fix(ds4): resolve repository imports to GGUF

Select a concrete DeepSeek V4 Flash GGUF from Hugging Face metadata while honoring quantization preference order. Reject unresolved repository imports instead of emitting an invalid bare URI.

Assisted-by: Codex:gpt-5

core/gallery/importers/ds4.go

@@ -2,6 +2,7 @@ package importers
 
 import (
 	"encoding/json"
+	"fmt"
 	"path/filepath"
 	"strings"
 
@@ -10,6 +11,7 @@ import (
 	"github.com/mudler/LocalAI/core/schema"
 	"github.com/mudler/LocalAI/pkg/downloader"
 	"github.com/mudler/LocalAI/pkg/functions"
+	hfapi "github.com/mudler/LocalAI/pkg/huggingface-api"
 	"go.yaml.in/yaml/v2"
 )
 
@@ -115,7 +117,48 @@ func (i *DS4Importer) Import(details Details) (gallery.ModelConfig, error) {
 	// filename to "ds4flash.gguf" to match ds4's own convention (its CLI
 	// defaults to that path), so users can run the model without extra
 	// config.
-	uri := downloader.URI(details.URI)
+	selectedURI := details.URI
+	uriFilename, _ := downloader.URI(details.URI).FilenameFromUrl()
+	if !strings.HasSuffix(strings.ToLower(uriFilename), ".gguf") {
+		if details.HuggingFace == nil {
+			return gallery.ModelConfig{}, fmt.Errorf("cannot select a compatible GGUF from repository %q without HuggingFace file metadata", details.URI)
+		}
+
+		quantizations, _ := preferencesMap["quantizations"].(string)
+		var compatible []hfapi.ModelFile
+		for _, file := range details.HuggingFace.Files {
+			base := strings.ToLower(filepath.Base(file.Path))
+			if strings.HasPrefix(base, "deepseek-v4-flash-") && strings.HasSuffix(base, ".gguf") {
+				compatible = append(compatible, file)
+			}
+		}
+		if len(compatible) == 0 {
+			return gallery.ModelConfig{}, fmt.Errorf("no compatible GGUF found in repository %q", details.URI)
+		}
+
+		selected := compatible[0].URL
+		if quantizations != "" {
+			selected = compatible[len(compatible)-1].URL
+			matched := false
+			for _, quantization := range strings.Split(quantizations, ",") {
+				quantization = strings.ToLower(strings.TrimSpace(quantization))
+				for _, file := range compatible {
+					base := strings.ToLower(filepath.Base(file.Path))
+					if quantTokenMatches(base, quantization) {
+						selected = file.URL
+						matched = true
+						break
+					}
+				}
+				if matched {
+					break
+				}
+			}
+		}
+		selectedURI = selected
+	}
+
+	uri := downloader.URI(selectedURI)
 	cfg.Files = append(cfg.Files, gallery.File{
 		Filename: "ds4flash.gguf",
 		URI:      string(uri),

core/gallery/importers/ds4_test.go

@@ -5,6 +5,7 @@ import (
 	"strings"
 
