app: expose server's default context length to UI

Parse the default_num_ctx from the server's "vram-based default context"
log line and expose it through the inference compute API. This eliminates
duplicate VRAM tier calculation logic in the frontend.

- Add InferenceInfo struct with Computes and DefaultContextLength
- Rename GetInferenceComputer to GetInferenceInfo
- Handle missing default context line gracefully (older servers)
- Add DefaultContextLength to InferenceComputeResponse
- Update Settings UI to use server's default, disable slider while loading
- Add disabled prop to Slider component (grays out + hides handle)
- Migrate existing users with context_length=4096 to 0 (auto mode)

app/server/server.go

@@ -41,6 +41,11 @@ type InferenceCompute struct {
 	VRAM    string
 }
 
+type InferenceInfo struct {
+	Computes             []InferenceCompute
+	DefaultContextLength int
+}
+
 func New(s *store.Store, devMode bool) *Server {
 	p := resolvePath("ollama")
 	return &Server{store: s, bin: p, dev: devMode}
@@ -262,9 +267,12 @@ func openRotatingLog() (io.WriteCloser, error) {
 
 // Attempt to retrieve inference compute information from the server
 // log.  Set ctx to timeout to control how long to wait for the logs to appear
-func GetInferenceComputer(ctx context.Context) ([]InferenceCompute, error) {
-	inference := []InferenceCompute{}
-	marker := regexp.MustCompile(`inference compute.*library=`)
+func GetInferenceInfo(ctx context.Context) (*InferenceInfo, error) {
+	info := &InferenceInfo{}
+	computeMarker := regexp.MustCompile(`inference compute.*library=`)
+	defaultCtxMarker := regexp.MustCompile(`vram-based default context`)
+	defaultCtxRegex := regexp.MustCompile(`default_num_ctx=(\d+)`)
+
 	q := `inference compute.*%s=["]([^"]*)["]`
 	nq := `inference compute.*%s=(\S+)\s`
 	type regex struct {
@@ -330,8 +338,8 @@ func GetInferenceComputer(ctx context.Context) ([]InferenceCompute, error) {
 		scanner := bufio.NewScanner(file)
 		for scanner.Scan() {
 			line := scanner.Text()
-			match := marker.FindStringSubmatch(line)
-			if len(match) > 0 {
+			// Check for inference compute lines
+			if computeMarker.MatchString(line) {
 				ic := InferenceCompute{
 					Library: get("library", line),
 					Variant: get("variant", line),
@@ -342,12 +350,25 @@ func GetInferenceComputer(ctx context.Context) ([]InferenceCompute, error) {
 				}
 
 				slog.Info("Matched", "inference compute", ic)
-				inference = append(inference, ic)
-			} else {
-				// Break out on first non matching line after we start matching
-				if len(inference) > 0 {
-					return inference, nil
+				info.Computes = append(info.Computes, ic)
+				continue
+			}
+			// Check for default context length line
+			if defaultCtxMarker.MatchString(line) {
+				match := defaultCtxRegex.FindStringSubmatch(line)
+				if len(match) > 1 {
+					numCtx, err := strconv.Atoi(match[1])
+					if err == nil {
+						info.DefaultContextLength = numCtx
+						slog.Info("Matched default context length", "default_num_ctx", numCtx)
+					}
 				}
+				return info, nil
+			}
+			// If we've found compute info but hit a non-matching line, return what we have
+			// This handles older server versions that don't log the default context line
+			if len(info.Computes) > 0 {
+				return info, nil
 			}
 		}
 		time.Sleep(100 * time.Millisecond)

