From f37774ff36915cd1da00219528cb3521333b7844 Mon Sep 17 00:00:00 2001
From: ValueOn AG <p.motsch@valueon.ch>
Date: Sun, 17 May 2026 20:38:40 +0200
Subject: [PATCH] db connection pooling and rag limit transparency

---
 src/api/connectionApi.ts              | 11 +++++-
 src/pages/RagInventoryPage.module.css | 16 ++++++++
 src/pages/RagInventoryPage.tsx        | 57 +++++++++++++++++++++++++--
 3 files changed, 79 insertions(+), 5 deletions(-)

diff --git a/src/api/connectionApi.ts b/src/api/connectionApi.ts
index 72b5097..828d2f1 100644
--- a/src/api/connectionApi.ts
+++ b/src/api/connectionApi.ts
@@ -348,6 +348,9 @@ export interface RagDataSourceDto {
   ragIndexEnabled: boolean;
   neutralize: boolean;
   lastIndexed: number | null;
+  /** Distinct files indexed for this DataSource (one row per source document). */
+  fileCount: number;
+  /** Embedding-sized text fragments (one per ContentChunk row, ~400 tokens each). */
   chunkCount: number;
 }
 
@@ -358,6 +361,7 @@ export interface RagConnectionDto {
   knowledgeIngestionEnabled: boolean;
   preferences: KnowledgePreferences;
   dataSources: RagDataSourceDto[];
+  totalFiles: number;
   totalChunks: number;
   runningJobs: { jobId: string; progress: number; progressMessage: string }[];
   lastError?: { jobId: string; errorMessage: string; finishedAt: number | null } | null;
@@ -369,12 +373,17 @@ export interface RagConnectionDto {
     skippedPolicy: number;
     failed: number;
     durationMs: number;
+    /** Name of the first budget that bit (e.g. "maxBytes", "maxItems", "maxTasks"); null if walk completed naturally. */
+    stoppedAtLimit?: string | null;
+    /** Effective limits used by the walker, for showing the value next to the limit name. */
+    limits?: Record<string, number>;
+    bytesProcessed?: number;
   } | null;
 }
 
 export interface RagInventoryDto {
   connections: RagConnectionDto[];
-  totals: { chunks: number; bytes?: number };
+  totals: { files: number; chunks: number; bytes?: number };
 }
 
 export interface RagActiveJobDto {
diff --git a/src/pages/RagInventoryPage.module.css b/src/pages/RagInventoryPage.module.css
index a04d275..8a72cd0 100644
--- a/src/pages/RagInventoryPage.module.css
+++ b/src/pages/RagInventoryPage.module.css
@@ -220,6 +220,22 @@
   opacity: 0.85;
 }
 
+/* Sync finished, but a hard limit (maxBytes/maxItems/...) cut the walk short.
+   Amber, not red — the data we DID index is valid; the user just needs to
+   know more would have been indexed without the limit. */
+.partialBanner {
+  display: flex;
+  align-items: center;
+  gap: 8px;
+  padding: 8px 12px;
+  background: #fffbeb;
+  border: 1px solid #fcd34d;
+  border-radius: 6px;
+  margin-bottom: 8px;
+  font-size: 0.8125rem;
+  color: #92400e;
+}
+
 .reindexBtn {
   display: flex;
   align-items: center;
diff --git a/src/pages/RagInventoryPage.tsx b/src/pages/RagInventoryPage.tsx
index 92dad83..7731773 100644
--- a/src/pages/RagInventoryPage.tsx
+++ b/src/pages/RagInventoryPage.tsx
@@ -139,6 +139,21 @@ export const RagInventoryPage: React.FC = () => {
     return `${Math.floor(ms / 60000)}m ${Math.floor((ms % 60000) / 1000)}s`;
   }, []);
 
