mirror of
https://github.com/danny-avila/LibreChat.git
synced 2025-12-18 17:30:16 +01:00
* 🔖 refactor: citation handling with support for both literal and Unicode formats * refactor: file search messages for edge cases in documents * 🔧 refactor: Enhance citation handling with detailed regex patterns for literal and Unicode formats * 🔧 refactor: Simplify file search query handling by removing unnecessary parameters and improving result formatting * ✨ test: Add comprehensive integration tests for citation processing flow with support for literal and Unicode formats * 🔧 refactor: Improve regex match handling and add performance tests for citation processing
262 lines
8.7 KiB
TypeScript
262 lines
8.7 KiB
TypeScript
import { visit } from 'unist-util-visit';
|
|
import type { Node } from 'unist';
|
|
import type { Citation, CitationNode } from './types';
|
|
import { SPAN_REGEX, STANDALONE_PATTERN, CLEANUP_REGEX, COMPOSITE_REGEX } from '~/utils/citations';
|
|
|
|
/**
|
|
* Checks if a standalone marker is truly standalone (not inside a composite block).
|
|
* A marker is inside a composite if there's an opening \ue200 without a closing \ue201 after it.
|
|
*
|
|
* Handles both literal text format ("\ue200") and actual Unicode (U+E200) by checking
|
|
* for both and using the rightmost occurrence. This correctly handles:
|
|
* - Pure literal format: "\ue200...\ue201"
|
|
* - Pure Unicode format: "..."
|
|
* - Mixed formats: "\ue200..." (different formats for open/close)
|
|
*/
|
|
function isStandaloneMarker(text: string, position: number): boolean {
|
|
const beforeText = text.substring(0, position);
|
|
|
|
// Find rightmost composite block start (either format)
|
|
const lastUe200Literal = beforeText.lastIndexOf('\\ue200');
|
|
const lastUe200Char = beforeText.lastIndexOf('\ue200');
|
|
const lastUe200 = Math.max(lastUe200Literal, lastUe200Char);
|
|
|
|
// Find rightmost composite block end (either format)
|
|
const lastUe201Literal = beforeText.lastIndexOf('\\ue201');
|
|
const lastUe201Char = beforeText.lastIndexOf('\ue201');
|
|
const lastUe201 = Math.max(lastUe201Literal, lastUe201Char);
|
|
|
|
// Standalone if: no opening marker OR closing marker appears after opening
|
|
return lastUe200 === -1 || (lastUe201 !== -1 && lastUe201 > lastUe200);
|
|
}
|
|
|
|
/**
|
|
* Find the next pattern match from the current position
|
|
*/
|
|
function findNextMatch(
|
|
text: string,
|
|
position: number,
|
|
): { type: string; match: RegExpExecArray | null; index: number } | null {
|
|
// Reset regex lastIndex to start from current position
|
|
SPAN_REGEX.lastIndex = position;
|
|
COMPOSITE_REGEX.lastIndex = position;
|
|
STANDALONE_PATTERN.lastIndex = position;
|
|
|
|
// Find next occurrence of each pattern
|
|
const spanMatch = SPAN_REGEX.exec(text);
|
|
const compositeMatch = COMPOSITE_REGEX.exec(text);
|
|
|
|
// For standalone, we need to check each match
|
|
let standaloneMatch: RegExpExecArray | null = null;
|
|
STANDALONE_PATTERN.lastIndex = position;
|
|
|
|
// Find the first standalone match that's not inside a composite block
|
|
let match: RegExpExecArray | null;
|
|
while (!standaloneMatch && (match = STANDALONE_PATTERN.exec(text)) !== null) {
|
|
if (isStandaloneMarker(text, match.index)) {
|
|
standaloneMatch = match;
|
|
}
|
|
}
|
|
|
|
// Find closest match
|
|
let nextMatch: RegExpExecArray | null = null;
|
|
let matchType = '';
|
|
let matchIndex = -1;
|
|
let typeIndex = -1;
|
|
|
|
if (spanMatch && (!nextMatch || spanMatch.index < matchIndex || matchIndex === -1)) {
|
|
nextMatch = spanMatch;
|
|
matchType = 'span';
|
|
matchIndex = spanMatch.index;
|
|
// We can use a counter for typeIndex if needed
|
|
typeIndex = 0;
|
|
}
|
|
|
|
if (compositeMatch && (!nextMatch || compositeMatch.index < matchIndex || matchIndex === -1)) {
|
|
nextMatch = compositeMatch;
|
|
matchType = 'composite';
|
|
matchIndex = compositeMatch.index;
|
|
typeIndex = 0;
|
|
}
|
|
|
|
if (standaloneMatch && (!nextMatch || standaloneMatch.index < matchIndex || matchIndex === -1)) {
|
|
nextMatch = standaloneMatch;
|
|
matchType = 'standalone';
|
|
matchIndex = standaloneMatch.index;
|
|
typeIndex = 0;
|
|
}
|
|
|
|
if (!nextMatch) return null;
|
|
|
|
return { type: matchType, match: nextMatch, index: typeIndex };
|
|
}
|
|
|
|
function processTree(tree: Node) {
|
|
visit(tree, 'text', (node, index, parent) => {
|
|
const textNode = node as CitationNode;
|
|
const parentNode = parent as CitationNode;
|
|
|
|
if (typeof textNode.value !== 'string') return;
|
|
|
|
const originalValue = textNode.value;