app/server/server_test.go

@@ -136,44 +136,50 @@ func TestServerCmd(t *testing.T) {
 	}
 }
 
-func TestGetInferenceComputer(t *testing.T) {
+func TestGetInferenceInfo(t *testing.T) {
 	tests := []struct {
-		name string
-		log  string
-		exp  []InferenceCompute
+		name             string
+		log              string
+		expComputes      []InferenceCompute
+		expDefaultCtxLen int
 	}{
 		{
 			name: "metal",
 			log: `time=2025-06-30T09:23:07.374-07:00 level=DEBUG source=sched.go:108 msg="starting llm scheduler"
 time=2025-06-30T09:23:07.416-07:00 level=INFO source=types.go:130 msg="inference compute" id=0 library=metal variant="" compute="" driver=0.0 name="" total="96.0 GiB" available="96.0 GiB"
+time=2025-06-30T09:23:07.417-07:00 level=INFO source=routes.go:1721 msg="vram-based default context" total_vram="96.0 GiB" default_num_ctx=262144
 time=2025-06-30T09:25:56.197-07:00 level=DEBUG source=ggml.go:155 msg="key not found" key=general.alignment default=32
 `,
-			exp: []InferenceCompute{{
+			expComputes: []InferenceCompute{{
 				Library: "metal",
 				Driver:  "0.0",
 				VRAM:    "96.0 GiB",
 			}},
+			expDefaultCtxLen: 262144,
 		},
 		{
 			name: "cpu",
 			log: `time=2025-07-01T17:59:51.470Z level=INFO source=gpu.go:377 msg="no compatible GPUs were discovered"
 time=2025-07-01T17:59:51.470Z level=INFO source=types.go:130 msg="inference compute" id=0 library=cpu variant="" compute="" driver=0.0 name="" total="31.3 GiB" available="30.4 GiB"
+time=2025-07-01T17:59:51.471Z level=INFO source=routes.go:1721 msg="vram-based default context" total_vram="31.3 GiB" default_num_ctx=32768
 [GIN] 2025/07/01 - 18:00:09 | 200 |      50.263µs | 100.126.204.152 | HEAD     "/"
 `,
-			exp: []InferenceCompute{{
+			expComputes: []InferenceCompute{{
 				Library: "cpu",
 				Driver:  "0.0",
 				VRAM:    "31.3 GiB",
 			}},
+			expDefaultCtxLen: 32768,
 		},
 		{
 			name: "cuda1",
 			log: `time=2025-07-01T19:33:43.162Z level=DEBUG source=amd_linux.go:419 msg="amdgpu driver not detected /sys/module/amdgpu"
 releasing cuda driver library
 time=2025-07-01T19:33:43.162Z level=INFO source=types.go:130 msg="inference compute" id=GPU-452cac9f-6960-839c-4fb3-0cec83699196 library=cuda variant=v12 compute=6.1 driver=12.7 name="NVIDIA GeForce GT 1030" total="3.9 GiB" available="3.9 GiB"
+time=2025-07-01T19:33:43.163Z level=INFO source=routes.go:1721 msg="vram-based default context" total_vram="3.9 GiB" default_num_ctx=4096
 [GIN] 2025/07/01 - 18:00:09 | 200 |      50.263µs | 100.126.204.152 | HEAD     "/"
 `,
-			exp: []InferenceCompute{{
+			expComputes: []InferenceCompute{{
 				Library: "cuda",
 				Variant: "v12",
 				Compute: "6.1",
@@ -181,6 +187,7 @@ time=2025-07-01T19:33:43.162Z level=INFO source=types.go:130 msg="inference comp
 				Name:    "NVIDIA GeForce GT 1030",
