update

cyyeh · cyyeh · commit 277480c514ef · 2025-08-28T14:04:05.000+08:00
diff --git a/wren-ai-service/src/pipelines/generation/data_assistance.py b/wren-ai-service/src/pipelines/generation/data_assistance.py
@@ -20,7 +20,7 @@
 data_assistance_system_prompt = """
 ### TASK ###
 You are a data analyst great at answering user's questions about given database schema.
-Please carefully read user's question and database schema to answer it in easy to understand manner
+Please carefully read user's question, intent for the question, and database schema to answer it in easy to understand manner
 using the Markdown format. Your goal is to help guide user understand its database!
 
 ### INSTRUCTIONS ###
@@ -50,6 +50,7 @@
 
 ### INPUT ###
 User's question: {{query}}
+Intent for user's question: {{intent_reasoning}}
 Language: {{language}}
 
 Custom Instruction: {{ custom_instruction }}
@@ -62,6 +63,7 @@
 @observe(capture_input=False)
 def prompt(
     query: str,
+    intent_reasoning: str,
     db_schemas: list[str],
     language: str,
     histories: list[AskHistory],
@@ -70,6 +72,7 @@ def prompt(
 ) -> dict:
     _prompt = prompt_builder.run(
         query=query,
+        intent_reasoning=intent_reasoning,
         histories=histories,
         db_schemas=db_schemas,
         language=language,
@@ -153,6 +156,7 @@ async def _get_streaming_results(query_id):
     async def run(
         self,
         query: str,
+        intent_reasoning: str,
         db_schemas: list[str],
         language: str,
         query_id: Optional[str] = None,
@@ -164,6 +168,7 @@ async def run(
             ["data_assistance"],
             inputs={
                 "query": query,
+                "intent_reasoning": intent_reasoning,
                 "db_schemas": db_schemas,
                 "language": language,
                 "query_id": query_id or "",
diff --git a/wren-ai-service/src/pipelines/generation/data_exploration_assistance.py b/wren-ai-service/src/pipelines/generation/data_exploration_assistance.py
@@ -19,8 +19,9 @@
 
 data_exploration_assistance_system_prompt = """
 You are a great data analyst good at exploring data.
-You are given a user question and a sql data.
-You need to understand the user question and the sql data, and then answer the user question.
+You are given a user question, an intent for the question, and a sql data.
+You need to understand the user question, the intent for the question, and the sql data, and then answer the user question.
+
 ### INSTRUCTIONS ###
 1. Your answer should be in the same language as the language user provided.
 2. You must follow the sql data to answer the user question.
@@ -30,6 +31,7 @@
 - provide insights and trends in the data
 - find out anomalies and outliers in the data
 5. You only need to use the skills required to answer the user question based on the user question and the sql data.
+
 ### OUTPUT FORMAT ###
 Please provide your response in proper Markdown format without ```markdown``` tags.
 """
@@ -44,6 +46,7 @@
 
 ### INPUT ###
 User Question: {{query}}
+Intent for user's question: {{intent_reasoning}}
 Language: {{language}}
 SQL Data:
 {{ sql_data }}
@@ -58,6 +61,7 @@
 @observe(capture_input=False)
 def prompt(
     query: str,
+    intent_reasoning: str,
     histories: list[AskHistory],
     language: str,
     sql_data: dict,
@@ -66,6 +70,7 @@ def prompt(
 ) -> dict:
     _prompt = prompt_builder.run(
         query=query,
+        intent_reasoning=intent_reasoning,
         language=language,
         sql_data=sql_data,
         histories=histories,
@@ -144,6 +149,7 @@ async def _get_streaming_results(query_id):
     async def run(
         self,
         query: str,
+        intent_reasoning: str,
         sql_data: dict,
         language: str,
         query_id: Optional[str] = None,
@@ -155,6 +161,7 @@ async def run(
             ["data_exploration_assistance"],
             inputs={
                 "query": query,
+                "intent_reasoning": intent_reasoning,
                 "language": language,
                 "query_id": query_id or "",
                 "sql_data": sql_data,
diff --git a/wren-ai-service/src/pipelines/generation/misleading_assistance.py b/wren-ai-service/src/pipelines/generation/misleading_assistance.py
@@ -19,8 +19,8 @@
 
