@juspay/neurolink 9.88.6 → 9.88.8

package/dist/lib/utils/csvProcessor.js

@@ -337,6 +337,20 @@ function calculateDataQualityScore(columns, warnings, totalRows) {
 /**
  * Analyze all columns in parsed CSV data
  */
+/**
+ * Confidence for a CSV processing result, weighted by data quality.
+ *
+ * The result used to report a static `confidence: 100` regardless of how clean
+ * the data was. This ties confidence to `dataQualityScore` so it actually
+ * varies with quality factors (ragged rows, high null rate, type
+ * inconsistency): a cleanly-parsed CSV stays at 100, a messy one is pulled
+ * halfway toward its quality score. It never drops below 50 for a CSV that
+ * parsed — the *format* is certain, only the *data* is imperfect.
+ */
+function csvResultConfidence(dataQualityScore) {
+    const clamped = Math.max(0, Math.min(100, dataQualityScore));
+    return Math.round((100 + clamped) / 2);
+}
 function analyzeColumns(rows) {
     if (rows.length === 0) {
         return {
@@ -425,6 +439,33 @@ function detectHasHeaders(headerValues, dataRows) {
  * - Lines with significantly different delimiter count than line 2
  * - Lines that don't match CSV structure of subsequent lines
  */
+/** Strip a leading UTF-8 BOM (U+FEFF) if present. */
+function stripBom(text) {
+    return text.charCodeAt(0) === 0xfeff ? text.slice(1) : text;
+}
+/**
+ * Count commas that fall OUTSIDE double-quoted fields (RFC 4180, escaped-quote
+ * aware). A quoted comma is field content, not a column separator, so counting
+ * naively would misclassify a quoted-comma header line.
+ */
+function countUnquotedCommas(line) {
+    let count = 0;
+    let inQuotes = false;
+    for (let i = 0; i < line.length; i++) {
+        const ch = line[i];
+        if (ch === '"') {
+            if (inQuotes && line[i + 1] === '"') {
+                i++;
+                continue;
+            }
+            inQuotes = !inQuotes;
+        }
+        else if (ch === "," && !inQuotes) {
+            count++;
+        }
+    }
+    return count;
+}
 function isMetadataLine(lines) {
     if (!lines[0] || lines.length < 2) {
         return false;
@@ -434,8 +475,8 @@ function isMetadataLine(lines) {
     if (firstLine.match(/^sep=/i)) {
         return true;
     }
-    const firstCommaCount = (firstLine.match(/,/g) || []).length;
-    const secondCommaCount = (secondLine.match(/,/g) || []).length;
+    const firstCommaCount = countUnquotedCommas(firstLine);
+    const secondCommaCount = countUnquotedCommas(secondLine);
     if (firstCommaCount === 0 && secondCommaCount > 0) {
         return true;
     }
@@ -574,7 +615,7 @@ export class CSVProcessor {
                 content: limitedCSV,
                 mimeType: "text/csv",
                 metadata: {
-                    confidence: 100,
+                    confidence: csvResultConfidence(dataQualityScore),
                     size: content.length,
                     rowCount,
                     totalLines: limitedLines.length,
@@ -648,7 +689,7 @@ export class CSVProcessor {
             content: formatted,
             mimeType: "text/csv",
             metadata: {
-                confidence: 100,
+                confidence: csvResultConfidence(dataQualityScore),
                 size: content.length,
                 rowCount,
                 columnCount,
@@ -676,6 +717,9 @@ export class CSVProcessor {
      * @returns Array of row objects
      */
     static async parseCSVFile(filePath, maxRows = 1000) {
+        if (typeof filePath !== "string" || filePath.trim().length === 0) {
+            throw new Error("CSVProcessor.parseCSVFile: filePath must be a non-empty string");
+        }
         const clampedMaxRows = Math.max(1, Math.min(10000, maxRows));
         const fs = await import("fs");
         logger.debug("[CSVProcessor] Starting file parsing", {
@@ -701,6 +745,8 @@ export class CSVProcessor {
                 }
             });
             lineReader.on("end", () => resolve());
+            // Metadata sniffing is best-effort — a read error here must not crash.
+            lineReader.on("error", () => resolve());
         });
         await fileHandle.close();
         const hasMetadataLine = isMetadataLine(firstLines);
@@ -718,6 +764,12 @@ export class CSVProcessor {
                 source.destroy();
                 parser.destroy();
             };
+            // .pipe() does not forward source errors (disk read/permission failures)
+            // to the parser, so listen on the source directly or it throws unhandled.
+            source.on("error", (error) => {
+                abort();
+                reject(new Error(`[CSVProcessor] CSV file read failed after ${count} row(s) (${filePath}): ${error.message}`));
+            });
             source
                 .pipe(parser)
                 .on("data", (row) => {
@@ -739,7 +791,7 @@ export class CSVProcessor {
             })
                 .on("error", (error) => {
                 logger.error("[CSVProcessor] File parsing failed:", error);
-                reject(error);
+                reject(new Error(`[CSVProcessor] CSV parsing failed after ${count} row(s): ${error.message}`));
             });
         });
     }
@@ -752,13 +804,19 @@ export class CSVProcessor {
      * @returns Array of row objects
      */
     static async parseCSVString(csvString, maxRows = 1000) {
+        if (typeof csvString !== "string" || csvString.trim().length === 0) {
+            throw new Error("CSVProcessor.parseCSVString: csvString must be a non-empty string");
+        }
+        // Strip a leading UTF-8 BOM (common in Excel exports) so it does not glue
+        // onto the first column name and break downstream key lookups.
+        const normalized = stripBom(csvString);
         const clampedMaxRows = Math.max(1, Math.min(10000, maxRows));
         logger.debug("[CSVProcessor] Starting string parsing", {
-            inputLength: csvString.length,
+            inputLength: normalized.length,
             maxRows: clampedMaxRows,
         });
         // Detect and skip metadata line
-        const lines = splitCsvLines(csvString);
+        const lines = splitCsvLines(normalized);
         const hasMetadataLine = isMetadataLine(lines);
         const csvData = hasMetadataLine
             ? lines.slice(1).join("\n")
@@ -775,6 +833,13 @@ export class CSVProcessor {
                 source.destroy();
                 parser.destroy();
             };
+            // .pipe() does not forward source-stream errors to the destination, so a
+            // source failure must be listened for directly or it throws as an
+            // unhandled EventEmitter error and can crash the process.
+            source.on("error", (error) => {
+                abort();
+                reject(new Error(`[CSVProcessor] CSV source stream failed after ${count} row(s): ${error.message}`));
+            });
             source
                 .pipe(parser)
                 .on("data", (row) => {
@@ -792,7 +857,7 @@ export class CSVProcessor {
             })
                 .on("error", (error) => {
                 logger.error("[CSVProcessor] Parsing failed:", error);
-                reject(error);
+                reject(new Error(`[CSVProcessor] CSV parsing failed after ${count} row(s): ${error.message}`));
             });
         });
     }

