v1.0.148: PDF parsing via PyPDF2 on VM (replaces broken PDFBox on Lucee)

deploy/convert_server.py

@@ -57,6 +57,13 @@ class Handler(BaseHTTPRequestHandler):
         self.send_header("Access-Control-Allow-Headers", "Content-Type")
         self.end_headers()
 
+    def do_GET(self):
+        parsed = urlparse(self.path)
+        if parsed.path == "/process-v2":
+            self._handle_process_v2(parsed)
+        else:
+            self.send_error(404)
+
     def do_POST(self):
         if self.path == "/llm-ops":
             self._handle_llm_ops()
@@ -64,6 +71,8 @@ class Handler(BaseHTTPRequestHandler):
             self._handle_unzip_upload()
         elif self.path == "/upload":
             self._handle_upload()
+        elif self.path == "/parse-pdf":
+            self._handle_parse_pdf()
         else:
             self._handle_doc_convert()
 
@@ -345,6 +354,43 @@ class Handler(BaseHTTPRequestHandler):
         else:
             self._send_error_cors(502, f"assemble failed: HTTP {resp.status_code}")
 
+    def _handle_parse_pdf(self):
+        """Распарсить PDF через PyPDF2 (получает base64 тело от Lucee)."""
+        import base64
+        from PyPDF2 import PdfReader
+        import io as io_module
+
+        length = int(self.headers.get("Content-Length", 0))
+        body = self.rfile.read(length).decode("utf-8", errors="ignore").strip()
+        raw = base64.b64decode(body)
+
+        try:
+            reader = PdfReader(io_module.BytesIO(raw))
+            elements = []
+            for page in reader.pages:
+                text = page.extract_text()
+                if text:
+                    for line in text.split('\n'):
+                        line = line.strip()
+                        if line:
+                            elements.append({"type": "paragraph", "text": line, "style": ""})
+            result = {
+                "ok": True,
+                "status": "parsed",
+                "element_count": len(elements),
+                "elements": elements,
+                "tables": 0,
+                "paragraphs": len(elements),
+            }
+        except Exception as e:
+            result = {"ok": False, "status": "error", "error": str(e), "element_count": 0}
+
+        self.send_response(200)
+        self.send_header("Content-Type", "application/json; charset=utf-8")
+        self.send_header("Access-Control-Allow-Origin", "*")
+        self.end_headers()
+        self.wfile.write(json.dumps(result, ensure_ascii=False).encode())
+
     def _handle_unzip_upload(self):
         """Распаковать ZIP и загрузить каждый файл в Lucee. Только файлы из корня архива."""
         import zipfile, io

deploy/nginx-contracts.conf

@@ -43,6 +43,11 @@ server {
         add_header Access-Control-Allow-Methods "GET, OPTIONS" always;
     }
 
+    location /parse-pdf {
+        proxy_pass http://127.0.0.1:8766;
+        proxy_read_timeout 60s;
+    }
+
     location /upload {
         if ($request_method = OPTIONS) {
             add_header Access-Control-Allow-Origin "*";

parser.cfm

@@ -10,42 +10,21 @@
     <cfargument name="fileBytes" type="binary">
     <cfset var out = {elements: [], errors: []}>
     <cftry>
-        <cfset var pdfDoc = 0>
-        <cftry>
-            <cfset pdfDoc = createObject("java", "org.apache.pdfbox.Loader").loadPDF(fileBytes)>
-            <cfcatch>
-                <cftry>
-                    <cfset pdfDoc = createObject("java", "org.apache.pdfbox.pdmodel.PDDocument").load(fileBytes)>
-                    <cfcatch>
-                        <cfset arrayAppend(out.errors, "PDFBox: cannot open — " & cfcatch.message)>
-        </cftry>
-        <cfif NOT isNumeric(pdfDoc)>
-            <cfset var stripper = createObject("java", "org.apache.pdfbox.text.PDFTextStripper")>
-            <cfset var pageCount = pdfDoc.getNumberOfPages()>
-            <cfloop from="1" to="#pageCount#" index="pn">
-                <cfset stripper.setStartPage(pn)>
-                <cfset stripper.setEndPage(pn)>
-                <cfset var pageText = stripper.getText(pdfDoc)>
-                <cfif len(trim(pageText))>
-                    <cfloop list="#pageText#" index="line" delimiters="#chr(10)#">
-                        <cfset line = trim(line)>
-                        <cfif len(line)>
-                            <cfset var el = structNew()>
-                            <cfset el.type = "paragraph">
-                            <cfset el.style = "">
-                            <cfset el.text = line>
-                            <cfset el.page = pn>
-                            <cfset arrayAppend(out.elements, el)>
-                        </cfif>
-                    </cfloop>
-                </cfif>
-            </cfloop>
-            <cfset pdfDoc.close()>
+        <cfhttp url="https://contracts.kube5s.ru/parse-pdf" method="post" timeout="30">
+            <cfhttpparam type="body" value="#binaryEncode(fileBytes, 'base64')#">
+        </cfhttp>
+        <cfif cfhttp.statusCode EQ 200>
+            <cfset var data = deserializeJSON(cfhttp.fileContent)>
+            <cfif data.ok>
+                <cfset out.elements = data.elements>
+            <cfelse>
+                <cfset arrayAppend(out.errors, "VM parse error: " & (structKeyExists(data, 'error') ? data.error : 'unknown'))>
+            </cfif>
+        <cfelse>
+            <cfset arrayAppend(out.errors, "VM HTTP error: " & cfhttp.statusCode)>
         </cfif>
-        <cfset bis.close()>
-        <cfset bais.close()>
         <cfcatch>
-            <cfset arrayAppend(out.errors, "PDFBox: " & cfcatch.message)>
+            <cfset arrayAppend(out.errors, "PDF parse failed: " & cfcatch.message)>
         </cfcatch>
     </cftry>
     <cfreturn out>