feat(api): expose endpoint to get raw file

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

Author: mudler

README.md

@@ -255,6 +255,14 @@ curl -X GET $BASE_URL/collections/myCollection/entries/file.txt
 
 Returns `collection`, `entry`, `chunks` (array of `id`, `content`, `metadata`), and `count`.
 
+- **Get Entry Raw File**:
+
+```sh
+curl -X GET $BASE_URL/collections/myCollection/entries/file.pdf/raw
+```
+
+Returns the original uploaded binary file with the appropriate Content-Type header.
+
 - **Search Collection**:
 
 ```sh

pkg/client/client.go

@@ -143,6 +143,24 @@ func (c *Client) GetEntryContent(collection, entry string) ([]EntryChunk, error)
 	return result.Data.Chunks, nil
 }
 
+// GetEntryRawFile returns the original uploaded binary file as a ReadCloser.
+// The caller is responsible for closing the returned ReadCloser.
+func (c *Client) GetEntryRawFile(collection, entry string) (io.ReadCloser, error) {
+	apiURL := fmt.Sprintf("%s/api/collections/%s/entries/%s/raw", c.BaseURL, collection, url.PathEscape(entry))
+
+	resp, err := http.Get(apiURL)
+	if err != nil {
+		return nil, fmt.Errorf("requesting raw file: %w", err)
+	}
+
+	if resp.StatusCode != http.StatusOK {
+		resp.Body.Close()
+		return nil, fmt.Errorf("raw file download failed (status %d)", resp.StatusCode)
+	}
+
+	return resp.Body, nil
+}
+
 // DeleteEntry deletes an Entry in a collection and return the entries left
 func (c *Client) DeleteEntry(collection, entry string) ([]string, error) {
 	url := fmt.Sprintf("%s/api/collections/%s/entry/delete", c.BaseURL, collection)

rag/persistency.go

@@ -235,6 +235,22 @@ func (db *PersistentKB) GetEntryContent(entry string) ([]types.Result, error) {
 	return results, nil
 }
 
+// GetEntryFilePath returns the filesystem path of the stored file for the given entry.
+func (db *PersistentKB) GetEntryFilePath(entry string) (string, error) {
+	db.Lock()
+	defer db.Unlock()
+
+	entry = filepath.Base(entry)
+	if _, ok := db.index[entry]; !ok {
+		return "", fmt.Errorf("entry not found: %s", entry)
+	}
+	fpath := filepath.Join(db.assetDir, entry)
+	if _, err := os.Stat(fpath); err != nil {
+		return "", fmt.Errorf("entry file not found: %s", entry)
+	}
+	return fpath, nil
+}
+
 // GetEntryFileContent returns the full content of the stored file (same text that was chunked, without overlap)
 // and the number of chunks it occupies. This avoids returning overlapping chunk content.
 func (db *PersistentKB) GetEntryFileContent(entry string) (content string, chunkCount int, err error) {
@@ -282,7 +298,11 @@ func (db *PersistentKB) storeFile(entry string, metadata map[string]string) erro
 	beforeCount := db.Engine.Count()
 	results, err := db.store(metadata, fileName)
 	if err != nil {
-		return fmt.Errorf("failed to store file: %w", err)
+		// File is already copied to assetDir. Index it with no chunks so it
+		// still appears in ListDocuments and can be served via GetEntryFilePath.
+		xlog.Warn("Chunking failed, storing file without chunks", "entry", entry, "error", err)
+		db.index[fileName] = nil
+		return db.save()
 	}
 	afterCount := db.Engine.Count()
 	xlog.Info("Stored file", "entry", entry, "fileName", fileName, "results_count", len(results), "count_before", beforeCount, "count_after", afterCount, "added_count", afterCount-beforeCount)

routes.go

@@ -127,6 +127,7 @@ func registerAPIRoutes(e *echo.Echo, openAIClient *openai.Client, maxChunkingSiz
 	e.GET("/api/collections", listCollections)
 	e.GET("/api/collections/:name/entries", listFiles(collections))
 	e.GET("/api/collections/:name/entries/:entry", getEntryContent(collections))
+	e.GET("/api/collections/:name/entries/:entry/raw", getEntryRawFile(collections))
 	e.POST("/api/collections/:name/search", search(collections))
 	e.POST("/api/collections/:name/reset", reset(collections))
 	e.DELETE("/api/collections/:name/entry/delete", deleteEntryFromCollection(collections))
@@ -309,6 +310,30 @@ func getEntryContent(collections collectionList) func(c echo.Context) error {
 	}
 }
 
+// getEntryRawFile returns the original uploaded binary file.
+func getEntryRawFile(collections collectionList) func(c echo.Context) error {
+	return func(c echo.Context) error {
+		name := c.Param("name")
+		collection, exists := collections[name]
+		if !exists {
+			return c.JSON(http.StatusNotFound, errorResponse(ErrCodeNotFound, "Collection not found", fmt.Sprintf("Collection '%s' does not exist", name)))
+		}
+
+		entryParam := c.Param("entry")
+		entry, err := url.PathUnescape(entryParam)
+		if err != nil {
+			entry = entryParam
+		}
+
+		fpath, err := collection.GetEntryFilePath(entry)
+		if err != nil {
+			return c.JSON(http.StatusNotFound, errorResponse(ErrCodeNotFound, "Entry not found", fmt.Sprintf("Entry '%s' does not exist in collection '%s'", entry, name)))
+		}
+
+		return c.File(fpath)
+	}
+}
+
 // uploadFile handles uploading files to a collection
 func uploadFile(collections collectionList, fileAssets string) func(c echo.Context) error {
 	return func(c echo.Context) error {