RAGFlow is an agent-native database: a Retrieval-Augmented Generation (RAG) engine that grew into context infrastructure.
Go agent-harnessagent-native-databaseagentic-aiagentic-retrievalagentic-searchaiai-agentscontext-enginecontext-engineeringcontext-managementharness-engineeringknowledge-compilationragretrieval-augmented-generationretrieval-harnesssearch-harness
Fix: Adjust ExecutionStrategy of THINKING_MODES. (#17862)
Kevin Hud2183e610d10
diff --git a/rag/advanced_rag/harness/config.py b/rag/advanced_rag/harness/config.pyindex b618495248..c7af23a0b4 100644--- a/rag/advanced_rag/harness/config.py+++ b/rag/advanced_rag/harness/config.py@@ -15,7 +15,7 @@ THINKING_MODES: dict[str, ExecutionStrategy] = {max_orchestrator_cycles=1,max_agent_cycles=0,max_parallel_agents=1,- available_tools=["hybrid_search"],+ available_tools=["hybrid_search", "web_search", "bm25_search"],sufficiency_threshold=0.85,partial_threshold=0.50,fallback_to_direct_llm=False,@@ -32,7 +32,7 @@ THINKING_MODES: dict[str, ExecutionStrategy] = {max_orchestrator_cycles=3,max_agent_cycles=0,max_parallel_agents=1,- available_tools=["hybrid_search"],+ available_tools=["hybrid_search", "web_search", "bm25_search"],sufficiency_threshold=0.75,partial_threshold=0.40,fallback_to_direct_llm=False,@@ -41,7 +41,7 @@ THINKING_MODES: dict[str, ExecutionStrategy] = {label="high",execution_strategy="agentic_research",requires_decomposition=True,- requires_agent_loop=True,+ requires_agent_loop=False,requires_sufficiency_judge=True,requires_selective_gen=True,allows_dynamic_claims=False,@@ -52,6 +52,7 @@ THINKING_MODES: dict[str, ExecutionStrategy] = {available_tools=["hybrid_search","web_search",+ "bm25_search","ontology_navigate","dataset_navigation_by_tree","graph_explore",diff --git a/rag/advanced_rag/knowlege_compile/wiki_incremental.py b/rag/advanced_rag/knowlege_compile/wiki_incremental.pyindex 1ee89ef68e..b9f5f727a9 100644--- a/rag/advanced_rag/knowlege_compile/wiki_incremental.py+++ b/rag/advanced_rag/knowlege_compile/wiki_incremental.py@@ -1669,12 +1669,14 @@ async def _wiki_refine_page(# Embed for searchfrom common.misc_utils import thread_pool_exec+ from rag.nlp import rag_tokenizerembeddings, _ = await thread_pool_exec(embd_mdl.encode, [summary or content[:200]])# Derive vector dimension from the embedding shapeemb_arr = np.asarray(embeddings[0])vec_dim = int(emb_arr.shape[0]) if emb_arr.ndim >= 1 and emb_arr.shape[0] else 768+ content_ltks = rag_tokenizer.tokenize(content)page = {"id": _stable_row_id(WIKI_PAGE_COMPILE_KWD, kb_id, page_id),@@ -1691,6 +1693,9 @@ async def _wiki_refine_page("page_type_kwd": page_type_kwd,"compile_kwd": WIKI_PAGE_COMPILE_KWD,"knowledge_graph_kwd": WIKI_PAGE_COMPILE_KWD,+ "title_tks": rag_tokenizer.tokenize(page_title),+ "content_ltks": content_ltks,+ "content_sm_ltks": rag_tokenizer.fine_grained_tokenize(content_ltks),}# Insert vector (adds q_{dim}_vec field)vec_col = f"q_{vec_dim}_vec"diff --git a/rag/llm/chat_model.py b/rag/llm/chat_model.pyindex d34d679d1a..5393b449d0 100644--- a/rag/llm/chat_model.py+++ b/rag/llm/chat_model.py@@ -1999,7 +1999,7 @@ class LiteLLMBase(ABC):content = json.dumps(result, ensure_ascii=False)else:content = str(result)- hist.append({"role": "tool", "tool_call_id": tc.id, "content": content.replace("</think>", "")})+ hist.append({"role": "tool", "tool_call_id": tc.id, "content": content.replace("</think>", "") + ("</think>" if content.find("<think>") >= 0 else "")})return histdef bind_tools(self, toolcall_session=None, tools=None):