 				VRAM:    "3.9 GiB",
 			}},
+			expDefaultCtxLen: 4096,
 		},
 		{
 			name: "frank",
@@ -188,9 +195,10 @@ time=2025-07-01T19:33:43.162Z level=INFO source=types.go:130 msg="inference comp
 		releasing cuda driver library
 		time=2025-07-01T19:36:13.315Z level=INFO source=types.go:130 msg="inference compute" id=GPU-d6de3398-9932-6902-11ec-fee8e424c8a2 library=cuda variant=v12 compute=7.5 driver=12.8 name="NVIDIA GeForce RTX 2080 Ti" total="10.6 GiB" available="10.4 GiB"
 		time=2025-07-01T19:36:13.315Z level=INFO source=types.go:130 msg="inference compute" id=GPU-9abb57639fa80c50 library=rocm variant="" compute=gfx1030 driver=6.3 name=1002:73bf total="16.0 GiB" available="1.3 GiB"
+		time=2025-07-01T19:36:13.316Z level=INFO source=routes.go:1721 msg="vram-based default context" total_vram="26.6 GiB" default_num_ctx=32768
 		[GIN] 2025/07/01 - 18:00:09 | 200 |      50.263µs | 100.126.204.152 | HEAD     "/"
 		`,
-			exp: []InferenceCompute{
+			expComputes: []InferenceCompute{
 				{
 					Library: "cuda",
 					Variant: "v12",
@@ -207,6 +215,20 @@ time=2025-07-01T19:33:43.162Z level=INFO source=types.go:130 msg="inference comp
 					VRAM:    "16.0 GiB",
 				},
 			},
+			expDefaultCtxLen: 32768,
+		},
+		{
+			name: "missing_default_context",
+			log: `time=2025-06-30T09:23:07.374-07:00 level=DEBUG source=sched.go:108 msg="starting llm scheduler"
+time=2025-06-30T09:23:07.416-07:00 level=INFO source=types.go:130 msg="inference compute" id=0 library=metal variant="" compute="" driver=0.0 name="" total="96.0 GiB" available="96.0 GiB"
+time=2025-06-30T09:25:56.197-07:00 level=DEBUG source=ggml.go:155 msg="key not found" key=general.alignment default=32
+`,
+			expComputes: []InferenceCompute{{
+				Library: "metal",
+				Driver:  "0.0",
+				VRAM:    "96.0 GiB",
+			}},
+			expDefaultCtxLen: 0, // No default context line, should return 0
 		},
 	}
 	for _, tt := range tests {
@@ -219,18 +241,21 @@ time=2025-07-01T19:33:43.162Z level=INFO source=types.go:130 msg="inference comp
 			}
 			ctx, cancel := context.WithTimeout(t.Context(), 10*time.Millisecond)
 			defer cancel()
-			ics, err := GetInferenceComputer(ctx)
+			info, err := GetInferenceInfo(ctx)
 			if err != nil {
-				t.Fatalf(" failed to get inference compute: %v", err)
+				t.Fatalf("failed to get inference info: %v", err)
 			}
-			if !reflect.DeepEqual(ics, tt.exp) {
-				t.Fatalf("got:\n%#v\nwant:\n%#v", ics, tt.exp)
+			if !reflect.DeepEqual(info.Computes, tt.expComputes) {
+				t.Fatalf("computes mismatch\ngot:\n%#v\nwant:\n%#v", info.Computes, tt.expComputes)
+			}
+			if info.DefaultContextLength != tt.expDefaultCtxLen {
+				t.Fatalf("default context length mismatch: got %d, want %d", info.DefaultContextLength, tt.expDefaultCtxLen)
 			}
 		})
 	}
 }
 
