feat: add tag-based filtering to dataset hub search

DashAI/back/api/api_v1/endpoints/dataset_source.py

@@ -52,6 +52,7 @@ async def search_datasets(
     q: str = Query(default="", description="Search query"),
     limit: int = Query(default=20, ge=1, le=100),
     cursor: str = Query(default="", description="Pagination cursor from previous page"),
+    tags: list[str] = Query(default=[]),
     registry: "ComponentRegistry" = Depends(lambda: di["component_registry"]),
 ) -> Dict[str, Any]:
     """Search for datasets in a registered source.
@@ -67,6 +68,9 @@ async def search_datasets(
     cursor : str
         Opaque pagination token returned by the previous call.  Empty string
         means first page.
+    tags : list[str]
+        Repeated tag filter strings (e.g. ``?tags=nlp&tags=tabular``).  Passed
+        through to the datasource via ``**filters``.
     registry : ComponentRegistry
         Injected component registry.
 
@@ -76,7 +80,7 @@ async def search_datasets(
         ``{"results": [...], "next_cursor": str | null}``
     """
     source = _get_source(source_name, registry)
-    page = source.search(q, limit=limit, cursor=cursor or None)
+    page = source.search(q, limit=limit, cursor=cursor or None, tags=tags)
     return {
         "results": [
             {

DashAI/back/dataset_sources/huggingface_dataset_source.py

@@ -99,7 +99,9 @@ def search(
             Pagination cursor returned by the previous call (encoded numeric
             offset).  ``None`` fetches the first page.
         **filters : Any
-            Unused; reserved for future tag/task filters.
+            Supported keys:
+              tags (list[str]): Filter by HuggingFace tag strings.
+                Passed directly as the ``filter`` argument to ``list_datasets``.
 
         Returns
         -------
@@ -110,11 +112,13 @@ def search(
 
         try:
             offset = int(cursor) if cursor else 0
+            tags: list[str] = filters.get("tags") or []
 
             iterator = HfApi().list_datasets(
                 search=query or None,
                 full=True,
                 limit=offset + limit + 1,
+                filter=tags if tags else None,
             )
             window = list(islice(iterator, offset, offset + limit + 1))
             has_next = len(window) > limit

DashAI/back/dataset_sources/openml_dataset_source.py

@@ -146,7 +146,9 @@ def search(
             Opaque pagination token (encodes the numeric offset).  ``None``
             fetches the first page.
         **filters : Any
-            Unused; reserved for future filters.
+            Supported keys:
+              tags (list[str]): Filter by OpenML tag. Only the first tag is
+                used; OpenML's API accepts a single ``tag`` parameter.
 
         Returns
         -------
@@ -157,12 +159,21 @@ def search(
             offset = int(cursor) if cursor else 0
             list_kwargs: dict[str, Any] = {
                 "offset": offset,
-                "size": limit,
+                "size": limit + 1,
                 "status": "active",
                 "output_format": "dataframe",
             }
             if query:
                 list_kwargs["data_name"] = query
+            tags: list[str] = filters.get("tags") or []
+            if tags:
+                list_kwargs["tag"] = tags[0]
+                if len(tags) > 1:
+                    log.warning(
+                        "OpenML supports only one tag filter; using %r, ignoring %r",
+                        tags[0],
+                        tags[1:],
+                    )
 
             result = openml.datasets.list_datasets(**list_kwargs)
 
@@ -188,22 +199,25 @@ def _meta(did: str) -> tuple[str, tuple[str, ...]] | None:
             entries = []
             for row, did, meta in zip(rows, ids, metas):
                 description = ""
-                tags: list[str] = []
+                dataset_tags: list[str] = []
                 if meta is not None:
                     description, real_tags = meta
-                    tags = list(real_tags)
+                    dataset_tags = list(real_tags)
                 entries.append(
                     DatasetEntry(
                         id=did,
                         name=row.get("name", "") or "",
                         description=description,
-                        tags=tags,
+                        tags=dataset_tags,
                         size_bytes=None,
                         url=f"https://www.openml.org/d/{did}",
                         source=self.__class__.__name__,
                     )
                 )
-            next_cursor = str(offset + limit) if len(entries) == limit else None
+            has_next = len(entries) > limit
+            if has_next:
+                entries = entries[:limit]
+            next_cursor = str(offset + limit) if has_next else None
             return SearchPage(entries=entries, next_cursor=next_cursor)
         except Exception:
             log.exception("Error searching OpenML datasets")

DashAI/back/dataset_sources/zenodo_dataset_source.py

@@ -19,6 +19,11 @@
 _ZENODO_API = "https://zenodo.org/api"
 
 
+def _escape_lucene_phrase(s: str) -> str:
+    """Escape backslash and double-quote inside a Lucene quoted-string phrase."""
+    return s.replace("\\", "\\\\").replace('"', '\\"')
+
+
 def _strip_html(text: str) -> str:
     """Remove HTML tags from a string.
 
