Fix: split process bug in graphrag extract (#6423)

### What problem does this PR solve?

1. miss completion delimiter.
2. miss bracket process.
3. doc_ids return by update_graph is a set, and insert operation in
extract_community need a list.


### Type of change

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

graphrag/general/graph_extractor.py

@@ -13,7 +13,7 @@ import trio
 
 from graphrag.general.extractor import Extractor, ENTITY_EXTRACTION_MAX_GLEANINGS, DEFAULT_ENTITY_TYPES
 from graphrag.general.graph_prompt import GRAPH_EXTRACTION_PROMPT, CONTINUE_PROMPT, LOOP_PROMPT
-from graphrag.utils import ErrorHandlerFn, perform_variable_replacements, chat_limiter
+from graphrag.utils import ErrorHandlerFn, perform_variable_replacements, chat_limiter, split_string_by_multi_markers
 from rag.llm.chat_model import Base as CompletionLLM
 import networkx as nx
 from rag.utils import num_tokens_from_string
@@ -121,8 +121,7 @@ class GraphExtractor(Extractor):
 
         # Repeat to ensure we maximize entity count
         for i in range(self._max_gleanings):
-            text = perform_variable_replacements(CONTINUE_PROMPT, history=history, variables=variables)
+            history.append({"role": "user", "content": CONTINUE_PROMPT})
-            history.append({"role": "user", "content": text})
             async with chat_limiter:
                 response = await trio.to_thread.run_sync(lambda: self._chat("", history, gen_conf))
             token_count += num_tokens_from_string("\n".join([m["content"] for m in history]) + response)
@@ -138,11 +137,19 @@ class GraphExtractor(Extractor):
             token_count += num_tokens_from_string("\n".join([m["content"] for m in history]) + response)
             if continuation != "YES":
                 break
-        record_delimiter = variables.get(self._record_delimiter_key, DEFAULT_RECORD_DELIMITER)
+
-        tuple_delimiter = variables.get(self._tuple_delimiter_key, DEFAULT_TUPLE_DELIMITER)
+        records = split_string_by_multi_markers(
-        records = [re.sub(r"^\(|\)$", "", r.strip()) for r in results.split(record_delimiter)]
+            results,
-        records = [r for r in records if r.strip()]
+            [self._prompt_variables[self._record_delimiter_key], self._prompt_variables[self._completion_delimiter_key]],
-        maybe_nodes, maybe_edges = self._entities_and_relations(chunk_key, records, tuple_delimiter)
+        )
+        rcds = []
+        for record in records:
+            record = re.search(r"\((.*)\)", record)
+            if record is None:
+                continue
+            rcds.append(record.group(1))
+        records = rcds
+        maybe_nodes, maybe_edges = self._entities_and_relations(chunk_key, records, self._prompt_variables[self._tuple_delimiter_key])
         out_results.append((maybe_nodes, maybe_edges, token_count))
         if self.callback:
             self.callback(0.5+0.1*len(out_results)/num_chunks, msg = f"Entities extraction of chunk {chunk_seq} {len(out_results)}/{num_chunks} done, {len(maybe_nodes)} nodes, {len(maybe_edges)} edges, {token_count} tokens.")

graphrag/general/index.py

@@ -343,7 +343,7 @@ async def extract_community(
             "entities_kwd": stru["entities"],
             "important_kwd": stru["entities"],
             "kb_id": kb_id,
-            "source_id": doc_ids,
+            "source_id": list(doc_ids),
             "available_int": 0,
         }
         chunk["content_sm_ltks"] = rag_tokenizer.fine_grained_tokenize(