-func TestGetInferenceComputerTimeout(t *testing.T) {
+func TestGetInferenceInfoTimeout(t *testing.T) {
 	ctx, cancel := context.WithTimeout(t.Context(), 10*time.Millisecond)
 	defer cancel()
 	tmpDir := t.TempDir()
@@ -239,7 +264,7 @@ func TestGetInferenceComputerTimeout(t *testing.T) {
 	if err != nil {
 		t.Fatalf("failed to write log file %s: %s", serverLogPath, err)
 	}
-	_, err = GetInferenceComputer(ctx)
+	_, err = GetInferenceInfo(ctx)
 	if err == nil {
 		t.Fatal("expected timeout")
 	}

app/store/database.go

@@ -14,7 +14,7 @@ import (
 
 // currentSchemaVersion defines the current database schema version.
 // Increment this when making schema changes that require migrations.
-const currentSchemaVersion = 12
+const currentSchemaVersion = 13
 
 // database wraps the SQLite connection.
 // SQLite handles its own locking for concurrent access:
@@ -73,7 +73,7 @@ func (db *database) init() error {
 		agent BOOLEAN NOT NULL DEFAULT 0,
 		tools BOOLEAN NOT NULL DEFAULT 0,
 		working_dir TEXT NOT NULL DEFAULT '',
-		context_length INTEGER NOT NULL DEFAULT 4096,
+		context_length INTEGER NOT NULL DEFAULT 0,
 		window_width INTEGER NOT NULL DEFAULT 0,
 		window_height INTEGER NOT NULL DEFAULT 0,
 		config_migrated BOOLEAN NOT NULL DEFAULT 0,
@@ -244,6 +244,12 @@ func (db *database) migrate() error {
 				return fmt.Errorf("migrate v11 to v12: %w", err)
 			}
 			version = 12
+		case 12:
+			// change default context_length from 4096 to 0 (VRAM-based tiered defaults)
+			if err := db.migrateV12ToV13(); err != nil {
+				return fmt.Errorf("migrate v12 to v13: %w", err)
+			}
+			version = 13
 		default:
 			// If we have a version we don't recognize, just set it to current
 			// This might happen during development
@@ -452,6 +458,23 @@ func (db *database) migrateV11ToV12() error {
 	return nil
 }
 
+// migrateV12ToV13 changes the default context_length from 4096 to 0
+// When context_length is 0, the ollama server uses VRAM-based tiered defaults
+func (db *database) migrateV12ToV13() error {
+	// Update users who have the old default of 4096 to the new default of 0
+	_, err := db.conn.Exec(`UPDATE settings SET context_length = 0 WHERE context_length = 4096`)
+	if err != nil {
+		return fmt.Errorf("update context_length default: %w", err)
+	}
+
+	_, err = db.conn.Exec(`UPDATE settings SET schema_version = 13`)
+	if err != nil {
+		return fmt.Errorf("update schema version: %w", err)
+	}
+
+	return nil
+}
+
 // cleanupOrphanedData removes orphaned records that may exist due to the foreign key bug
 func (db *database) cleanupOrphanedData() error {
 	_, err := db.conn.Exec(`

app/store/testdata/schema.sql

@@ -13,7 +13,7 @@ CREATE TABLE IF NOT EXISTS settings (
     agent BOOLEAN NOT NULL DEFAULT 0,
     tools BOOLEAN NOT NULL DEFAULT 0,
     working_dir TEXT NOT NULL DEFAULT '',
-    context_length INTEGER NOT NULL DEFAULT 4096,
+    context_length INTEGER NOT NULL DEFAULT 0,
     window_width INTEGER NOT NULL DEFAULT 0,
     window_height INTEGER NOT NULL DEFAULT 0,
     config_migrated BOOLEAN NOT NULL DEFAULT 0,

app/ui/app/codegen/gotypes.gen.ts

@@ -289,10 +289,12 @@ export class InferenceCompute {
 }
 export class InferenceComputeResponse {
     inferenceComputes: InferenceCompute[];
+    defaultContextLength: number;
 
     constructor(source: any = {}) {
         if ('string' === typeof source) source = JSON.parse(source);
         this.inferenceComputes = this.convertValues(source["inferenceComputes"], InferenceCompute);
+        this.defaultContextLength = source["defaultContextLength"];
     }
 
 	convertValues(a: any, classs: any, asMap: boolean = false): any {

app/ui/app/src/api.ts

@@ -4,7 +4,6 @@ import {
   ChatEvent,
   DownloadEvent,
   ErrorEvent,
-  InferenceCompute,
   InferenceComputeResponse,
   ModelCapabilitiesResponse,
   Model,
@@ -379,7 +378,7 @@ export async function* pullModel(
   }
 }
 