 misleading_assistance_system_prompt = """
 ### TASK ###
-You are a helpful assistant that can help users understand their data better. Currently, you are given a user's question that is potentially misleading.
-Your goal is to help guide user understand its data better and suggest few better questions to ask.
+You are a helpful assistant that can help users understand their data better. Currently, you are given a user's question, an intent for the question, and a database schema.
+Your goal is to help guide user understand its data better and suggest few better questions to ask based on the intent for the question and the database schema.
 
 ### INSTRUCTIONS ###
 
@@ -50,6 +50,7 @@
 
 ### INPUT ###
 User's question: {{query}}
+Intent for user's question: {{intent_reasoning}}
 Language: {{language}}
 
 Custom Instruction: {{ custom_instruction }}
@@ -62,6 +63,7 @@
 @observe(capture_input=False)
 def prompt(
     query: str,
+    intent_reasoning: str,
     db_schemas: list[str],
     language: str,
     histories: list[AskHistory],
@@ -70,6 +72,7 @@ def prompt(
 ) -> dict:
     _prompt = prompt_builder.run(
         query=query,
+        intent_reasoning=intent_reasoning,
         histories=histories,
         db_schemas=db_schemas,
         language=language,
@@ -153,6 +156,7 @@ async def _get_streaming_results(query_id):
     async def run(
         self,
         query: str,
+        intent_reasoning: str,
         db_schemas: list[str],
         language: str,
         query_id: Optional[str] = None,
@@ -164,6 +168,7 @@ async def run(
             ["misleading_assistance"],
             inputs={
                 "query": query,
+                "intent_reasoning": intent_reasoning,
                 "db_schemas": db_schemas,
                 "language": language,
                 "query_id": query_id or "",
diff --git a/wren-ai-service/src/pipelines/generation/user_clarification_assistance.py b/wren-ai-service/src/pipelines/generation/user_clarification_assistance.py
@@ -18,16 +18,48 @@
 
 
 user_clarification_assistance_system_prompt = """
+You are a helpful assistant that can help users understand their data better. Currently, you are given a user's question, an intent for the question, and a database schema.
+You should tell user why the question is not clear enough or vague and suggest a better question based on the intent for the question and the database schema.
+
+### INSTRUCTIONS ###
+1. Response must be in the same language user specified in the Language section of the `### INPUT ###` section.
+2. There should be proper line breaks, whitespace, and Markdown formatting(headers, lists, tables, etc.) in your response.
+3. MUST NOT add SQL code in your response.
+4. MUST consider database schema when suggesting better questions.
+5. The maximum response length is 100 words.
+6. If the user provides a custom instruction, it should be followed strictly and you should use it to change the style of response.
+
+### OUTPUT FORMAT ###
+Please provide your response in proper Markdown format without ```markdown``` tags.
 """
 
 user_clarification_assistance_user_prompt_template = """
+{% if histories %}
+### PREVIOUS QUESTIONS ###
+{% for history in histories %}
+    {{ history.question }}
+{% endfor %}
+{% endif %}
+
+### DATABASE SCHEMA ###
+{% for db_schema in db_schemas %}
+    {{ db_schema }}
+{% endfor %}
+
+### INPUT ###
+User Question: {{query}}
+Intent for user's question: {{intent_reasoning}}
+Language: {{language}}
+
+Custom Instruction: {{ custom_instruction }}
 """
 
