enlarge the default token length of RAPTOR summarization (#3454)

### What problem does this PR solve?

#3426

### Type of change

- [x] New Feature (non-breaking change which adds functionality)

api/apps/document_app.py

@@ -13,6 +13,7 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License
 #
+import json
 import os.path
 import pathlib
 import re
@@ -533,7 +534,7 @@ def parse():
                 data=False, message='The URL format is invalid', code=settings.RetCode.ARGUMENT_ERROR)
         download_path = os.path.join(get_project_base_directory(), "logs/downloads")
         os.makedirs(download_path, exist_ok=True)
-        from selenium.webdriver import Chrome, ChromeOptions
+        from seleniumwire.webdriver import Chrome, ChromeOptions
         options = ChromeOptions()
         options.add_argument('--headless')
         options.add_argument('--disable-gpu')
@@ -547,11 +548,32 @@ def parse():
         })
         driver = Chrome(options=options)
         driver.get(url)
-        print(driver.get_downloadable_files())
+        res_headers = [r.response.headers for r in driver.requests]
+        if len(res_headers) > 1:
             sections = RAGFlowHtmlParser().parser_txt(driver.page_source)
-        driver.close()
+            driver.quit()
             return get_json_result(data="\n".join(sections))
 
+        class File:
+            filename: str
+            filepath: str
+
+            def __init__(self, filename, filepath):
+                self.filename = filename
+                self.filepath = filepath
+
+            def read(self):
+                with open(self.filepath, "r") as f:
+                    return f.read()
+
+        r = re.search(r"filename=\"([^\"])\"", json.dumps(res_headers))
+        if not r or r.group(1):
+            return get_json_result(
+                data=False, message="Can't not identify downloaded file", code=RetCode.ARGUMENT_ERROR)
+        f = File(r.group(1), os.path.join(download_path, r.group(1)))
+        txt = FileService.parse_docs([f], current_user.id)
+        return get_json_result(data=txt)
+
     if 'file' not in request.files:
         return get_json_result(
             data=False, message='No file part!', code=settings.RetCode.ARGUMENT_ERROR)

graphrag/search.py

@@ -68,7 +68,7 @@ class KGSearch(Dealer):
 
         ent_res = self.dataStore.search(src, list(), condition, [matchText, matchDense, fusionExpr], OrderByExpr(), 0, 32, idxnm, kb_ids)
         ent_res_fields = self.dataStore.getFields(ent_res, src)
-        entities = [d["name_kwd"] for d in ent_res_fields.values()]
+        entities = [d.get["name_kwd"] for d in ent_res_fields.values() if d.get("name_kwd")]
         ent_ids = self.dataStore.getChunkIds(ent_res)
         ent_content = merge_into_first(ent_res_fields, "-Entities-")
         if ent_content:

pyproject.toml

@@ -87,6 +87,7 @@ ruamel-base = "1.0.0"
 scholarly = "1.7.11"
 scikit-learn = "1.5.0"
 selenium = "4.22.0"
+selenium-wire = "5.1.0"
 setuptools = "^75.2.0"
 shapely = "2.0.5"
 six = "1.16.0"

rag/raptor.py

@@ -26,7 +26,7 @@ from rag.utils import truncate
 
 
 class RecursiveAbstractiveProcessing4TreeOrganizedRetrieval:
-    def __init__(self, max_cluster, llm_model, embd_model, prompt, max_token=256, threshold=0.1):
+    def __init__(self, max_cluster, llm_model, embd_model, prompt, max_token=512, threshold=0.1):
         self._max_cluster = max_cluster
         self._llm_model = llm_model
         self._embd_model = embd_model