Feat/mrope clean

.gitattributes

@@ -19,6 +19,8 @@ ml/backend/**/*.comp linguist-vendored
 ml/backend/**/*.glsl linguist-vendored
 ml/backend/**/CMakeLists.txt linguist-vendored
 
+app/webview linguist-vendored
+
 llama/build-info.cpp linguist-generated
 ml/backend/ggml/ggml/src/ggml-metal/ggml-metal-embed.s linguist-generated
 

.github/workflows/test.yaml

@@ -228,7 +228,7 @@ jobs:
 
       - uses: golangci/golangci-lint-action@v9
         with:
-          only-new-issues: true
+          args: --new-from-merge-base ${{ github.base_ref }}
 
   patches:
     runs-on: ubuntu-latest

.golangci.yaml

@@ -1,77 +1,51 @@
 version: "2"
 linters:
-  default: none
   enable:
     - asasalint
     - bidichk
     - bodyclose
     - containedctx
-    - copyloopvar
-    - errcheck
-    - errorlint
-    - exptostd
     - gocheckcompilerdirectives
-    - gocritic
-    - govet
-    - ineffassign
     - intrange
     - makezero
     - misspell
-    - modernize
     - nilerr
-    - nilnil
     - nolintlint
     - nosprintfhostport
-    - perfsprint
-    - prealloc
-    - sloglint
-    - staticcheck
     - unconvert
-    - unused
-    - usestdlibvars
     - usetesting
     - wastedassign
     - whitespace
+  disable:
+    - errcheck
+    - usestdlibvars
   settings:
-    errcheck:
-      exclude-functions:
-        - fmt.Fprintf
-    perfsprint:
-      strconcat: false
-      concat-loop: false
+    govet:
+      disable:
+        - unusedresult
     staticcheck:
       checks:
         - all
-        # Using a deprecated function, variable, constant or field.
-        # https://staticcheck.dev/docs/checks/#SA1019
+        - -QF* # disable quick fix suggestions
         - -SA1019
-        # Incorrect or missing package comment.
-        # https://staticcheck.dev/docs/checks/#ST1000
-        - -ST1000
-        # Poorly chosen identifier.
-        # https://staticcheck.dev/docs/checks/#ST1003
-        - -ST1003
-        # The documentation of an exported function should start with the function's name.
-        # https://staticcheck.dev/docs/checks/#ST1020
-        - -ST1020
-        # The documentation of an exported type should start with type's name.
-        # https://staticcheck.dev/docs/checks/#ST1021
-        - -ST1021
-        # The documentation of an exported variable or constant should start with variable's name.
-        # https://staticcheck.dev/docs/checks/#ST1022
-        - -ST1022
-    usestdlibvars:
-      http-method: false
-      http-status-code: false
-
+        - -ST1000 # package comment format
+        - -ST1003 # underscores in package names
+        - -ST1005 # error strings should not be capitalized
+        - -ST1012 # error var naming (ErrFoo)
+        - -ST1016 # receiver name consistency
+        - -ST1020 # comment on exported function format
+        - -ST1021 # comment on exported type format
+        - -ST1022 # comment on exported var format
+        - -ST1023 # omit type from declaration
+severity:
+  default: error
+  rules:
+    - linters:
+        - gofmt
+        - goimports
+        - intrange
+      severity: info
 formatters:
   enable:
-    - gci
     - gofmt
     - gofumpt
-  settings:
-    gci:
-      sections:
-        - standard
-        - default
-        - localmodule

Makefile.sync

@@ -1,6 +1,6 @@
 UPSTREAM=https://github.com/ggml-org/llama.cpp.git
 WORKDIR=llama/vendor
-FETCH_HEAD=3cfa9c3f125763305b4226bc032f1954f08990dc
+FETCH_HEAD=7f8ef50cce40e3e7e4526a3696cb45658190e69a
 
