feat: upgrade to async extractor, add RAG processing, link healing and Docker support

2026-05-07 18:43:43 -03:00
parent 68dc35abbd
commit 570292d8a9
116 changed files with 16277 additions and 388 deletions
@@ -0,0 +1,90 @@
+import os
+import glob
+import re
+
+def heal_links():
+    docs_dir = "fluig_rag_docs"
+    # 1. Mapear todos os arquivos reais existentes
+    all_files = glob.glob(os.path.join(docs_dir, "**", "*"), recursive=True)
+    existing_paths = {os.path.abspath(f): f for f in all_files if os.path.isfile(f)}
+    
+    md_files = [f for f in all_files if f.endswith(".md")]
+    
+    # Regex que suporta parênteses aninhados (até 1 nível)
+    link_pattern = re.compile(r"\[(.*?)\]\(((?:[^()]+|\([^()]*\))*)\)")
+    
+    healed_count = 0
+    print(f"Iniciando cura profunda de links em {len(md_files)} arquivos...")
+    
+    for file_path in md_files:
+        with open(file_path, "r", encoding="utf-8") as f:
+            content = f.read()
+            
+        current_dir = os.path.dirname(file_path)
+        
+        def replace_link(match):
+            nonlocal healed_count
+            text = match.group(1)
+            link = match.group(2)
+            
+            # Pular links externos
+            if link.startswith("http") or ":" in link and not link.startswith("."):
+                return match.group(0)
+            
+            # Correção 1: Remover ".md/" no meio do caminho (bug de substituição parcial)
+            # Ex: Path.md/Subpath.md -> Path/Subpath.md
+            if ".md/" in link:
+                link = link.replace(".md/", "/")
+            
+            # Separar âncora
+            parts = link.split("#")
+            path_part = parts[0]
+            anchor = "#" + parts[1] if len(parts) > 1 else ""
+            
+            if not path_part: return f"[{text}]({link})"
+            
+            target_path = os.path.abspath(os.path.join(current_dir, path_part))
+            
+            # Se o link já existe, apenas devolve (com correção 1 aplicada se houve)
+            if os.path.exists(target_path):
+                if link != match.group(2):
+                    healed_count += 1
+                return f"[{text}]({link})"
+            
+            # Correção 2: Tentar adicionar extensões ou fechar parênteses
+            candidates = [
+                path_part + ".md",
+                path_part + ").md",
+                path_part + ")",
+                path_part.rstrip("(") + ").md"
+            ]
+            
+            for cand in candidates:
+                cand_path = os.path.abspath(os.path.join(current_dir, cand))
+                if os.path.exists(cand_path):
+                    healed_count += 1
+                    new_rel = os.path.relpath(cand_path, current_dir).replace("\\", "/")
+                    return f"[{text}]({new_rel}{anchor})"
+            
+            # Correção 3: Busca aproximada
+            base_name = os.path.basename(path_part).lower().strip()
+            if base_name:
+                for abs_p, rel_p in existing_paths.items():
+                    target_base = os.path.basename(rel_p).lower()
+                    if base_name in target_base:
+                        healed_count += 1
+                        new_rel = os.path.relpath(abs_p, current_dir).replace("\\", "/")
+                        return f"[{text}]({new_rel}{anchor})"
+
+            return match.group(0)
+
+        new_content = link_pattern.sub(replace_link, content)
+        
+        if new_content != content:
+            with open(file_path, "w", encoding="utf-8") as f:
+                f.write(new_content)
+
+    print(f"\nCura concluída! {healed_count} links foram analisados/corrigidos.")
+
+if __name__ == "__main__":
+    heal_links()