package/dist/lib/utils/fileDetector.js

@@ -2166,13 +2166,38 @@ class ContentHeuristicStrategy {
             const hasUniformLengths = stdDev / avgLength < 0.75;
             return hasReasonableLengths && noBinaryChars && hasUniformLengths;
         }
-        // Count delimiters per line and check consistency
-        const delimRegex = delimiter === "|" ? /\|/g : new RegExp(delimiter, "g");
-        const counts = lines.map((line) => (line.match(delimRegex) || []).length);
+        // Count delimiters per line and check consistency. Delimiters inside a
+        // double-quoted field are field content, not column separators (RFC 4180) —
+        // a naive count inflates rows with quoted delimiters (e.g. `"Smith, John"`),
+        // which used to make consistency collapse and reject a valid CSV.
+        const counts = lines.map((line) => ContentHeuristicStrategy.countDelimitersOutsideQuotes(line, delimiter));
         const firstCount = counts[0];
         const consistentLines = counts.filter((c) => c === firstCount).length;
         return consistentLines / lines.length >= 0.8;
     }
+    /**
+     * Count occurrences of `delimiter` in `line` that fall OUTSIDE double-quoted
+     * fields, honoring RFC-4180 escaped quotes (`""`). Used by CSV detection so a
+     * delimiter embedded in a quoted value is not mistaken for a column break.
+     */
+    static countDelimitersOutsideQuotes(line, delimiter) {
+        let count = 0;
+        let inQuotes = false;
+        for (let i = 0; i < line.length; i++) {
+            const ch = line[i];
+            if (ch === '"') {
+                if (inQuotes && line[i + 1] === '"') {
+                    i++; // escaped quote inside a quoted field — skip the pair
+                    continue;
+                }
+                inQuotes = !inQuotes;
+            }
+            else if (ch === delimiter && !inQuotes) {
+                count++;
+            }
+        }
+        return count;
+    }
     looksLikeJSON(text) {
         // hasJsonMarkers now does full validation including JSON.parse
         return hasJsonMarkers(text);