@@ -122,7 +127,10 @@ def search(
             Opaque pagination token (encodes the page number as a string).
             Pass ``None`` to fetch the first page.
         **filters : Any
-            Unused; reserved for future filters.
+            Supported keys:
+              tags (list[str]): Filter by Zenodo keywords. Each tag is
+                appended to the query using Lucene ``keywords:"<tag>"``
+                syntax (AND logic).
 
         Returns
         -------
@@ -131,8 +139,18 @@ def search(
         """
         try:
             page = int(cursor) if cursor else 1
+            tags: list[str] = filters.get("tags") or []
+            keyword_clauses = " AND ".join(
+                f'keywords:"{_escape_lucene_phrase(t)}"' for t in tags
+            )
+            if query and tags:
+                zenodo_q = f"({query}) AND {keyword_clauses}"
+            elif tags:
+                zenodo_q = keyword_clauses
+            else:
+                zenodo_q = query
             params: dict[str, Any] = {
-                "q": query,
+                "q": zenodo_q,
                 "type": "dataset",
                 "page": page,
                 "size": limit,

DashAI/front/src/api/api.ts

@@ -4,6 +4,17 @@ import i18n from "i18next";
 
 const api: AxiosInstance = axios.create({
   baseURL: process.env.REACT_APP_API_URL,
+  paramsSerializer: (params) => {
+    const sp = new URLSearchParams();
+    for (const [key, val] of Object.entries(params)) {
+      if (Array.isArray(val)) {
+        val.forEach((v) => sp.append(key, String(v)));
+      } else if (val !== null && val !== undefined) {
+        sp.append(key, String(val));
+      }
+    }
+    return sp.toString();
+  },
 });
 
 api.interceptors.request.use((config) => {

DashAI/front/src/api/hub.ts

@@ -35,10 +35,11 @@ export const searchDatasets = async (
   query: string,
   limit = 20,
   cursor: string | null = null,
+  tags: string[] = [],
 ): Promise<DatasetSearchPage> => {
   const response = await api.get<DatasetSearchPage>(
     `${hubEndpoint}/${sourceName}/search`,
-    { params: { q: query, limit, cursor: cursor ?? "" } },
+    { params: { q: query, limit, cursor: cursor ?? "", tags } },
   );
   return response.data;
 };

DashAI/front/src/components/hub/DatasetGrid.jsx

@@ -2,11 +2,13 @@ import { useCallback, useEffect, useRef, useState } from "react";
 import {
   Box,
   Button,
+  Chip,
   CircularProgress,
   InputAdornment,
   TextField,
   Typography,
 } from "@mui/material";
+import LocalOfferIcon from "@mui/icons-material/LocalOffer";
 import SearchIcon from "@mui/icons-material/Search";
 import { useTranslation } from "react-i18next";
 import { searchDatasets } from "../../api/hub";
@@ -31,21 +33,26 @@ export default function DatasetGrid({
 }) {
   const { t } = useTranslation(["hub", "common"]);
   const [query, setQuery] = useState("");
+  const [tags, setTags] = useState([]);
+  const [tagInput, setTagInput] = useState("");
   const [datasets, setDatasets] = useState([]);
   const [nextCursor, setNextCursor] = useState(null);
   const [hasMore, setHasMore] = useState(false);
   const [loading, setLoading] = useState(false);
   const [loadingMore, setLoadingMore] = useState(false);
   const debounceRef = useRef(null);
+  const reqIdRef = useRef(0);
 
   const loadPage = useCallback(
-    (q, cursor, append) => {
+    (q, activeTags, cursor, append) => {
       if (!sourceName) return;
       if (append) setLoadingMore(true);
       else setLoading(true);
+      const reqId = ++reqIdRef.current;
 
-      searchDatasets(sourceName, q, PAGE_SIZE, cursor)
+      searchDatasets(sourceName, q, PAGE_SIZE, cursor, activeTags)
         .then(({ results, next_cursor }) => {
+          if (reqId !== reqIdRef.current) return;
           setDatasets((prev) =>
             append
               ? [
@@ -59,11 +66,13 @@ export default function DatasetGrid({
           setHasMore(next_cursor !== null);
         })
         .catch(() => {
+          if (reqId !== reqIdRef.current) return;
           if (!append) setDatasets([]);
           setNextCursor(null);
           setHasMore(false);
         })
         .finally(() => {
+          if (reqId !== reqIdRef.current) return;
           if (append) setLoadingMore(false);
           else setLoading(false);
         });
@@ -75,20 +84,50 @@ export default function DatasetGrid({
     setDatasets([]);
     setNextCursor(null);
     setQuery("");
+    setTags([]);
+    setTagInput("");
     setHasMore(false);
-    if (sourceName) loadPage("", null, false);
+    if (sourceName) loadPage("", [], null, false);
   }, [sourceName]);
 
