Merge pull request #394 from openchatai/feat/summarizer

Adding a summarization column to be used when summarizing responses

llm-server/custom_types/bot_message.py

@@ -1,17 +1,17 @@
-from typing import List, Optional
+from typing import List
 from langchain.pydantic_v1 import BaseModel, Field
 from langchain.output_parsers import PydanticOutputParser
 
+
 class BotMessage(BaseModel):
     bot_message: str = Field(description="Message from the bot")
     ids: List[str] = Field(description="List of IDs")
-    missing_information: Optional[str] = Field(description="Incase of ambiguity ask user follow up question")
-
-
+
+
 # Set up a parser + inject instructions into the prompt template.
 bot_message_parser = PydanticOutputParser(pydantic_object=BotMessage)
 
 
 # bot_message_parser.parse(input_string)
 def parse_bot_message(input: str) -> BotMessage:
-    return bot_message_parser.parse(input)
+    return bot_message_parser.parse(input)

llm-server/models/migrations/versions/d845330c4432_add_system_summary_prompt_to_chatbot.py

@@ -0,0 +1,41 @@
+"""Add system_summary_prompt to Chatbot
+
+Revision ID: d845330c4432
+Revises: 86c78095b920
+Create Date: 2023-12-12 14:35:53.182454
+
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+
+# revision identifiers, used by Alembic.
+revision: str = "d845330c4432"
+down_revision: Union[str, None] = "86c78095b920"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade():
+    if (
+        not op.get_bind()
+        .execute(sa.text("SHOW COLUMNS FROM chatbots LIKE 'summary_prompt'"))
+        .fetchone()
+    ):
+        op.add_column("chatbots", sa.Column("summary_prompt", sa.Text(), nullable=True))
+
+    op.execute(
+        """
+        UPDATE chatbots
+        SET summary_prompt = "Given a JSON response, summarize the key information in a concise manner. Include relevant details, references, and links if present. Format the summary in Markdown for clarity and readability."
+        """
+    )
+
+    op.alter_column(
+        "chatbots", "summary_prompt", existing_type=sa.TEXT(), nullable=False
+    )
+
+
+def downgrade():
+    op.drop_column("chatbots", "system_summary_prompt")

llm-server/routes/chat/chat_controller.py

@@ -158,10 +158,13 @@ async def send_chat():
             headers=headers_from_json,
             server_base_url=server_base_url,
             app=app_name,
+            summary_prompt=str(bot.summary_prompt),
         )
 
         if response_data["response"]:
-            upsert_analytics_record(chatbot_id=str(bot.id), successful_operations=1, total_operations=1)
+            upsert_analytics_record(
+                chatbot_id=str(bot.id), successful_operations=1, total_operations=1
+            )
             create_chat_history(str(bot.id), session_id, True, message)
             create_chat_history(
                 str(bot.id),
@@ -170,7 +173,12 @@ async def send_chat():
                 response_data["response"] or response_data["error"] or "",
             )
         elif response_data["error"]:
-            upsert_analytics_record(chatbot_id=str(bot.id), successful_operations=0, total_operations=1, logs=response_data["error"])
+            upsert_analytics_record(
+                chatbot_id=str(bot.id),
+                successful_operations=0,
+                total_operations=1,
+                logs=response_data["error"],
+            )
 
         return jsonify(
             {"type": "text", "response": {"text": response_data["response"]}}

llm-server/routes/root_service.py

@@ -47,6 +47,7 @@
 
 chat = get_chat_model()
 
+
 def validate_steps(steps: List[str], swagger_doc: ResolvingParser):
     try:
         paths = swagger_doc.specification.get("paths", {})
@@ -67,15 +68,18 @@ def validate_steps(steps: List[str], swagger_doc: ResolvingParser):
         if all(x in operationIds for x in steps):
             return True
         else:
-            logger.warn("Model has hallucinated, made up operation id", steps=steps, operationIds=operationIds)
+            logger.warn(
+                "Model has hallucinated, made up operation id",
+                steps=steps,
+                operationIds=operationIds,
+            )
             return False
 
     except Exception as e:
         logger.error(f"An error occurred: {str(e)}")
         return False
 
 
-
 async def handle_request(
     text: str,
     swagger_url: str,
@@ -85,6 +89,7 @@ async def handle_request(
     headers: Dict[str, str],
     server_base_url: str,
     app: Optional[str],
+    summary_prompt: str,
 ) -> ResponseDict:
     log_user_request(text)
     check_required_fields(base_prompt, text, swagger_url)
@@ -112,32 +117,26 @@ async def handle_request(
             prev_conversations=prev_conversations,
             flows=flows,
             bot_id=bot_id,
-            base_prompt=base_prompt
+            base_prompt=base_prompt,
         )
-
-        if step.missing_information is not None and len(step.missing_information) >= 10:
-            return {
-                "error": None,
-                "response": step.bot_message + "\n" + step.missing_information
-            }
 
         if len(step.ids) > 0:
             swagger_doc = get_swagger_doc(swagger_url)
             fl = validate_steps(step.ids, swagger_doc)
-            
+
             if fl is False:
                 return {"error": None, "response": step.bot_message}
-            
+
             response = await handle_api_calls(
                 ids=step.ids,
                 swagger_doc=swagger_doc,
                 app=app,
                 bot_id=bot_id,
                 headers=headers,
                 server_base_url=server_base_url,
-                session_id=session_id,
                 text=text,
                 swagger_url=swagger_url,
+                summary_prompt=summary_prompt,
             )
 
             logger.info(
@@ -196,6 +195,7 @@ def get_swagger_doc(swagger_url: str) -> ResolvingParser:
     else:
         return ResolvingParser(spec_string=swagger_doc)
 