+  /** Render the budget value next to its name. Bytes get MB units so the user
+   *  immediately recognises the 200 MB default; everything else stays raw. */
+  const _formatLimit = useCallback((name: string, budget: number | undefined, bytesProcessed: number | undefined): string => {
+    if (budget == null) return name;
+    if (name === 'maxBytes') {
+      const mb = Math.round(budget / 1024 / 1024);
+      const procMb = bytesProcessed != null ? ` (${(bytesProcessed / 1024 / 1024).toFixed(0)} MB ${t('verarbeitet')})` : '';
+      return `${name}=${mb} MB${procMb}`;
+    }
+    if (name === 'maxFileSize') {
+      return `${name}=${Math.round(budget / 1024 / 1024)} MB`;
+    }
+    return `${name}=${budget}`;
+  }, [t]);
+
   const scopeOptions = useMemo(() => {
     const opts: { value: string; label: string }[] = [
       { value: 'personal', label: t('Meine Verbindungen') },
@@ -193,7 +208,9 @@ export const RagInventoryPage: React.FC = () => {
       {inventory && (
         <div className={styles.content}>
           <div className={styles.totals}>
-            <span className={styles.totalLabel}>{t('Total Chunks')}:</span>
+            <span className={styles.totalLabel}>{t('Total Dateien')}:</span>
+            <strong className={styles.totalValue}>{inventory.totals?.files ?? 0}</strong>
+            <span className={styles.totalLabel} title={t('Embedding-Fragmente (~400 Tokens), die der RAG-Retrieval trifft')}>{t('Total Chunks')}:</span>
             <strong className={styles.totalValue}>{inventory.totals?.chunks ?? 0}</strong>
             {inventory.totals?.bytes != null && inventory.totals.bytes > 0 && (
               <span className={styles.totalBytes}>{(inventory.totals.bytes / 1024 / 1024).toFixed(1)} MB</span>
@@ -205,8 +222,13 @@ export const RagInventoryPage: React.FC = () => {
               <div className={styles.connectionHeader}>
                 <span className={styles.authority}>{conn.authority}</span>
                 <span className={styles.email}>{conn.externalEmail}</span>
-                {conn.totalChunks > 0 && (
-                  <span className={styles.connChunks}>{conn.totalChunks} chunks</span>
+                {(conn.totalFiles > 0 || conn.totalChunks > 0) && (
+                  <span
+                    className={styles.connChunks}
+                    title={t('Embedding-Fragmente (~400 Tokens), die der RAG-Retrieval trifft')}
+                  >
+                    {t('{f} Dateien · {c} Chunks', { f: conn.totalFiles, c: conn.totalChunks })}
+                  </span>
                 )}
                 <button
                   className={styles.consentToggle}
@@ -261,6 +283,28 @@ export const RagInventoryPage: React.FC = () => {
                     s.skippedPolicy > 0 ? t('{n} übersprungen', { n: s.skippedPolicy }) : null,
                     s.failed > 0 ? t('{n} fehler', { n: s.failed }) : null,
                   ].filter(Boolean).join(' · ');
+
+                  const stop = s.stoppedAtLimit;
+                  if (stop) {
+                    const budget = s.limits?.[stop];
+                    const limitText = _formatLimit(stop, budget, s.bytesProcessed);
+                    return (
+                      <div className={styles.partialBanner}>
+                        <FaExclamationTriangle />
+                        <span>
+                          <strong>{t('Sync abgeschlossen, Korpus aber unvollständig')}</strong> ({_formatRelative(okAt)})
+                          {' — '}
+                          {t('Limit {l} erreicht', { l: limitText })}.
+                          {stats && <> {stats}.</>}{' '}
+                          {t('Weitere Dateien wurden NICHT indexiert. Limit erhöhen oder DataSource enger eingrenzen, dann erneut starten.')}
+                        </span>
+                        <button className={styles.reindexBtn} onClick={() => _handleReindex(conn.id)} title={t('Erneut indexieren')}>
+                          <FaRedo size={12} /> {t('Erneut indexieren')}
+                        </button>
+                      </div>
+                    );
+                  }
+
                   return (
                     <div className={styles.successBanner}>
                       <FaCheckCircle />
@@ -293,7 +337,12 @@ export const RagInventoryPage: React.FC = () => {
                   <div key={ds.id} className={`${styles.dsRow} ${ds.ragIndexEnabled ? styles.dsActive : ''}`}>
                     <span className={styles.dsLabel}>{ds.label || ds.path}</span>
                     <span className={styles.dsType}>{ds.sourceType}</span>
-                    <span className={styles.dsChunks}>{ds.chunkCount} chunks</span>
+                    <span
+                      className={styles.dsChunks}
+                      title={t('{f} indizierte Dateien · {c} Embedding-Chunks (~400 Tokens)', { f: ds.fileCount, c: ds.chunkCount })}
+                    >
+                      {ds.fileCount} {t('Dateien')} · {ds.chunkCount} {t('Chunks')}
+                    </span>
                     <span className={styles.dsIndex}>{ds.ragIndexEnabled ? '\uD83E\uDDE0' : '\u2014'}</span>
                   </div>
                 ))}