feat: enhance AutoBIDSify, AI search display, and preview reliability

src/components/SearchPage/DatasetCard.tsx

@@ -40,6 +40,7 @@ interface DatasetCardProps {
     value: {
       name?: string;
       readme?: string;
+      aisummary?: string | Record<string, string>;
       modality?: string[];
       subj?: string[];
       info?: {
@@ -94,15 +95,30 @@ const containsKeyword = (text?: string, kw?: string) => {
   return words.some((w) => t.includes(w));
 };
 
+/** AISummary can be a plain string OR a sectioned object
+ *  ({Introduction, Methods, Results, Conclusion}). Flatten to one string so we
+ *  can search/snippet it — mirrors the detail page's handling. */
+const flattenAiSummary = (s: any): string | undefined => {
+  if (!s) return undefined;
+  if (typeof s === "string") return s;
+  if (typeof s === "object")
+    return Object.values(s).filter(Boolean).join(" ");
+  return undefined;
+};
+
 /** Find a short snippet in secondary fields if not already visible */
 function findMatchSnippet(
   v: any,
   kw?: string
 ): { label: string; html: string } | null {
   if (!kw) return null;
 
-  // Which fields to scan (can add/remove fields here)
+  // Which fields to scan (can add/remove fields here).
+  // "AI Summary" is first so a topic-word hit in the generated summary is the
+  // explanation shown (its text lives in the dbinfo view's `aisummary` field).
   const CANDIDATE_FIELDS: Array<[string, (v: any) => string | undefined]> = [
+    ["AI Summary", (v) => flattenAiSummary(v?.aisummary)],
+    ["Description", (v) => v?.info?.Description],
     ["Acknowledgements", (v) => v?.info?.Acknowledgements],
     [
       "Funding",
@@ -377,7 +393,8 @@ const DatasetCard: React.FC<DatasetCardProps> = ({
                 paragraph
                 sx={{ textOverflow: "ellipsis" }}
               >
-                <strong>Summary:</strong> {highlightKeyword(readme, keyword)}
+                <strong>README:</strong> {highlightKeyword(readme, keyword)}
+                {readme.length >= 256 ? "…" : ""}
               </Typography>
             )}
           </Stack>

src/components/User/Dashboard/DatasetOrganizer/LLMPanel.tsx

@@ -1590,6 +1590,7 @@ const LLMPanel: React.FC<LLMPanelProps> = ({
                   </MenuItem>
                   <MenuItem value="mri">MRI</MenuItem>
                   <MenuItem value="nirs">NIRS</MenuItem>
+                  <MenuItem value="eeg">EEG</MenuItem>
                   <MenuItem value="mixed">Mixed</MenuItem>
                 </Select>
                 {modalityError && (
@@ -1605,7 +1606,7 @@ const LLMPanel: React.FC<LLMPanelProps> = ({
 
               <TextField
                 label="Describe your dataset (optional)"
-                placeholder='e.g. "DICOM files from 2 subjects, one male one female"'
+                placeholder='e.g. "DICOM files from 2 subjects, one male one female" or "EDF recordings from 5 subjects, rest and task runs"'
                 value={describeText}
                 onChange={(e) => setDescribeText(e.target.value)}
                 size="small"

src/components/User/Dashboard/DatasetOrganizer/utils/fileAnalyzers.ts

@@ -81,6 +81,19 @@ export const categorizeFile = (file: FileItem): string => {
   )
     return "nirs";
 
+  // eeg — mirrors EEG_EXT = {'.edf', '.vhdr', '.set', '.bdf'}
+  if ([".edf", ".vhdr", ".set", ".bdf"].some((e) => name.endsWith(e)))
+    return "eeg";
+
+  // eeg_aux — mirrors EEG_AUX_EXT {'.vmrk', '.eeg', '.fdt'}
+  //           + EEG_EVENT_EXT {'.event', '.events', '.evt', '.mrk'}
+  if (
+    [".vmrk", ".eeg", ".fdt", ".event", ".events", ".evt", ".mrk"].some((e) =>
+      name.endsWith(e)
+    )
+  )
+    return "eeg_aux";
+
   // mri — mirrors MRI_EXT = {'.nii', '.dcm'} + .nii.gz
   if (
     name.endsWith(".nii.gz") ||
@@ -140,6 +153,13 @@ export const detectModality = (files: FileItem[]): string => {
     files.some((f) => f.name.endsWith(".snirf"))
   )
     return "nirs";
+  // eeg — fileType keys from fileProcessors.ts getFileType()
+  if (
+    counts.eegEdf > 0 ||
+    counts.eegBrainvision > 0 ||
+    counts.eegEeglab > 0
+  )
+    return "eeg";
   return "mixed";
 };
 

src/components/User/Dashboard/DatasetOrganizer/utils/fileProcessors.ts

@@ -30,6 +30,9 @@ export const getFileType = (name: string): string => {
     matlab: ["mat"],
     dicom: ["dcm"],
     nirs: ["nirs"],
+    eegEdf: ["edf", "bdf"],
+    eegBrainvision: ["vhdr", "vmrk", "eeg"],
+    eegEeglab: ["set", "fdt"],
   };
 
   for (const [type, extensions] of Object.entries(fileTypes)) {

src/components/User/Dashboard/DatasetOrganizer/utils/llm.ts

@@ -375,6 +375,15 @@ fNIRS FORMATS (modality: nirs):
   • Homer3 (.nirs)         → convert_to: snirf
   • MATLAB (.mat)          → convert_to: snirf
 