 	. "github.com/mudler/LocalAI/core/gallery/importers"
+	hfapi "github.com/mudler/LocalAI/pkg/huggingface-api"
 	. "github.com/onsi/ginkgo/v2"
 	. "github.com/onsi/gomega"
 )
@@ -65,5 +66,71 @@ var _ = Describe("DS4Importer", func() {
 				"ConfigFile must specify backend: ds4, got: %s", cfg.ConfigFile)
 			Expect(strings.Contains(cfg.ConfigFile, "use_tokenizer_template: true")).To(BeTrue())
 		})
+
+		It("preserves an explicit GGUF URI with a query string", func() {
+			uri := "https://huggingface.co/antirez/deepseek-v4-gguf/resolve/main/DeepSeek-V4-Flash-UD-IQ3_XXS.gguf?download=true"
+
+			cfg, err := importer.Import(Details{URI: uri})
+
+			Expect(err).NotTo(HaveOccurred())
+			Expect(cfg.Files).To(HaveLen(1))
+			Expect(cfg.Files[0].URI).To(Equal(uri))
+		})
+
+		It("selects the highest-priority requested quantization from a repository", func() {
+			details := Details{
+				URI:         "huggingface://antirez/deepseek-v4-gguf",
+				Preferences: json.RawMessage(`{"quantizations":"UD-IQ3_XXS,IQ2_XXS"}`),
+				HuggingFace: &hfapi.ModelDetails{Files: []hfapi.ModelFile{
+					{Path: "DeepSeek-V4-Flash-IQ2_XXS.gguf", URL: "https://huggingface.co/antirez/deepseek-v4-gguf/resolve/main/DeepSeek-V4-Flash-IQ2_XXS.gguf"},
+					{Path: "DeepSeek-V4-Flash-UD-IQ3_XXS.gguf", URL: "https://huggingface.co/antirez/deepseek-v4-gguf/resolve/main/DeepSeek-V4-Flash-UD-IQ3_XXS.gguf"},
+				}},
+			}
+
+			cfg, err := importer.Import(details)
+
+			Expect(err).NotTo(HaveOccurred())
+			Expect(cfg.Files).To(HaveLen(1))
+			Expect(cfg.Files[0].URI).To(Equal("https://huggingface.co/antirez/deepseek-v4-gguf/resolve/main/DeepSeek-V4-Flash-UD-IQ3_XXS.gguf"))
+		})
+
+		It("falls back to the last compatible GGUF when requested quantizations do not match", func() {
+			details := Details{
+				URI:         "huggingface://antirez/deepseek-v4-gguf",
+				Preferences: json.RawMessage(`{"quantizations":"F16"}`),
+				HuggingFace: &hfapi.ModelDetails{Files: []hfapi.ModelFile{
+					{Path: "DeepSeek-V4-Flash-BF16.gguf", URL: "https://huggingface.co/antirez/deepseek-v4-gguf/resolve/main/DeepSeek-V4-Flash-BF16.gguf"},
+					{Path: "README.md", URL: "https://huggingface.co/antirez/deepseek-v4-gguf/resolve/main/README.md"},
+					{Path: "DeepSeek-V4-Flash-Q8_0.gguf", URL: "https://huggingface.co/antirez/deepseek-v4-gguf/resolve/main/DeepSeek-V4-Flash-Q8_0.gguf"},
+				}},
+			}
+
+			cfg, err := importer.Import(details)
+
+			Expect(err).NotTo(HaveOccurred())
+			Expect(cfg.Files).To(HaveLen(1))
+			Expect(cfg.Files[0].URI).To(Equal("https://huggingface.co/antirez/deepseek-v4-gguf/resolve/main/DeepSeek-V4-Flash-Q8_0.gguf"))
+		})
+
+		It("rejects a repository URI when no HuggingFace file metadata is available", func() {
+			_, err := importer.Import(Details{URI: "huggingface://antirez/deepseek-v4-gguf"})
+
+			Expect(err).To(MatchError(ContainSubstring("compatible GGUF")))
+		})
+
+		It("rejects repository metadata without a compatible GGUF", func() {
+			details := Details{
+				URI: "huggingface://antirez/deepseek-v4-gguf",
+				HuggingFace: &hfapi.ModelDetails{Files: []hfapi.ModelFile{
+					{Path: "README.md", URL: "https://huggingface.co/antirez/deepseek-v4-gguf/resolve/main/README.md"},
+					{Path: "unrelated.gguf", URL: "https://huggingface.co/antirez/deepseek-v4-gguf/resolve/main/unrelated.gguf"},
+				}},
+			}
+
+			cfg, err := importer.Import(details)
+
+			Expect(err).To(MatchError(ContainSubstring("compatible GGUF")))
+			Expect(cfg.Files).To(BeEmpty())
+		})
 	})
 })