 .PHONY: help
 help:

api/client.go

@@ -226,7 +226,14 @@ func (c *Client) stream(ctx context.Context, method, path string, data any, fn f
 
 		bts := scanner.Bytes()
 		if err := json.Unmarshal(bts, &errorResponse); err != nil {
-			return fmt.Errorf("unmarshal: %w", err)
+			if response.StatusCode >= http.StatusBadRequest {
+				return StatusError{
+					StatusCode:   response.StatusCode,
+					Status:       response.Status,
+					ErrorMessage: string(bts),
+				}
+			}
+			return errors.New(string(bts))
 		}
 
 		if response.StatusCode == http.StatusUnauthorized {

api/client_test.go

@@ -55,6 +55,7 @@ func TestClientFromEnvironment(t *testing.T) {
 type testError struct {
 	message    string
 	statusCode int
+	raw        bool // if true, write message as-is instead of JSON encoding
 }
 
 func (e testError) Error() string {
@@ -111,6 +112,20 @@ func TestClientStream(t *testing.T) {
 				},
 			},
 		},
+		{
+			name: "plain text error response",
+			responses: []any{
+				"internal server error",
+			},
+			wantErr: "internal server error",
+		},
+		{
+			name: "HTML error page",
+			responses: []any{
+				"<html><body>404 Not Found</body></html>",
+			},
+			wantErr: "404 Not Found",
+		},
 	}
 
 	for _, tc := range testCases {
@@ -135,6 +150,12 @@ func TestClientStream(t *testing.T) {
 						return
 					}
 
+					if str, ok := resp.(string); ok {
+						fmt.Fprintln(w, str)
+						flusher.Flush()
+						continue
+					}
+
 					if err := json.NewEncoder(w).Encode(resp); err != nil {
 						t.Fatalf("failed to encode response: %v", err)
 					}
@@ -173,25 +194,28 @@ func TestClientStream(t *testing.T) {
 
 func TestClientDo(t *testing.T) {
 	testCases := []struct {
-		name     string
-		response any
-		wantErr  string
+		name           string
+		response       any
+		wantErr        string
+		wantStatusCode int
 	}{
 		{
 			name: "immediate error response",
 			response: testError{
 				message:    "test error message",
 				statusCode: http.StatusBadRequest,
 			},
-			wantErr: "test error message",
+			wantErr:        "test error message",
+			wantStatusCode: http.StatusBadRequest,
 		},
 		{
 			name: "server error response",
 			response: testError{
 				message:    "internal error",
 				statusCode: http.StatusInternalServerError,
 			},
-			wantErr: "internal error",
+			wantErr:        "internal error",
+			wantStatusCode: http.StatusInternalServerError,
 		},
 		{
 			name: "successful response",
@@ -203,18 +227,43 @@ func TestClientDo(t *testing.T) {
 				Success: true,
 			},
 		},
+		{
+			name: "plain text error response",
+			response: testError{
+				message:    "internal server error",
+				statusCode: http.StatusInternalServerError,
+				raw:        true,
+			},
+			wantErr:        "internal server error",
+			wantStatusCode: http.StatusInternalServerError,
+		},
+		{
+			name: "HTML error page",
+			response: testError{
+				message:    "<html><body>404 Not Found</body></html>",
+				statusCode: http.StatusNotFound,
+				raw:        true,
+			},
+			wantErr:        "<html><body>404 Not Found</body></html>",
+			wantStatusCode: http.StatusNotFound,
+		},
 	}
 
 	for _, tc := range testCases {
 		t.Run(tc.name, func(t *testing.T) {
 			ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 				if errResp, ok := tc.response.(testError); ok {
 					w.WriteHeader(errResp.statusCode)
-					err := json.NewEncoder(w).Encode(map[string]string{
-						"error": errResp.message,
-					})
-					if err != nil {
-						t.Fatal("failed to encode error response:", err)
+					if !errResp.raw {
+						err := json.NewEncoder(w).Encode(map[string]string{
+							"error": errResp.message,
+						})
+						if err != nil {
+							t.Fatal("failed to encode error response:", err)
+						}
+					} else {
+						// Write raw message (simulates non-JSON error responses)
+						fmt.Fprint(w, errResp.message)
 					}
 					return
 				}
@@ -241,6 +290,15 @@ func TestClientDo(t *testing.T) {
 				if err.Error() != tc.wantErr {
 					t.Errorf("error message mismatch: got %q, want %q", err.Error(), tc.wantErr)
 				}
+				if tc.wantStatusCode != 0 {
+					if statusErr, ok := err.(StatusError); ok {
+						if statusErr.StatusCode != tc.wantStatusCode {
+							t.Errorf("status code mismatch: got %d, want %d", statusErr.StatusCode, tc.wantStatusCode)
+						}
+					} else {
+						t.Errorf("expected StatusError, got %T", err)
+					}
+				}
 				return
 			}
 

