Fix: Tika server upgrades. (#12073)

### What problem does this PR solve?

#12037

### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)

rag/app/book.py

@@ -91,7 +91,7 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
             filename, binary=binary, from_page=from_page, to_page=to_page)
         remove_contents_table(sections, eng=is_english(
             random_choices([t for t, _ in sections], k=200)))
-        tbls=vision_figure_parser_docx_wrapper(sections=sections,tbls=tbls,callback=callback,**kwargs)
+        tbls = vision_figure_parser_docx_wrapper(sections=sections,tbls=tbls,callback=callback,**kwargs)
         # tbls = [((None, lns), None) for lns in tbls]
         sections=[(item[0],item[1] if item[1] is not None else "") for item in sections if not isinstance(item[1], Image.Image)]
         callback(0.8, "Finish parsing.")
@@ -147,9 +147,16 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
 
     elif re.search(r"\.doc$", filename, re.IGNORECASE):
         callback(0.1, "Start to parse.")
-        with BytesIO(binary) as binary:
-            binary = BytesIO(binary)
-            doc_parsed = parser.from_buffer(binary)
+        try:
+            from tika import parser as tika_parser
+        except Exception as e:
+            callback(0.8, f"tika not available: {e}. Unsupported .doc parsing.")
+            logging.warning(f"tika not available: {e}. Unsupported .doc parsing for {filename}.")
+            return []
+
+        binary = BytesIO(binary)
+        doc_parsed = tika_parser.from_buffer(binary)
+        if doc_parsed.get('content', None) is not None:
             sections = doc_parsed['content'].split('\n')
             sections = [(line, "") for line in sections if line]
             remove_contents_table(sections, eng=is_english(

rag/app/manual.py

@@ -312,7 +312,7 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
             tk_cnt = num_tokens_from_string(txt)
             if sec_id > -1:
                 last_sid = sec_id
-        tbls=vision_figure_parser_pdf_wrapper(tbls=tbls,callback=callback,**kwargs)
+        tbls = vision_figure_parser_pdf_wrapper(tbls=tbls,callback=callback,**kwargs)
         res = tokenize_table(tbls, doc, eng)
         res.extend(tokenize_chunks(chunks, doc, eng, pdf_parser))
         table_ctx = max(0, int(parser_config.get("table_context_size", 0) or 0))
@@ -325,7 +325,7 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
         docx_parser = Docx()
         ti_list, tbls = docx_parser(filename, binary,
                                     from_page=0, to_page=10000, callback=callback)
-        tbls=vision_figure_parser_docx_wrapper(sections=ti_list,tbls=tbls,callback=callback,**kwargs)
+        tbls = vision_figure_parser_docx_wrapper(sections=ti_list,tbls=tbls,callback=callback,**kwargs)
         res = tokenize_table(tbls, doc, eng)
         for text, image in ti_list:
             d = copy.deepcopy(doc)

rag/app/one.py

@@ -76,7 +76,7 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
     if re.search(r"\.docx$", filename, re.IGNORECASE):
         callback(0.1, "Start to parse.")
         sections, tbls = naive.Docx()(filename, binary)
-        tbls=vision_figure_parser_docx_wrapper(sections=sections,tbls=tbls,callback=callback,**kwargs)
+        tbls = vision_figure_parser_docx_wrapper(sections=sections, tbls=tbls, callback=callback, **kwargs)
         sections = [s for s, _ in sections if s]
         for (_, html), _ in tbls:
             sections.append(html)
@@ -142,10 +142,18 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
 
     elif re.search(r"\.doc$", filename, re.IGNORECASE):
         callback(0.1, "Start to parse.")
+        try:
+            from tika import parser as tika_parser
+        except Exception as e:
+            callback(0.8, f"tika not available: {e}. Unsupported .doc parsing.")
+            logging.warning(f"tika not available: {e}. Unsupported .doc parsing for {filename}.")
+            return []
+
         binary = BytesIO(binary)
-        doc_parsed = parser.from_buffer(binary)
-        sections = doc_parsed['content'].split('\n')
-        sections = [s for s in sections if s]
+        doc_parsed = tika_parser.from_buffer(binary)
+        if doc_parsed.get('content', None) is not None:
+            sections = doc_parsed['content'].split('\n')
+            sections = [s for s in sections if s]
         callback(0.8, "Finish parsing.")
 
     else:

rag/flow/parser/parser.py

@@ -650,7 +650,7 @@ class Parser(ProcessBase):
             tmpf.flush()
             tmp_path = os.path.abspath(tmpf.name)
 
-            seq2txt_mdl = LLMBundle(self._canvas.get_tenant_id(), LLMType.SPEECH2TEXT)
+            seq2txt_mdl = LLMBundle(self._canvas.get_tenant_id(), LLMType.SPEECH2TEXT, llm_name=conf["llm_id"])
             txt = seq2txt_mdl.transcription(tmp_path)
 
             self.set_output("text", txt)

rag/svr/task_executor.py

@@ -374,7 +374,7 @@ async def build_chunks(task, progress_callback):
         chat_mdl = LLMBundle(task["tenant_id"], LLMType.CHAT, llm_name=task["llm_id"], lang=task["language"])
 
         async def gen_metadata_task(chat_mdl, d):
-            cached = get_llm_cache(chat_mdl.llm_name, d["content_with_weight"], "metadata")
+            cached = get_llm_cache(chat_mdl.llm_name, d["content_with_weight"], "metadata", {})
             if not cached:
                 async with chat_limiter:
                     cached = await gen_metadata(chat_mdl,
@@ -852,7 +852,7 @@ async def do_handle_task(task):
     task_tenant_id = task["tenant_id"]
     task_embedding_id = task["embd_id"]
     task_language = task["language"]
-    task_llm_id = task["llm_id"]
+    task_llm_id = task["parser_config"].get("llm_id") or task["llm_id"]
     task_dataset_id = task["kb_id"]
     task_doc_id = task["doc_id"]
     task_document_name = task["name"]