fix position extraction bug (#93)

* fix position extraction bug

* remove delimiter for naive parser

api/apps/chunk_app.py

@@ -60,18 +60,21 @@ def list():
         for id in sres.ids:
             d = {
                 "chunk_id": id,
-                "content_with_weight": rmSpace(sres.highlight[id]) if question else sres.field[id].get("content_with_weight", ""),
+                "content_with_weight": rmSpace(sres.highlight[id]) if question else sres.field[id].get(
+                    "content_with_weight", ""),
                 "doc_id": sres.field[id]["doc_id"],
                 "docnm_kwd": sres.field[id]["docnm_kwd"],
                 "important_kwd": sres.field[id].get("important_kwd", []),
                 "img_id": sres.field[id].get("img_id", ""),
                 "available_int": sres.field[id].get("available_int", 1),
                 "positions": sres.field[id].get("position_int", "").split("\t")
             }
-            poss = []
-            for i in range(0, len(d["positions"]), 5):
-                poss.append([float(d["positions"][i]), float(d["positions"][i+1]), float(d["positions"][i+2]), float(d["positions"][i+3]), float(d["positions"][i+4])])
-            d["positions"] = poss
+            if len(d["positions"]) % 5 == 0:
+                poss = []
+                for i in range(0, len(d["positions"]), 5):
+                    poss.append([float(d["positions"][i]), float(d["positions"][i + 1]), float(d["positions"][i + 2]),
+                                 float(d["positions"][i + 3]), float(d["positions"][i + 4])])
+                d["positions"] = poss
             res["chunks"].append(d)
         return get_json_result(data=res)
     except Exception as e:
@@ -137,10 +140,10 @@ def set():
             return get_data_error_result(retmsg="Document not found!")
 
         if doc.parser_id == ParserType.QA:
-            arr = [t for t in re.split(r"[\n\t]", req["content_with_weight"]) if len(t)>1]
+            arr = [t for t in re.split(r"[\n\t]", req["content_with_weight"]) if len(t) > 1]
             if len(arr) != 2: return get_data_error_result(retmsg="Q&A must be separated by TAB/ENTER key.")
             q, a = rmPrefix(arr[0]), rmPrefix[arr[1]]
-            d = beAdoc(d, arr[0], arr[1], not any([huqie.is_chinese(t) for t in q+a]))
+            d = beAdoc(d, arr[0], arr[1], not any([huqie.is_chinese(t) for t in q + a]))
 
         v, c = embd_mdl.encode([doc.name, req["content_with_weight"]])
         v = 0.1 * v[0] + 0.9 * v[1] if doc.parser_id != ParserType.QA else v[1]
@@ -189,7 +192,8 @@ def create():
     md5 = hashlib.md5()
     md5.update((req["content_with_weight"] + req["doc_id"]).encode("utf-8"))
     chunck_id = md5.hexdigest()
-    d = {"id": chunck_id, "content_ltks": huqie.qie(req["content_with_weight"]), "content_with_weight": req["content_with_weight"]}
+    d = {"id": chunck_id, "content_ltks": huqie.qie(req["content_with_weight"]),
+         "content_with_weight": req["content_with_weight"]}
     d["content_sm_ltks"] = huqie.qieqie(d["content_ltks"])
     d["important_kwd"] = req.get("important_kwd", [])
     d["important_tks"] = huqie.qie(" ".join(req.get("important_kwd", [])))

api/db/db_models.py

@@ -527,7 +527,7 @@ class Dialog(DataBaseModel):
     tenant_id = CharField(max_length=32, null=False)
     name = CharField(max_length=255, null=True, help_text="dialog application name")
     description = TextField(null=True, help_text="Dialog description")
-    icon = CharField(max_length=16, null=False, help_text="dialog icon")
+    icon = TextField(null=True, help_text="icon base64 string")
     language = CharField(max_length=32, null=True, default="Chinese", help_text="English|Chinese")
     llm_id = CharField(max_length=32, null=False, help_text="default llm ID")
     llm_setting = JSONField(null=False, default={"temperature": 0.1, "top_p": 0.3, "frequency_penalty": 0.7,

deepdoc/parser/pdf_parser.py

@@ -35,6 +35,7 @@ def __init__(self):
             self.updown_cnt_mdl.set_param({"device": "cuda"})
         self.updown_cnt_mdl.load_model(hf_hub_download(repo_id="InfiniFlow/text_concat_xgb_v1.0",
                                                        filename="updown_concat_xgb.model"))
+        self.page_from = 0
         """
         If you have trouble downloading HuggingFace models, -_^ this might help!!
 
@@ -683,7 +684,7 @@ def cropout(bxs, ltype, poss):
                             "layoutno", "")))
 
                 left, top, right, bott = b["x0"], b["top"], b["x1"], b["bottom"]
-                poss.append((pn, left, right, top, bott))
+                poss.append((pn+self.page_from, left, right, top, bott))
                 return self.page_images[pn] \
                     .crop((left * ZM, top * ZM,
                            right * ZM, bott * ZM))
@@ -863,6 +864,7 @@ def __images__(self, fnm, zoomin=3, page_from=0, page_to=299):
         self.garbages = {}
         self.page_cum_height = [0]
         self.page_layout = []
+        self.page_from = page_from
         try:
             self.pdf = pdfplumber.open(fnm) if isinstance(fnm, str) else pdfplumber.open(BytesIO(fnm))
             self.page_images = [p.to_image(resolution=72 * zoomin).annotated for i, p in
@@ -947,7 +949,9 @@ def crop(self, text, ZM=3, need_position=False):
             left, right, top, bottom = float(left), float(
                 right), float(top), float(bottom)
             poss.append(([int(p) - 1 for p in pn.split("-")], left, right, top, bottom))
-        if not poss: return
+        if not poss:
+            if need_position: return None, None
+            return
 