api/examples/chat/main.go

@@ -15,19 +15,19 @@ func main() {
 	}
 
 	messages := []api.Message{
-		api.Message{
+		{
 			Role:    "system",
 			Content: "Provide very brief, concise responses",
 		},
-		api.Message{
+		{
 			Role:    "user",
 			Content: "Name some unusual animals",
 		},
-		api.Message{
+		{
 			Role:    "assistant",
 			Content: "Monotreme, platypus, echidna",
 		},
-		api.Message{
+		{
 			Role:    "user",
 			Content: "which of these is the most dangerous?",
 		},

app/ui/app/src/api.ts

@@ -204,12 +204,10 @@ export async function* sendMessage(
     data: uint8ArrayToBase64(att.data),
   }));
 
-  // Only send think parameter when actually requesting thinking
-  // Don't send false as it causes issues with some providers
+  // Send think parameter when it's explicitly set (true, false, or a non-empty string).
   const shouldSendThink =
     think !== undefined &&
-    ((typeof think === "boolean" && think) ||
-      (typeof think === "string" && think !== ""));
+    (typeof think === "boolean" || (typeof think === "string" && think !== ""));
 
   const response = await fetch(`${API_BASE}/api/v1/chat/${chatId}`, {
     method: "POST",

cmd/bench/bench.go

@@ -48,8 +48,8 @@ func OutputMetrics(w io.Writer, format string, metrics []Metrics, verbose bool)
 	case "benchstat":
 		if verbose {
 			printHeader := func() {
-				fmt.Printf("sysname: %s\n", runtime.GOOS)
-				fmt.Printf("machine: %s\n", runtime.GOARCH)
+				fmt.Fprintf(w, "sysname: %s\n", runtime.GOOS)
+				fmt.Fprintf(w, "machine: %s\n", runtime.GOARCH)
 			}
 			once.Do(printHeader)
 		}
@@ -147,6 +147,17 @@ func BenchmarkChat(fOpt flagOptions) error {
 		return err
 	}
 
+	var out io.Writer = os.Stdout
+	if fOpt.outputFile != nil && *fOpt.outputFile != "" {
+		f, err := os.OpenFile(*fOpt.outputFile, os.O_CREATE|os.O_WRONLY, 0o644)
+		if err != nil {
+			fmt.Fprintf(os.Stderr, "ERROR: cannot open output file %s: %v\n", *fOpt.outputFile, err)
+			return err
+		}
+		defer f.Close()
+		out = f
+	}
+
 	for _, model := range models {
 		for range *fOpt.epochs {
 			options := make(map[string]interface{})
@@ -241,13 +252,14 @@ func BenchmarkChat(fOpt flagOptions) error {
 				},
 			}
 
-			OutputMetrics(os.Stdout, *fOpt.format, metrics, *fOpt.verbose)
+			OutputMetrics(out, *fOpt.format, metrics, *fOpt.verbose)
 
 			if *fOpt.keepAlive > 0 {
 				time.Sleep(time.Duration(*fOpt.keepAlive*float64(time.Second)) + 200*time.Millisecond)
 			}
 		}
 	}
+
 	return nil
 }
 

cmd/cmd.go

@@ -1430,7 +1430,7 @@ func chat(cmd *cobra.Command, opts runOptions) (*api.Message, error) {
 		latest.Summary()
 	}
 
-	return &api.Message{Role: role, Content: fullResponse.String()}, nil
+	return &api.Message{Role: role, Thinking: thinkingContent.String(), Content: fullResponse.String()}, nil
 }
 
 func generate(cmd *cobra.Command, opts runOptions) error {

convert/convert.go

@@ -208,6 +208,8 @@ func ConvertModel(fsys fs.FS, f *os.File) error {
 		conv = &gptossModel{}
 	case "DeepseekOCRForCausalLM":
 		conv = &deepseekocr{}
+	case "DeepseekV3ForCausalLM":
+		conv = &deepseek2Model{}
 	default:
 		return fmt.Errorf("unsupported architecture %q", p.Architectures[0])
 	}