|
|
const segments: Array<CitationNode> = [];
|
|
|
|
// Single-pass processing through the string
|
|
let currentPosition = 0;
|
|
|
|
// Important change: Create a map to track citation IDs by their position
|
|
// This ensures consistent IDs across multiple segments
|
|
const citationIds = new Map<number, string>();
|
|
const typeCounts = { span: 0, composite: 0, standalone: 0 };
|
|
|
|
while (currentPosition < originalValue.length) {
|
|
const nextMatchInfo = findNextMatch(originalValue, currentPosition);
|
|
|
|
if (!nextMatchInfo) {
|
|
// No more matches, add remaining content with cleanup
|
|
const remainingText = originalValue.substring(currentPosition).replace(CLEANUP_REGEX, '');
|
|
if (remainingText) {
|
|
segments.push({ type: 'text', value: remainingText });
|
|
}
|
|
break;
|
|
}
|
|
|
|
const { type, match } = nextMatchInfo;
|
|
const matchIndex = match!.index;
|
|
const matchText = match![0];
|
|
|
|
// Add cleaned text before this match
|
|
if (matchIndex > currentPosition) {
|
|
const textBeforeMatch = originalValue
|
|
.substring(currentPosition, matchIndex)
|
|
.replace(CLEANUP_REGEX, '');
|
|
|
|
if (textBeforeMatch) {
|
|
segments.push({ type: 'text', value: textBeforeMatch });
|
|
}
|
|
}
|
|
|
|
// Generate a unique ID for this citation based on its position in the text
|
|
const citationId = `${type}-${typeCounts[type as keyof typeof typeCounts]}-${matchIndex}`;
|
|
citationIds.set(matchIndex, citationId);
|
|
|
|
// Process based on match type
|
|
switch (type) {
|
|
case 'span': {
|
|
const spanText = matchText;
|
|
const cleanText = spanText.replace(/\\ue203|\\ue204/g, '');
|
|
|
|
// Look ahead for associated citation
|
|
let associatedCitationId: string | null = null;
|
|
const endOfSpan = matchIndex + matchText.length;
|
|
|
|
// Check if there's a citation right after this span
|
|
const nextCitation = findNextMatch(originalValue, endOfSpan);
|
|
if (
|
|
nextCitation &&
|
|
(nextCitation.type === 'standalone' || nextCitation.type === 'composite') &&
|
|
nextCitation.match!.index - endOfSpan < 5
|
|
) {
|
|
// Use the ID that will be generated for the next citation
|
|
const nextIndex = nextCitation.match!.index;
|
|
const nextType = nextCitation.type;
|
|
associatedCitationId = `${nextType}-${typeCounts[nextType as keyof typeof typeCounts]}-${nextIndex}`;
|
|
}
|
|
|
|
segments.push({
|
|
type: 'highlighted-text',
|
|
data: {
|
|
hName: 'highlighted-text',
|
|
hProperties: { citationId: associatedCitationId },
|
|
},
|
|
children: [{ type: 'text', value: cleanText }],
|
|
});
|
|
|
|
typeCounts.span++;
|
|
break;
|
|
}
|
|
|
|
case 'composite': {
|
|
const compositeText = matchText;
|
|
|
|
// Use a regular expression to extract reference indices
|
|
const compositeRefRegex = new RegExp(STANDALONE_PATTERN.source, 'g');
|
|
let refMatch: RegExpExecArray | null;
|
|
const citations: Array<Citation> = [];
|
|
|
|
while ((refMatch = compositeRefRegex.exec(compositeText)) !== null) {
|
|
const turn = Number(refMatch[1]);
|
|
const refType = refMatch[2];
|
|
const refIndex = Number(refMatch[3]);
|
|
|
|
citations.push({
|
|
turn,
|
|
refType,
|
|
index: refIndex,
|
|
});
|
|
}
|
|
|
|
if (citations.length > 0) {
|
|
segments.push({
|
|
type: 'composite-citation',
|
|
data: {
|
|
hName: 'composite-citation',
|
|
hProperties: {
|
|
citations,
|
|
citationId: citationId,
|
|
},
|
|
},
|
|
});
|
|
}
|
|
|
|
typeCounts.composite++;
|
|
break;
|
|
}
|
|
|
|
case 'standalone': {
|
|
// Extract reference info
|
|
const turn = Number(match![1]);
|
|
const refType = match![2];
|
|
const refIndex = Number(match![3]);
|
|
|
|
segments.push({
|
|
type: 'citation',
|
|
data: {
|
|
hName: 'citation',
|
|
hProperties: {
|
|
citation: {
|
|
turn,
|
|
refType,
|
|
index: refIndex,
|
|
},
|
|
citationType: 'standalone',
|
|
citationId: citationId,
|
|
},
|
|
},
|
|
});
|
|
|
|
typeCounts.standalone++;
|
|
break;
|
|
}
|
|
}
|
|
|
|
// Move position forward
|
|
currentPosition = matchIndex + matchText.length;
|
|
}
|
|
|
|
// Replace the original node with our segments or clean up the original
|
|
if (segments.length > 0 && index !== undefined) {
|
|
parentNode.children?.splice(index, 1, ...segments);
|
|
return index + segments.length;
|
|
} else if (textNode.value !== textNode.value.replace(CLEANUP_REGEX, '')) {
|
|
// If we didn't create segments but there are markers to clean up
|
|
textNode.value = textNode.value.replace(CLEANUP_REGEX, '');
|
|
}
|
|
});
|
|
}
|
|
|
|
export function unicodeCitation() {
|
|
return (tree: Node) => {
|
|
processTree(tree);
|
|
};
|
|
}
|