wip: snapshot das alterações não commitadas (2026-10-20)

Origem: D:/Documents/Github/OllaHub
Arquivos omitidos (segredos, bancos de dados, logs, >50 MB): 0

lib/knowledge-base-processor.ts

@@ -81,16 +81,12 @@ export function condenseKnowledgeBase(
   options: CondensationOptions = {}
 ): CondensationResult {
   const condensationStart = Date.now();
-  
+
   chatLog.info(`\n[KnowledgeBase] ========== CONDENSATION START ==========`);
   chatLog.info(`[KnowledgeBase] Query: "${query}"`);
   chatLog.info(`[KnowledgeBase] Entries: ${entries.length}`);
-  chatLog.info(`[KnowledgeBase] Options:`, {
-    maxTokens: options.maxTokens || DEFAULT_OPTIONS.maxTokens,
-    minRelevanceScore: options.minRelevanceScore || DEFAULT_OPTIONS.minRelevanceScore,
-    autoSummarize: options.autoSummarize ?? DEFAULT_OPTIONS.autoSummarize,
-  });
-  
+  chatLog.info(`[KnowledgeBase] Options: maxTokens=${options.maxTokens || DEFAULT_OPTIONS.maxTokens}, minRelevanceScore=${options.minRelevanceScore || DEFAULT_OPTIONS.minRelevanceScore}, autoSummarize=${options.autoSummarize ?? DEFAULT_OPTIONS.autoSummarize}`);
+
   const opts = { ...DEFAULT_OPTIONS, ...options };
 
   if (entries.length === 0) {
@@ -110,7 +106,7 @@ export function condenseKnowledgeBase(
   // Calcular tokens originais
   const originalContent = entries.map(e => e.content).join('\n\n');
   const originalTokens = estimateTokens(originalContent);
-  
+
   chatLog.info(`[KnowledgeBase] Original content: ${originalContent.length} chars`);
   chatLog.info(`[KnowledgeBase] Original tokens (estimated): ${originalTokens}`);
 
@@ -123,14 +119,14 @@ export function condenseKnowledgeBase(
   }> = [];
 
   chatLog.info(`[KnowledgeBase] Processing ${entries.length} entries into chunks...`);
-  
+
   entries.forEach((entry, entryIndex) => {
     const entryStart = Date.now();
     chatLog.info(`\n[KnowledgeBase] --- Processing Entry ${entryIndex + 1}/${entries.length} ---`);
     chatLog.info(`[KnowledgeBase] Title: ${entry.title}`);
     chatLog.info(`[KnowledgeBase] URL: ${entry.sourceUrl}`);
     chatLog.info(`[KnowledgeBase] Content length: ${entry.content.length} chars`);
-    
+
     const processed = processContent(entry.content, query, {
       summarizeThreshold: opts.summarizeThreshold,
     });
@@ -145,7 +141,7 @@ export function condenseKnowledgeBase(
 
     const entryDuration = Date.now() - entryStart;
     chatLog.info(`[KnowledgeBase] Generated ${chunksWithSource.length} chunks in ${entryDuration}ms`);
-    
+
     // Log detalhes dos chunks
     if (chunksWithSource.length > 0) {
       chunksWithSource.slice(0, 3).forEach((chunk, chunkIdx) => {
@@ -164,7 +160,7 @@ export function condenseKnowledgeBase(
 
     allChunks.push(...chunksWithSource);
   });
-  
+
   chatLog.info(`[KnowledgeBase] Total chunks generated: ${allChunks.length}`);
 
   // Calcular tokens disponíveis (reservar 20% para margem)
@@ -175,7 +171,7 @@ export function condenseKnowledgeBase(
   // Selecionar chunks mais relevantes (usando contexto se disponível)
   const selectionStart = Date.now();
   chatLog.info(`[KnowledgeBase] Selecting most relevant chunks...`);
-  
+
   let selectedChunks = scoreAndSelectChunks(
     allChunks,
     query,
@@ -183,10 +179,10 @@ export function condenseKnowledgeBase(
     opts.minRelevanceScore,
     opts.context
   );
-  
+
   const selectionDuration = Date.now() - selectionStart;
   chatLog.info(`[KnowledgeBase] Selected ${selectedChunks.length} chunks from ${allChunks.length} total in ${selectionDuration}ms`);
-  
+
   // Log top chunks selecionados
   if (selectedChunks.length > 0) {
     chatLog.info(`[KnowledgeBase] Top 5 selected chunks:`);
@@ -208,7 +204,7 @@ export function condenseKnowledgeBase(
   } else {
     // Construir contexto a partir dos chunks selecionados
     const chunksBySource = new Map<number, Chunk[]>();
-    
+
     selectedChunks.forEach(chunk => {
       if (!chunksBySource.has(chunk.sourceIndex)) {
         chunksBySource.set(chunk.sourceIndex, []);
@@ -264,14 +260,14 @@ export function condenseKnowledgeBase(
   chatLog.info(`[KnowledgeBase] Compression ratio: ${(compressionRatio * 100).toFixed(1)}%`);
   chatLog.info(`[KnowledgeBase] Chunks used: ${selectedChunks.length} / ${allChunks.length}`);
   chatLog.info(`[KnowledgeBase] Final context length: ${context.length} chars`);
-  
+
   // Log estatísticas por fonte
   chatLog.info(`[KnowledgeBase] Chunks per source:`);
   entries.forEach((entry, idx) => {
     const chunksUsed = selectedChunks.filter(c => c.sourceIndex === idx).length;
     chatLog.info(`  ${entry.title}: ${chunksUsed} chunks`);
   });
-  
+
   chatLog.info(`[KnowledgeBase] ============================================\n`);
 
   return {
@@ -327,7 +323,7 @@ export function validateCondensedContext(
   const normalizedContext = result.context.toLowerCase();
   const queryWords = normalizedQuery.split(/\s+/).filter(w => w.length > 3);
   const matchingWords = queryWords.filter(word => normalizedContext.includes(word));
-  
+
   if (matchingWords.length < queryWords.length * 0.5) {
     warnings.push('Poucas palavras-chave da query aparecem no contexto');
   }