 16 K — CPU KV cache is RAM-backed; keep conservative
+                n_batch: 512,
+                n_ubatch: 128,
                 flash_attn: false,
                 n_threads: num_cpus(),
             },