package/dist/utils/csvProcessor.js

@@ -337,6 +337,20 @@ function calculateDataQualityScore(columns, warnings, totalRows) {
 /**
  * Analyze all columns in parsed CSV data
  */
+/**
+ * Confidence for a CSV processing result, weighted by data quality.
+ *
+ * The result used to report a static `confidence: 100` regardless of how clean
+ * the data was. This ties confidence to `dataQualityScore` so it actually
+ * varies with quality factors (ragged rows, high null rate, type
+ * inconsistency): a cleanly-parsed CSV stays at 100, a messy one is pulled
+ * halfway toward its quality score. It never drops below 50 for a CSV that
+ * parsed — the *format* is certain, only the *data* is imperfect.
+ */
+function csvResultConfidence(dataQualityScore) {
+    const clamped = Math.max(0, Math.min(100, dataQualityScore));
+    return Math.round((100 + clamped) / 2);
+}
 function analyzeColumns(rows) {
     if (rows.length === 0) {
         return {
@@ -425,6 +439,33 @@ function detectHasHeaders(headerValues, dataRows) {
  * - Lines with significantly different delimiter count than line 2
  * - Lines that don't match CSV structure of subsequent lines
  */
+/** Strip a leading UTF-8 BOM (U+FEFF) if present. */
+function stripBom(text) {
+    return text.charCodeAt(0) === 0xfeff ? text.slice(1) : text;
+}
+/**
+ * Count commas that fall OUTSIDE double-quoted fields (RFC 4180, escaped-quote
+ * aware). A quoted comma is field content, not a column separator, so counting
+ * naively would misclassify a quoted-comma header line.
+ */
+function countUnquotedCommas(line) {
+    let count = 0;
+    let inQuotes = false;
+    for (let i = 0; i < line.length; i++) {
+        const ch = line[i];
+        if (ch === '"') {
+            if (inQuotes && line[i + 1] === '"') {
+                i++;
+                continue;
+            }
+            inQuotes = !inQuotes;
+        }
+        else if (ch === "," && !inQuotes) {
+            count++;
+        }
+    }
+    return count;
+}
 function isMetadataLine(lines) {
     if (!lines[0] || lines.length < 2) {
         return false;
@@ -434,8 +475,8 @@ function isMetadataLine(lines) {
     if (firstLine.match(/^sep=/i)) {
         return true;
     }
-    const firstCommaCount = (firstLine.match(/,/g) || []).length;
-    const secondCommaCount = (secondLine.match(/,/g) || []).length;
+    const firstCommaCount = countUnquotedCommas(firstLine);
+    const secondCommaCount = countUnquotedCommas(secondLine);
     if (firstCommaCount === 0 && secondCommaCount > 0) {
         return true;
     }
@@ -574,7 +615,7 @@ export class CSVProcessor {
                 content: limitedCSV,
                 mimeType: "text/csv",
                 metadata: {
-                    confidence: 100,
+                    confidence: csvResultConfidence(dataQualityScore),
                     size: content.length,
                     rowCount,
                     totalLines: limitedLines.length,
@@ -648,7 +689,7 @@ export class CSVProcessor {
             content: formatted,
             mimeType: "text/csv",
             metadata: {
-                confidence: 100,
+                confidence: csvResultConfidence(dataQualityScore),
                 size: content.length,
                 rowCount,
                 columnCount,
@@ -676,6 +717,9 @@ export class CSVProcessor {
      * @returns Array of row objects
      */
     static async parseCSVFile(filePath, maxRows = 1000) {
+        if (typeof filePath !== "string" || filePath.trim().length === 0) {
+            throw new Error("CSVProcessor.parseCSVFile: filePath must be a non-empty string");
+        }
         const clampedMaxRows = Math.max(1, Math.min(10000, maxRows));
         const fs = await import("fs");
         logger.debug("[CSVProcessor] Starting file parsing", {
@@ -701,6 +745,8 @@ export class CSVProcessor {
                 }
             });
             lineReader.on("end", () => resolve());
+            // Metadata sniffing is best-effort — a read error here must not crash.
+            lineReader.on("error", () => resolve());
         });
         await fileHandle.close();
         const hasMetadataLine = isMetadataLine(firstLines);
@@ -718,6 +764,12 @@ export class CSVProcessor {
                 source.destroy();
                 parser.destroy();
             };
+            // .pipe() does not forward source errors (disk read/permission failures)
+            // to the parser, so listen on the source directly or it throws unhandled.
+            source.on("error", (error) => {
+                abort();
+                reject(new Error(`[CSVProcessor] CSV file read failed after ${count} row(s) (${filePath}): ${error.message}`));
+            });
             source
                 .pipe(parser)
                 .on("data", (row) => {
@@ -739,7 +791,7 @@ export class CSVProcessor {
             })
                 .on("error", (error) => {
                 logger.error("[CSVProcessor] File parsing failed:", error);
-                reject(error);
+                reject(new Error(`[CSVProcessor] CSV parsing failed after ${count} row(s): ${error.message}`));
             });
         });
     }
@@ -752,13 +804,19 @@ export class CSVProcessor {
      * @returns Array of row objects
      */
     static async parseCSVString(csvString, maxRows = 1000) {
+        if (typeof csvString !== "string" || csvString.trim().length === 0) {
+            throw new Error("CSVProcessor.parseCSVString: csvString must be a non-empty string");
+        }
+        // Strip a leading UTF-8 BOM (common in Excel exports) so it does not glue
+        // onto the first column name and break downstream key lookups.
+        const normalized = stripBom(csvString);
         const clampedMaxRows = Math.max(1, Math.min(10000, maxRows));
         logger.debug("[CSVProcessor] Starting string parsing", {
-            inputLength: csvString.length,
+            inputLength: normalized.length,
             maxRows: clampedMaxRows,
         });
         // Detect and skip metadata line
-        const lines = splitCsvLines(csvString);
+        const lines = splitCsvLines(normalized);
         const hasMetadataLine = isMetadataLine(lines);
         const csvData = hasMetadataLine
             ? lines.slice(1).join("\n")
@@ -775,6 +833,13 @@ export class CSVProcessor {
                 source.destroy();
                 parser.destroy();
             };
+            // .pipe() does not forward source-stream errors to the destination, so a
+            // source failure must be listened for directly or it throws as an
+            // unhandled EventEmitter error and can crash the process.
+            source.on("error", (error) => {
+                abort();
+                reject(new Error(`[CSVProcessor] CSV source stream failed after ${count} row(s): ${error.message}`));
+            });
             source
                 .pipe(parser)
                 .on("data", (row) => {
@@ -792,7 +857,7 @@ export class CSVProcessor {
             })
                 .on("error", (error) => {
                 logger.error("[CSVProcessor] Parsing failed:", error);
-                reject(error);
+                reject(new Error(`[CSVProcessor] CSV parsing failed after ${count} row(s): ${error.message}`));
             });
         });
     }

