fix: skip reference instructions in LLM prompt when include_references=false

lightrag/operate.py

@@ -3186,7 +3186,12 @@ async def kg_query(
     )
 
     # Build system prompt
-    sys_prompt_temp = system_prompt if system_prompt else PROMPTS["rag_response"]
+    if system_prompt:
+        sys_prompt_temp = system_prompt
+    elif not query_param.include_references:
+        sys_prompt_temp = PROMPTS["rag_response_no_ref"]
+    else:
+        sys_prompt_temp = PROMPTS["rag_response"]
     sys_prompt = sys_prompt_temp.format(
         response_type=response_type,
         user_prompt=user_prompt,
@@ -4011,11 +4016,18 @@ async def _build_context_str(
     )
 
     # Get the system prompt template from PROMPTS or global_config
-    sys_prompt_template = global_config.get(
-        "system_prompt_template", PROMPTS["rag_response"]
-    )
+    custom_sys_prompt = global_config.get("system_prompt_template")
+    if custom_sys_prompt:
+        sys_prompt_template = custom_sys_prompt
+    elif not query_param.include_references:
+        sys_prompt_template = PROMPTS["rag_response_no_ref"]
+    else:
+        sys_prompt_template = PROMPTS["rag_response"]
 
-    kg_context_template = PROMPTS["kg_query_context"]
+    if not query_param.include_references:
+        kg_context_template = PROMPTS["kg_query_context_no_ref"]
+    else:
+        kg_context_template = PROMPTS["kg_query_context"]
     user_prompt = query_param.user_prompt if query_param.user_prompt else ""
     response_type = (
         query_param.response_type
@@ -4087,11 +4099,14 @@ async def _build_context_str(
     text_units_str = "\n".join(
         json.dumps(text_unit, ensure_ascii=False) for text_unit in chunks_context
     )
-    reference_list_str = "\n".join(
-        f"[{ref['reference_id']}] {ref['file_path']}"
-        for ref in reference_list
-        if ref["reference_id"]
-    )
+    if query_param.include_references:
+        reference_list_str = "\n".join(
+            f"[{ref['reference_id']}] {ref['file_path']}"
+            for ref in reference_list
+            if ref["reference_id"]
+        )
+    else:
+        reference_list_str = ""
 
     logger.info(
         f"Final context: {len(entities_context)} entities, {len(relations_context)} relations, {len(chunks_context)} chunks"
@@ -4938,9 +4953,12 @@ async def naive_query(
     )
 
     # Use the provided system prompt or default
-    sys_prompt_template = (
-        system_prompt if system_prompt else PROMPTS["naive_rag_response"]
-    )
+    if system_prompt:
+        sys_prompt_template = system_prompt
+    elif not query_param.include_references:
+        sys_prompt_template = PROMPTS["naive_rag_response_no_ref"]
+    else:
+        sys_prompt_template = PROMPTS["naive_rag_response"]
 
     # Create a preliminary system prompt with empty content_data to calculate overhead
     pre_sys_prompt = sys_prompt_template.format(
@@ -5012,13 +5030,17 @@ async def naive_query(
     text_units_str = "\n".join(
         json.dumps(text_unit, ensure_ascii=False) for text_unit in chunks_context
     )
-    reference_list_str = "\n".join(
-        f"[{ref['reference_id']}] {ref['file_path']}"
-        for ref in reference_list
-        if ref["reference_id"]
-    )
+    if query_param.include_references:
+        reference_list_str = "\n".join(
+            f"[{ref['reference_id']}] {ref['file_path']}"
+            for ref in reference_list
+            if ref["reference_id"]
+        )
+        naive_context_template = PROMPTS["naive_query_context"]
+    else:
+        reference_list_str = ""
+        naive_context_template = PROMPTS["naive_query_context_no_ref"]
 
-    naive_context_template = PROMPTS["naive_query_context"]
     context_content = naive_context_template.format(
         text_chunks_str=text_units_str,
         reference_list_str=reference_list_str,