 
 ## Start of Pipeline
 @observe(capture_input=False)
 def prompt(
     query: str,
+    intent_reasoning: str,
     db_schemas: list[str],
     language: str,
     histories: list[AskHistory],
@@ -36,6 +68,7 @@ def prompt(
 ) -> dict:
     _prompt = prompt_builder.run(
         query=query,
+        intent_reasoning=intent_reasoning,
         histories=histories,
         db_schemas=db_schemas,
         language=language,
@@ -119,6 +152,7 @@ async def _get_streaming_results(query_id):
     async def run(
         self,
         query: str,
+        intent_reasoning: str,
         db_schemas: list[str],
         language: str,
         query_id: Optional[str] = None,
@@ -130,6 +164,7 @@ async def run(
             ["user_clarification_assistance"],
             inputs={
                 "query": query,
+                "intent_reasoning": intent_reasoning,
                 "db_schemas": db_schemas,
                 "language": language,
                 "query_id": query_id or "",
diff --git a/wren-ai-service/src/pipelines/generation/user_guide_assistance.py b/wren-ai-service/src/pipelines/generation/user_guide_assistance.py
@@ -19,8 +19,8 @@
 
 user_guide_assistance_system_prompt = """
 You are a helpful assistant that can help users understand Wren AI. 
-You are given a user question and a user guide.
-You need to understand the user question and the user guide, and then answer the user question.
+You are given a user question, an intent for the question, and a user guide.
+You need to understand the user question, the intent reasoning, and the user guide, and then answer the user question.
 
 ### INSTRUCTIONS ###
 1. Your answer should be in the same language as the language user provided.
@@ -44,6 +44,7 @@
 
 ### INPUT ###
 User Question: {{query}}
+Intent for user's question: {{intent_reasoning}}
 Language: {{language}}
 User Guide:
 {% for doc in docs %}
@@ -60,6 +61,7 @@
 @observe(capture_input=False)
 def prompt(
     query: str,
+    intent_reasoning: str,
     language: str,
     wren_ai_docs: list[dict],
     histories: list[AskHistory],
@@ -68,6 +70,7 @@ def prompt(
 ) -> dict:
     _prompt = prompt_builder.run(
         query=query,
+        intent_reasoning=intent_reasoning,
         histories=histories,
         language=language,
         docs=wren_ai_docs,
@@ -153,6 +156,7 @@ async def _get_streaming_results(query_id):
     async def run(
         self,
         query: str,
+        intent_reasoning: str,
         language: str,
         query_id: Optional[str] = None,
         histories: Optional[list[AskHistory]] = None,
@@ -163,6 +167,7 @@ async def run(
             ["user_guide_assistance"],
             inputs={
                 "query": query,
+                "intent_reasoning": intent_reasoning,
                 "language": language,
                 "query_id": query_id or "",
                 "histories": histories or [],
diff --git a/wren-ai-service/src/web/v1/services/ask.py b/wren-ai-service/src/web/v1/services/ask.py
@@ -290,6 +290,7 @@ async def ask(
                             asyncio.create_task(
                                 self._pipelines["misleading_assistance"].run(
                                     query=user_query,
+                                    intent_reasoning=intent_reasoning,
                                     histories=histories,
                                     db_schemas=table_ddls,
                                     language=ask_request.configurations.language,
@@ -313,6 +314,7 @@ async def ask(
                             asyncio.create_task(
                                 self._pipelines["data_assistance"].run(
                                     query=user_query,
+                                    intent_reasoning=intent_reasoning,
                                     histories=histories,
                                     db_schemas=table_ddls,
                                     language=ask_request.configurations.language,
@@ -336,6 +338,7 @@ async def ask(
                             asyncio.create_task(
                                 self._pipelines["user_guide_assistance"].run(
                                     query=user_query,
+                                    intent_reasoning=intent_reasoning,
                                     histories=histories,
                                     language=ask_request.configurations.language,
                                     query_id=ask_request.query_id,
@@ -358,6 +361,7 @@ async def ask(
                             asyncio.create_task(
                                 self._pipelines["data_exploration_assistance"].run(
                                     query=user_query,
+                                    intent_reasoning=intent_reasoning,
                                     histories=histories,
                                     sql_data=last_sql_data,
                                     language=ask_request.configurations.language,
@@ -381,6 +385,7 @@ async def ask(
                             asyncio.create_task(
                                 self._pipelines["user_clarification_assistance"].run(
                                     query=user_query,
+                                    intent_reasoning=intent_reasoning,
                                     histories=histories,
                                     db_schemas=table_ddls,
                                     language=ask_request.configurations.language,