-export async function getInferenceCompute(): Promise<InferenceCompute[]> {
+export async function getInferenceCompute(): Promise<InferenceComputeResponse> {
   const response = await fetch(`${API_BASE}/api/v1/inference-compute`);
   if (!response.ok) {
     throw new Error(
@@ -388,8 +387,7 @@ export async function getInferenceCompute(): Promise<InferenceCompute[]> {
   }
 
   const data = await response.json();
-  const inferenceComputeResponse = new InferenceComputeResponse(data);
-  return inferenceComputeResponse.inferenceComputes || [];
+  return new InferenceComputeResponse(data);
 }
 
 export async function fetchHealth(): Promise<boolean> {

app/ui/app/src/components/Settings.tsx

@@ -19,7 +19,7 @@ import { Settings as SettingsType } from "@/gotypes";
 import { useNavigate } from "@tanstack/react-router";
 import { useUser } from "@/hooks/useUser";
 import { useQuery, useMutation, useQueryClient } from "@tanstack/react-query";
-import { getSettings, updateSettings } from "@/api";
+import { getSettings, updateSettings, getInferenceCompute } from "@/api";
 
 function AnimatedDots() {
   return (
@@ -65,6 +65,13 @@ export default function Settings() {
 
   const settings = settingsData?.settings || null;
 
+  const { data: inferenceComputeResponse } = useQuery({
+    queryKey: ["inferenceCompute"],
+    queryFn: getInferenceCompute,
+  });
+
+  const defaultContextLength = inferenceComputeResponse?.defaultContextLength;
+
   const updateSettingsMutation = useMutation({
     mutationFn: updateSettings,
     onSuccess: () => {
@@ -148,7 +155,7 @@ export default function Settings() {
         Models: "",
         Agent: false,
         Tools: false,
-        ContextLength: 4096,
+        ContextLength: 0,
         AirplaneMode: false,
       });
       updateSettingsMutation.mutate(defaultSettings);
@@ -419,13 +426,11 @@ export default function Settings() {
                     </Description>
                     <div className="mt-3">
                       <Slider
-                        value={(() => {
-                          // Otherwise use the settings value
-                          return settings.ContextLength || 4096;
-                        })()}
+                        value={settings.ContextLength || defaultContextLength || 0}
                         onChange={(value) => {
                           handleChange("ContextLength", value);
                         }}
+                        disabled={!defaultContextLength}
                         options={[
                           { value: 4096, label: "4k" },
                           { value: 8192, label: "8k" },
@@ -440,6 +445,7 @@ export default function Settings() {
                   </div>
                 </div>
               </Field>
+
               {/* Airplane Mode */}
               <Field>
                 <div className="flex items-start justify-between gap-4">

app/ui/app/src/components/ui/slider.tsx

@@ -6,10 +6,11 @@ export interface SliderProps {
   value?: number;
   onChange?: (value: number) => void;
   className?: string;
+  disabled?: boolean;
 }
 
 const Slider = React.forwardRef<HTMLDivElement, SliderProps>(
-  ({ label, options, value = 0, onChange }, ref) => {
+  ({ label, options, value = 0, onChange, disabled = false }, ref) => {
     const [selectedValue, setSelectedValue] = React.useState(value);
     const [isDragging, setIsDragging] = React.useState(false);
     const containerRef = React.useRef<HTMLDivElement>(null);
@@ -20,6 +21,7 @@ const Slider = React.forwardRef<HTMLDivElement, SliderProps>(
     }, [value]);
 
     const handleClick = (optionValue: number) => {
+      if (disabled) return;
       setSelectedValue(optionValue);
       onChange?.(optionValue);
     };
@@ -39,6 +41,7 @@ const Slider = React.forwardRef<HTMLDivElement, SliderProps>(
     };
 
     const handleMouseDown = (e: React.MouseEvent) => {
+      if (disabled) return;
       setIsDragging(true);
       e.preventDefault();
     };
@@ -77,7 +80,7 @@ const Slider = React.forwardRef<HTMLDivElement, SliderProps>(
     }
 