package/dist/utils/fileDetector.js

@@ -2166,13 +2166,38 @@ class ContentHeuristicStrategy {
             const hasUniformLengths = stdDev / avgLength < 0.75;
             return hasReasonableLengths && noBinaryChars && hasUniformLengths;
         }
-        // Count delimiters per line and check consistency
-        const delimRegex = delimiter === "|" ? /\|/g : new RegExp(delimiter, "g");
-        const counts = lines.map((line) => (line.match(delimRegex) || []).length);
+        // Count delimiters per line and check consistency. Delimiters inside a
+        // double-quoted field are field content, not column separators (RFC 4180) —
+        // a naive count inflates rows with quoted delimiters (e.g. `"Smith, John"`),
+        // which used to make consistency collapse and reject a valid CSV.
+        const counts = lines.map((line) => ContentHeuristicStrategy.countDelimitersOutsideQuotes(line, delimiter));
         const firstCount = counts[0];
         const consistentLines = counts.filter((c) => c === firstCount).length;
         return consistentLines / lines.length >= 0.8;
     }
+    /**
+     * Count occurrences of `delimiter` in `line` that fall OUTSIDE double-quoted
+     * fields, honoring RFC-4180 escaped quotes (`""`). Used by CSV detection so a
+     * delimiter embedded in a quoted value is not mistaken for a column break.
+     */
+    static countDelimitersOutsideQuotes(line, delimiter) {
+        let count = 0;
+        let inQuotes = false;
+        for (let i = 0; i < line.length; i++) {
+            const ch = line[i];
+            if (ch === '"') {
+                if (inQuotes && line[i + 1] === '"') {
+                    i++; // escaped quote inside a quoted field — skip the pair
+                    continue;
+                }
+                inQuotes = !inQuotes;
+            }
+            else if (ch === delimiter && !inQuotes) {
+                count++;
+            }
+        }
+        return count;
+    }
     looksLikeJSON(text) {
         // hasJsonMarkers now does full validation including JSON.parse
         return hasJsonMarkers(text);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@juspay/neurolink",
-  "version": "9.88.6",
+  "version": "9.88.8",
   "packageManager": "pnpm@10.15.1",
   "description": "TypeScript AI SDK with 24+ LLM providers behind one consistent API. MCP-native (58+ servers), voice TTS/STT/realtime, RAG, agents, memory, context compaction. OpenAI · Anthropic · Gemini · Bedrock · Azure · Ollama · DeepSeek · NVIDIA NIM and more.",
   "author": {