+EEG FORMATS (modality: eeg):
+  • EDF/EDF+ (.edf)        → format_ready: true  (copy directly)
+  • BrainVision (.vhdr)    → format_ready: true  (copy directly)
+  • EEGLAB (.set)          → format_ready: true  (copy directly)
+  • Biosemi (.bdf)         → format_ready: true  (copy directly)
+  CRITICAL: EEG files are NEVER converted. Always format_ready: true, convert_to: none.
+  CRITICAL: EEG bids_template MUST end with '_eeg.<original_ext>' (e.g. '_eeg.edf').
+            NEVER use NIfTI suffixes (T1w, T2w, bold) for EEG data.
+
 ═══════════════════════════════════════════════════════════════════════
 SUBJECT IDENTIFICATION — MOST IMPORTANT STEP
 ═══════════════════════════════════════════════════════════════════════
@@ -495,6 +504,23 @@ For MRI: use acq- to distinguish different scan series from same subject.
   VHFCT1mm-Ankle.dcm → acq-ankle_T1w
   VHFCT1mm-Head.dcm  → acq-head_T1w
 
+For EEG: infer task and run from filename suffixes or directory names.
+  RULE 1 — If each subject has multiple EEG files, each file is a separate scan.
+    Identify what differs between files of the same subject (suffix, keyword, directory).
+    Map each variant to a distinct task- or run- label from user description.
+    If task labels cannot be inferred, use run-1, run-2, run-N.
+  RULE 2 — Create one mapping entry per unique file variant across subjects.
+  RULE 3 — BIDS directory for EEG is always 'eeg/', never 'anat/' or 'nirs/'.
+  RULE 4 — BIDS filename suffix is always '_eeg' + original extension.
+
+EEG FILENAME EXAMPLES (CRITICAL — follow exactly):
+  ✓ sub-01_task-rest_eeg.edf
+  ✓ sub-01_task-arithmetic_eeg.edf
+  ✓ sub-01_run-1_eeg.edf
+  ✗ sub-01_T1w.nii.gz          ← NEVER for EEG
+  ✗ sub-01_unknown.nii.gz      ← NEVER for EEG
+  ✗ sub-01_bold.nii.gz         ← NEVER for EEG
+
 ═══════════════════════════════════════════════════════════════════════
 OUTPUT FORMAT
 ═══════════════════════════════════════════════════════════════════════
@@ -527,6 +553,20 @@ mappings:
       - match_pattern: '.*'
         bids_template: 'sub-X_task-walking_nirs.snirf'
 
+  # EEG example — when each subject has ONE edf file:
+  - modality: eeg
+    match: ['**/*.edf']
+    exclude: []
+    format_ready: true
+    convert_to: none
+    filename_rules:
+      - match_pattern: '.*'
+        bids_template: 'sub-X_task-rest_eeg.edf'
+
+  # EEG example — when each subject has MULTIPLE edf files (different tasks/runs):
+  # Create one mapping entry per task/run, use match_pattern to distinguish them.
+  # The match_pattern must be derived from what actually differs in the filenames.
+
 OUTPUT: Raw YAML only (no markdown, no explanation)`;
 
 export const PROMPT_MAT_SNIRF_MAPPING = `You are an fNIRS data format expert.

src/components/User/Dashboard/DatasetOrganizer/utils/plannerHelpers.ts

@@ -70,9 +70,10 @@ export interface BuildBidsPlanResult {
 // Mirrors planner.py _DATA_EXTS, evidence.py TRIO_NAMES
 // ============================================================================
 
+// Mirrors planner.py _DATA_EXTS — EEG primary formats (edf/vhdr/set/bdf)
+// are data files; EEG aux (.vmrk/.eeg/.fdt) are handled separately, not here.
 const DATA_EXTENSIONS =
-  //   /\.(snirf|nii|nii\.gz|dcm|mat|nirs|jnii|bnii|h5|hdf5|edf|bdf)$/i;
-  /\.(snirf|nii|nii\.gz|dcm|mat|nirs|jnii|bnii)$/i;
+  /\.(snirf|nii|nii\.gz|dcm|mat|nirs|jnii|bnii|edf|vhdr|set|bdf)$/i;
 
 const TRIO_FILENAMES = new Set([
   "dataset_description.json",

src/redux/projects/types/projects.interface.ts

@@ -14,6 +14,9 @@ export interface FileItem {
     | "matlab"
     | "dicom"
     | "nirs"
+    | "eegEdf"
+    | "eegBrainvision"
+    | "eegEeglab"
     | "array"
     | "other";
   content?: string;