     return (
-      <div className="space-y-2" ref={ref}>
+      <div className={`space-y-2 ${disabled ? "opacity-50" : ""}`} ref={ref}>
         {label && <label className="text-sm font-medium">{label}</label>}
         <div className="relative">
           <div className="absolute top-[9px] left-2 right-2 h-1 bg-neutral-200 dark:bg-neutral-700 pointer-events-none rounded-full" />
@@ -88,10 +91,11 @@ const Slider = React.forwardRef<HTMLDivElement, SliderProps>(
                 <button
                   onClick={() => handleClick(option.value)}
                   onMouseDown={handleMouseDown}
-                  className="relative px-3 py-6 -mx-3 -my-6 z-10 cursor-pointer"
+                  disabled={disabled}
+                  className={`relative px-3 py-6 -mx-3 -my-6 z-10 ${disabled ? "cursor-not-allowed" : "cursor-pointer"}`}
                 >
                   <div className="relative w-5 h-5 flex items-center justify-center">
-                    {selectedValue === option.value && (
+                    {selectedValue === option.value && !disabled && (
                       <div className="w-4 h-4 bg-white dark:bg-white border border-neutral-400 dark:border-neutral-500 rounded-full cursor-grab active:cursor-grabbing" />
                     )}
                   </div>

app/ui/app/src/hooks/useSelectedModel.ts

@@ -26,12 +26,14 @@ export function useSelectedModel(currentChatId?: string, searchQuery?: string) {
     currentChatId && currentChatId !== "new" ? currentChatId : "",
   );
 
-  const { data: inferenceComputes = [] } = useQuery({
-    queryKey: ["inference-compute"],
+  const { data: inferenceComputeResponse } = useQuery({
+    queryKey: ["inferenceCompute"],
     queryFn: getInferenceCompute,
     enabled: !settings.selectedModel, // Only fetch if no model is selected
   });
 
+  const inferenceComputes = inferenceComputeResponse?.inferenceComputes || [];
+
   const totalVRAM = useMemo(
     () => getTotalVRAM(inferenceComputes),
     [inferenceComputes],

app/ui/responses/types.go

@@ -45,7 +45,8 @@ type InferenceCompute struct {
 }
 
 type InferenceComputeResponse struct {
-	InferenceComputes []InferenceCompute `json:"inferenceComputes"`
+	InferenceComputes    []InferenceCompute `json:"inferenceComputes"`
+	DefaultContextLength int                `json:"defaultContextLength"`
 }
 
 type ModelCapabilitiesResponse struct {

app/ui/ui.go

@@ -1417,11 +1417,6 @@ func (s *Server) getSettings(w http.ResponseWriter, r *http.Request) error {
 		settings.Models = envconfig.Models()
 	}
 
-	// set default context length if not set
-	if settings.ContextLength == 0 {
-		settings.ContextLength = 4096
-	}
-
 	// Include current runtime settings
 	settings.Agent = s.Agent
 	settings.Tools = s.Tools
@@ -1463,14 +1458,14 @@ func (s *Server) settings(w http.ResponseWriter, r *http.Request) error {
 func (s *Server) getInferenceCompute(w http.ResponseWriter, r *http.Request) error {
 	ctx, cancel := context.WithTimeout(r.Context(), 500*time.Millisecond)
 	defer cancel()
-	serverInferenceComputes, err := server.GetInferenceComputer(ctx)
+	info, err := server.GetInferenceInfo(ctx)
 	if err != nil {
-		s.log().Error("failed to get inference compute", "error", err)
-		return fmt.Errorf("failed to get inference compute: %w", err)
+		s.log().Error("failed to get inference info", "error", err)
+		return fmt.Errorf("failed to get inference info: %w", err)
 	}
 
-	inferenceComputes := make([]responses.InferenceCompute, len(serverInferenceComputes))
-	for i, ic := range serverInferenceComputes {
+	inferenceComputes := make([]responses.InferenceCompute, len(info.Computes))
+	for i, ic := range info.Computes {
 		inferenceComputes[i] = responses.InferenceCompute{
 			Library: ic.Library,
 			Variant: ic.Variant,
@@ -1482,7 +1477,8 @@ func (s *Server) getInferenceCompute(w http.ResponseWriter, r *http.Request) err
 	}
 
 	response := responses.InferenceComputeResponse{
-		InferenceComputes: inferenceComputes,
+		InferenceComputes:    inferenceComputes,
+		DefaultContextLength: info.DefaultContextLength,
 	}
 
 	w.Header().Set("Content-Type", "application/json")