fix(service): fix exception when no chunks

gnes/service/encoder.py

@@ -38,17 +38,19 @@ def embed_chunks_in_docs(self, docs: Union[List['gnes_pb2.Document'], 'gnes_pb2.
         embeds = None
 
         for d in docs:
+            if not d.chunks:
+                self.logger.warning('document (doc_id=%s) contains no chunks!' % d.doc_id)
+                continue
+
             for c in d.chunks:
-                chunks.append(c)
                 if d.doc_type == gnes_pb2.Document.TEXT:
                     contents.append(c.text)
                 elif getattr(c, c.WhichOneof('content')) == 'blob':
                     contents.append(blob2array(c.blob))
                 else:
-                    raise ServiceError(
-                        'chunk content is in type: %s, dont kow how to handle that' % c.WhichOneof('content'))
-            else:
-                self.logger.warning('document (doc_id=%s) contains no chunks!' % d.doc_id)
+                    self.logger.warning(
+                        'chunk content is in type: %s, dont kow how to handle that, ignored' % c.WhichOneof('content'))
+                chunks.append(c)
 
         if do_encoding:
             embeds = self._model.encode(contents)

gnes/service/indexer.py

@@ -46,11 +46,12 @@ def _handler_chunk_index(self, msg: 'gnes_pb2.Message'):
         for d in msg.request.index.docs:
             if not d.chunks:
                 self.logger.warning('document (doc_id=%s) contains no chunks!' % d.doc_id)
-            else:
-                vecs += [blob2array(c.embedding) for c in d.chunks]
-                doc_ids += [d.doc_id] * len(d.chunks)
-                offsets += [c.offset for c in d.chunks]
-                weights += [c.weight for c in d.chunks]
+                continue
+
+            vecs += [blob2array(c.embedding) for c in d.chunks]
+            doc_ids += [d.doc_id] * len(d.chunks)
+            offsets += [c.offset for c in d.chunks]
+            weights += [c.weight for c in d.chunks]
 
         if vecs:
             self._model.add(list(zip(doc_ids, offsets)), np.concatenate(vecs, 0), weights)