lightrag/prompt.py

@@ -324,6 +324,76 @@
 6. Additional Instructions: {user_prompt}
 
 
+---Context---
+
+{content_data}
+"""
+
+PROMPTS["rag_response_no_ref"] = """---Role---
+
+You are an expert AI assistant specializing in synthesizing information from a provided knowledge base. Your primary function is to answer user queries accurately by ONLY using the information within the provided **Context**.
+
+---Goal---
+
+Generate a comprehensive, well-structured answer to the user query.
+The answer must integrate relevant facts from the Knowledge Graph and Document Chunks found in the **Context**.
+Consider the conversation history if provided to maintain conversational flow and avoid repeating information.
+
+---Instructions---
+
+1. Step-by-Step Instruction:
+  - Carefully determine the user's query intent in the context of the conversation history to fully understand the user's information need.
+  - Scrutinize both `Knowledge Graph Data` and `Document Chunks` in the **Context**. Identify and extract all pieces of information that are directly relevant to answering the user query.
+  - Weave the extracted facts into a coherent and logical response. Your own knowledge must ONLY be used to formulate fluent sentences and connect ideas, NOT to introduce any external information.
+
+2. Content & Grounding:
+  - Strictly adhere to the provided context from the **Context**; DO NOT invent, assume, or infer any information not explicitly stated.
+  - If the answer cannot be found in the **Context**, state that you do not have enough information to answer. Do not attempt to guess.
+
+3. Formatting & Language:
+  - The response MUST be in the same language as the user query.
+  - The response MUST utilize Markdown formatting for enhanced clarity and structure (e.g., headings, bold text, bullet points).
+  - The response should be presented in {response_type}.
+  - Do not include a references or citations section in the response.
+
+4. Additional Instructions: {user_prompt}
+
+
+---Context---
+
+{context_data}
+"""
+
+PROMPTS["naive_rag_response_no_ref"] = """---Role---
+
+You are an expert AI assistant specializing in synthesizing information from a provided knowledge base. Your primary function is to answer user queries accurately by ONLY using the information within the provided **Context**.
+
+---Goal---
+
+Generate a comprehensive, well-structured answer to the user query.
+The answer must integrate relevant facts from the Document Chunks found in the **Context**.
+Consider the conversation history if provided to maintain conversational flow and avoid repeating information.
+
+---Instructions---
+
+1. Step-by-Step Instruction:
+  - Carefully determine the user's query intent in the context of the conversation history to fully understand the user's information need.
+  - Scrutinize `Document Chunks` in the **Context**. Identify and extract all pieces of information that are directly relevant to answering the user query.
+  - Weave the extracted facts into a coherent and logical response. Your own knowledge must ONLY be used to formulate fluent sentences and connect ideas, NOT to introduce any external information.
+
+2. Content & Grounding:
+  - Strictly adhere to the provided context from the **Context**; DO NOT invent, assume, or infer any information not explicitly stated.
+  - If the answer cannot be found in the **Context**, state that you do not have enough information to answer. Do not attempt to guess.
+
+3. Formatting & Language:
+  - The response MUST be in the same language as the user query.
+  - The response MUST utilize Markdown formatting for enhanced clarity and structure (e.g., headings, bold text, bullet points).
+  - The response should be presented in {response_type}.
+  - Do not include a references or citations section in the response.
+
+4. Additional Instructions: {user_prompt}
+
+
 ---Context---
 
 {content_data}
@@ -371,6 +441,36 @@
 
 """
 
+PROMPTS["kg_query_context_no_ref"] = """
+Knowledge Graph Data (Entity):
+
+```json
+{entities_str}
+```
+
+Knowledge Graph Data (Relationship):
+
+```json
+{relations_str}
+```
+
+Document Chunks:
+
+```json
+{text_chunks_str}
+```
+
+"""
+
+PROMPTS["naive_query_context_no_ref"] = """
+Document Chunks:
+
+```json
+{text_chunks_str}
+```
+
+"""
+
 PROMPTS["keywords_extraction"] = """---Role---
 You are an expert keyword extractor, specializing in analyzing user queries for a Retrieval-Augmented Generation (RAG) system. Your purpose is to identify both high-level and low-level keywords in the user's query that will be used for effective document retrieval.
 

tests/test_include_references_prompt.py

@@ -0,0 +1,73 @@
+"""Tests that include_references=False removes reference instructions from LLM prompts."""
+
+import pytest
+
+from lightrag.prompt import PROMPTS
+
+pytestmark = pytest.mark.offline
+
+
+class TestRagResponsePromptVariants:
+    """Verify rag_response and rag_response_no_ref prompt templates."""
+
+    def test_rag_response_contains_reference_instructions(self):
+        prompt = PROMPTS["rag_response"]
+        assert "References Section Format" in prompt
+        assert "reference_id" in prompt
+        assert "Reference Document List" in prompt
+
+    def test_rag_response_no_ref_omits_reference_instructions(self):
+        prompt = PROMPTS["rag_response_no_ref"]
+        assert "References Section Format" not in prompt
+        assert "reference_id" not in prompt
+        assert "Reference Document List" not in prompt
+
+    def test_rag_response_no_ref_has_required_placeholders(self):
+        prompt = PROMPTS["rag_response_no_ref"]
+        assert "{response_type}" in prompt
+        assert "{user_prompt}" in prompt
+        assert "{context_data}" in prompt
+
+
+class TestNaiveRagResponsePromptVariants:
+    """Verify naive_rag_response and naive_rag_response_no_ref prompt templates."""
+
+    def test_naive_rag_response_contains_reference_instructions(self):
+        prompt = PROMPTS["naive_rag_response"]
+        assert "References Section Format" in prompt
+        assert "reference_id" in prompt
+
+    def test_naive_rag_response_no_ref_omits_reference_instructions(self):
+        prompt = PROMPTS["naive_rag_response_no_ref"]
+        assert "References Section Format" not in prompt
+        assert "reference_id" not in prompt
+
+    def test_naive_rag_response_no_ref_has_required_placeholders(self):
+        prompt = PROMPTS["naive_rag_response_no_ref"]
+        assert "{response_type}" in prompt
+        assert "{user_prompt}" in prompt
+        assert "{content_data}" in prompt
+
+
+class TestContextTemplateVariants:
+    """Verify context templates with and without reference sections."""
+
+    def test_kg_context_contains_reference_list(self):
+        template = PROMPTS["kg_query_context"]
+        assert "Reference Document List" in template
+        assert "{reference_list_str}" in template
+
+    def test_kg_context_no_ref_omits_reference_list(self):
+        template = PROMPTS["kg_query_context_no_ref"]
+        assert "Reference Document List" not in template
+        assert "{reference_list_str}" not in template
+
+    def test_naive_context_contains_reference_list(self):
+        template = PROMPTS["naive_query_context"]
+        assert "Reference Document List" in template
+        assert "{reference_list_str}" in template
+
+    def test_naive_context_no_ref_omits_reference_list(self):
+        template = PROMPTS["naive_query_context_no_ref"]
+        assert "Reference Document List" not in template
+        assert "{reference_list_str}" not in template