   const handleQueryChange = (e) => {
     const val = e.target.value;
     setQuery(val);
     setNextCursor(null);
     clearTimeout(debounceRef.current);
-    debounceRef.current = setTimeout(() => loadPage(val, null, false), 400);
+    debounceRef.current = setTimeout(
+      () => loadPage(val, tags, null, false),
+      400,
+    );
+  };
+
+  const handleTagInputKeyDown = (e) => {
+    if (e.key === "Enter" || e.key === ",") {
+      e.preventDefault();
+      const trimmed = tagInput.trim();
+      if (trimmed && !tags.includes(trimmed)) {
+        const newTags = [...tags, trimmed];
+        setTags(newTags);
+        setTagInput("");
+        setNextCursor(null);
+        clearTimeout(debounceRef.current);
+        loadPage(query, newTags, null, false);
+      } else {
+        setTagInput("");
+      }
+    }
+  };
+
+  const handleRemoveTag = (tagToRemove) => {
+    const newTags = tags.filter((t) => t !== tagToRemove);
+    setTags(newTags);
+    setNextCursor(null);
+    clearTimeout(debounceRef.current);
+    loadPage(query, newTags, null, false);
   };
 
   const handleLoadMore = () => {
-    loadPage(query, nextCursor, true);
+    loadPage(query, tags, nextCursor, true);
   };
 
   if (!sourceName) {
@@ -137,6 +176,39 @@ export default function DatasetGrid({
         }}
       />
 
+      <Box sx={{ display: "flex", flexDirection: "column", gap: 0.5 }}>
+        <TextField
+          size="small"
+          fullWidth
+          placeholder={t("hub:tagFilterPlaceholder")}
+          value={tagInput}
+          onChange={(e) => setTagInput(e.target.value)}
+          onKeyDown={handleTagInputKeyDown}
+          slotProps={{
+            input: {
+              startAdornment: (
+                <InputAdornment position="start">
+                  <LocalOfferIcon fontSize="small" />
+                </InputAdornment>
+              ),
+            },
+          }}
+        />
+        {tags.length > 0 && (
+          <Box sx={{ display: "flex", flexWrap: "wrap", gap: 0.5 }}>
+            {tags.map((tag) => (
+              <Chip
+                key={tag}
+                label={tag}
+                size="small"
+                variant="outlined"
+                onDelete={() => handleRemoveTag(tag)}
+              />
+            ))}
+          </Box>
+        )}
+      </Box>
+
       {loading ? (
         <Box sx={{ display: "flex", justifyContent: "center", pt: 4 }}>
           <CircularProgress />
@@ -160,7 +232,7 @@ export default function DatasetGrid({
           <Box
             sx={{
               display: "grid",
-              gap: 1.5,
+              gap: 2,
               gridTemplateColumns: {
                 xs: "1fr",
                 md: "repeat(2, 1fr)",

DashAI/front/src/utils/i18n/locales/de/hub.json

@@ -60,5 +60,6 @@
   "datasetDetails": "Dateidetails",
   "checkQueue": "Details in der Aufgabenwarteschlange prüfen",
   "deleteSuccess": "Datei \"{{name}}\" gelöscht.",
-  "deleteError": "Fehler beim Löschen der Datei."
+  "deleteError": "Fehler beim Löschen der Datei.",
+  "tagFilterPlaceholder": "Nach Tag filtern, Enter drücken zum Hinzufügen..."
 }

DashAI/front/src/utils/i18n/locales/en/hub.json

@@ -60,5 +60,6 @@
   "datasetDetails": "Datafile Details",
   "checkQueue": "Check the job queue for details",
   "deleteSuccess": "Datafile \"{{name}}\" deleted.",
-  "deleteError": "Failed to delete datafile."
+  "deleteError": "Failed to delete datafile.",
+  "tagFilterPlaceholder": "Filter by tag, press Enter to add..."
 }

DashAI/front/src/utils/i18n/locales/es/hub.json

@@ -60,5 +60,6 @@
   "datasetDetails": "Detalles del Datafile",
   "checkQueue": "Revisa la cola de trabajos para más detalles",
   "deleteSuccess": "Datafile \"{{name}}\" eliminado.",
-  "deleteError": "Error al eliminar el datafile."
+  "deleteError": "Error al eliminar el datafile.",
+  "tagFilterPlaceholder": "Filtrar por etiqueta, presiona Enter para agregar..."
 }

DashAI/front/src/utils/i18n/locales/pt/hub.json

@@ -60,5 +60,6 @@
   "datasetDetails": "Detalhes do Arquivo",
   "checkQueue": "Verifique a fila de trabalhos para mais detalhes",
   "deleteSuccess": "Arquivo \"{{name}}\" excluído.",
-  "deleteError": "Falha ao excluir o arquivo."
+  "deleteError": "Falha ao excluir o arquivo.",
+  "tagFilterPlaceholder": "Filtrar por tag, pressione Enter para adicionar..."
 }

DashAI/front/src/utils/i18n/locales/zh/hub.json

@@ -60,5 +60,6 @@
   "datasetDetails": "数据文件详情",
   "checkQueue": "在任务队列中查看详情",
   "deleteSuccess": "数据文件 \"{{name}}\" 已删除。",
-  "deleteError": "删除数据文件失败。"
+  "deleteError": "删除数据文件失败。",
+  "tagFilterPlaceholder": "按标签过滤，按 Enter 添加..."
 }