         max_width = np.max([right-left for (_, left, right, _, _) in poss])
         GAP = 6
@@ -969,7 +973,8 @@ def crop(self, text, ZM=3, need_position=False):
                     bottom, self.page_images[pns[0]].size[1])
                                                ))
             )
-            positions.append((pns[0], left, right, top, min(
+            if 0 < ii < len(poss)-1:
+                positions.append((pns[0]+self.page_from, left, right, top, min(
                     bottom, self.page_images[pns[0]].size[1])/ZM))
             bottom -= self.page_images[pns[0]].size[1]
             for pn in pns[1:]:
@@ -980,8 +985,9 @@ def crop(self, text, ZM=3, need_position=False):
                                                    self.page_images[pn].size[1])
                                                ))
                 )
-                positions.append((pn, left, right, 0, min(
-                    bottom, self.page_images[pn].size[1]) / ZM))
+                if 0 < ii < len(poss) - 1:
+                    positions.append((pn+self.page_from, left, right, 0, min(
+                        bottom, self.page_images[pn].size[1]) / ZM))
                 bottom -= self.page_images[pn].size[1]
 
         if not imgs:

docker/entrypoint.sh

@@ -10,7 +10,7 @@ PY=/root/miniconda3/envs/py11/bin/python
 
 function task_exe(){
   sleep 60;
-  while [ 1 -eq 1 ];do mpirun -n 2 --allow-run-as-root $PY rag/svr/task_executor.py ; done
+  while [ 1 -eq 1 ];do mpirun -n 4 --allow-run-as-root $PY rag/svr/task_executor.py ; done
 }
 
 function watch_broker(){

rag/app/book.py

@@ -41,7 +41,7 @@ def __call__(self, filename, binary=None, from_page=0,
         self._filter_forpages()
         self._merge_with_same_bullet()
         callback(0.75, "Text merging finished.")
-        tbls = self._extract_table_figure(True, zoomin, False, True)
+        tbls = self._extract_table_figure(True, zoomin, True, True)
 
         callback(0.8, "Text extraction finished")
 

rag/app/manual.py

@@ -33,7 +33,7 @@ def __call__(self, filename, binary=None, from_page=0,
         self._concat_downward(concat_between_pages=False)
         self._filter_forpages()
         callback(0.77, "Text merging finished")
-        tbls = self._extract_table_figure(True, zoomin, False, True)
+        tbls = self._extract_table_figure(True, zoomin, True, True)
 
         # clean mess
         for b in self.boxes:

rag/app/naive.py

@@ -40,7 +40,7 @@ def __call__(self, filename, binary=None, from_page=0,
         self._concat_downward(concat_between_pages=False)
         self._filter_forpages()
         callback(0.77, "Text merging finished")
-        tbls = self._extract_table_figure(True, zoomin, False, True)
+        tbls = self._extract_table_figure(True, zoomin, True, True)
 
         cron_logger.info("paddle layouts:".format((timer() - start) / (self.total_page + 0.1)))
         #self._naive_vertical_merge()

rag/app/paper.py

@@ -48,7 +48,7 @@ def __call__(self, filename, binary=None, from_page=0,
         self._concat_downward(concat_between_pages=False)
         self._filter_forpages()
         callback(0.75, "Text merging finished.")
-        tbls = self._extract_table_figure(True, zoomin, False, True)
+        tbls = self._extract_table_figure(True, zoomin, True, True)
 
         # clean mess
         if column_width < self.page_images[0].size[0] / zoomin / 2:

rag/nlp/__init__.py

@@ -246,6 +246,8 @@ def add_chunk(t, pos):
             tk_nums[-1] += tnum
 
     for sec, pos in sections:
+        add_chunk(sec, pos)
+        continue
         s, e = 0, 1
         while e < len(sec):
             if sec[e] in delimiter:

rag/nlp/search.py

@@ -83,7 +83,7 @@ def search(self, req, idxnm, emb_mdl=None):
             else:
                 s = s.sort(
                     {"page_num_int": {"order": "asc", "unmapped_type": "float"}},
-                    {"top_int": {"order": "asc", "unmapped_type": "float"}},
+                    {"top_int": {"order": "asc", "unmapped_type": "float", "mode" : "avg"}},
                     {"create_time": {"order": "desc", "unmapped_type": "date"}},
                     {"create_timestamp_flt": {"order": "desc", "unmapped_type": "float"}}
                 )

rag/svr/task_broker.py

@@ -83,10 +83,10 @@ def new_task():
             pages = PdfParser.total_page_number(r["name"], MINIO.get(r["kb_id"], r["location"]))
             for s,e in r["parser_config"].get("pages", [(0,100000)]):
                 e = min(e, pages)
-                for p in range(s, e, 10):
+                for p in range(s, e, 5):
                     task = new_task()
                     task["from_page"] = p
-                    task["to_page"] = min(p + 10, e)
+                    task["to_page"] = min(p + 5, e)
                     tsks.append(task)
         else:
             tsks.append(new_task())