+
 async def handle_api_calls(
     ids: List[str],
     swagger_doc: ResolvingParser,
@@ -204,8 +204,8 @@ async def handle_api_calls(
     server_base_url: str,
     swagger_url: Optional[str],
     app: Optional[str],
-    session_id: str,
     bot_id: str,
+    summary_prompt: str,
 ) -> ResponseDict:
     _workflow = create_workflow_from_operation_ids(ids, swagger_doc, text)
     output = await run_workflow(
@@ -214,6 +214,7 @@ async def handle_api_calls(
         WorkflowData(text, headers, server_base_url, swagger_url, app),
         app,
         bot_id=bot_id,
+        summary_prompt=summary_prompt,
     )
 
     _workflow["swagger_url"] = swagger_url

llm-server/routes/workflow/extractors/convert_json_to_text.py

@@ -17,20 +17,10 @@ def convert_json_to_text(
     api_response: Dict[str, Any],
     api_request_data: Dict[str, Any],
     bot_id: str,
+    summary_prompt: str,
 ) -> str:
     chat = get_chat_model()
-
-    api_summarizer_template = None
-    system_message = SystemMessage(
-        content="You are an ai assistant that can summarize api responses"
-    )
-    prompt_templates = load_prompts(bot_id)
-    api_summarizer_template = (
-        prompt_templates.api_summarizer if prompt_templates else None
-    )
-
-    if api_summarizer_template is not None:
-        system_message = SystemMessage(content=api_summarizer_template)
+    system_message = SystemMessage(content=summary_prompt)
 
     messages = [
         system_message,
@@ -41,12 +31,14 @@ def convert_json_to_text(
         HumanMessage(
             content="Here is the response from the apis: {}".format(api_response)
         ),
-        # HumanMessage(
-        #     content="Here is the api_request_data: {}".format(api_request_data)
-        # ),
     ]
 
     result = chat(messages)
-    logger.info("Convert json to text", content=result.content, incident="convert_json_to_text", api_request_data=api_request_data)
+    logger.info(
+        "Convert json to text",
+        content=result.content,
+        incident="convert_json_to_text",
+        api_request_data=api_request_data,
+    )
 
     return cast(str, result.content)

llm-server/routes/workflow/utils/process_conversation_step.py

@@ -28,15 +28,25 @@ def process_conversation_step(
     prev_conversations: List[BaseMessage],
     flows: List[WorkflowFlowType],
     bot_id: str,
-    base_prompt: str
+    base_prompt: str,
 ):
-    logger.info("planner data", context=context, api_summaries=api_summaries, prev_conversations=prev_conversations, flows=flows)
+    logger.info(
+        "planner data",
+        context=context,
+        api_summaries=api_summaries,
+        prev_conversations=prev_conversations,
+        flows=flows,
+    )
     if not session_id:
         raise ValueError("Session id must be defined for chat conversations")
     messages: List[BaseMessage] = []
     messages.append(SystemMessage(content=base_prompt))
-
-    messages.append(SystemMessage(content="You will have access to a list of api's and some useful information, called context."))
+
+    messages.append(
+        SystemMessage(
+            content="You will have access to a list of api's and some useful information, called context."
+        )
+    )
 
     if len(prev_conversations) > 0:
         messages.extend(prev_conversations)
@@ -71,24 +81,22 @@ def process_conversation_step(
 
     messages.append(
         HumanMessage(
-            content="""Based on the information provided to you and the conversation history of this conversation, I want you to answer the questions that follow. Your should respond with a json that looks like the following, you must always use the operationIds provided in api summaries. Do not make up an operation id - 
-    {
-        "ids": ["list", "of", "operationIds", "for apis to be called"],
-        "bot_message": "your response based on the instructions provided at the beginning",
-        "missing_information": "Optional Field; Incase of ambiguity ask clarifying questions. You should not worry about the api filters or query, that should be decided by a different agent."
-    }                
-    
-    Don't add operation ids if you can reply by merely looking in the conversation history.
-    """
+            content="""Based on the information provided to you and the conversation history of this conversation, I want you to answer the questions that follow. You should respond with a json that looks like the following, you must always use the operationIds provided in api summaries. Do not make up an operation id -
+{
+    "ids": ["list", "of", "operationIds", "for", "apis", "to", "be", "called"],
+    "bot_message": "your response based on the instructions provided at the beginning, this could also be clarification if the information provided by the user is not complete / accurate",  
+}
+
+Don't add operation ids if you can reply by merely looking in the conversation history.
+"""
         )
     )
     messages.append(
         HumanMessage(content="If you are unsure / confused, ask claryfying questions")
     )
 
     messages.append(HumanMessage(content=user_requirement))
-
-
+
     logger.info("messages array", messages=messages)
 
     content = cast(str, chat(messages=messages).content)
@@ -105,7 +113,7 @@ def process_conversation_step(
 
     except OutputParserException as e:
         logger.warn("Failed to parse json", data=content, err=str(e))
-        return BotMessage(bot_message=content, ids=[], missing_information=None)
+        return BotMessage(bot_message=content, ids=[])
     except Exception as e:
         logger.warn("unexpected error occured", err=str(e))
-        return BotMessage(ids=[], bot_message=str(e), missing_information=None)
+        return BotMessage(ids=[], bot_message=str(e))