langchain/agents/mrkl/prompt.py

# flake8: noqa
PREFIX = """Answer the following questions as best you can. You have access to the following tools:"""
FORMAT_INSTRUCTIONS = """
1.Use the following format to structure each question:

Question: the input question you must answer
Thought: you should always think about what to do
Plan: you should always plan the list of things you need to do to answer the question
Action: the action to take, should be one of [{tool_names}]
Action Input: the input to the action
Observation: the result of the action
... (this Thought/Plan/Action/Action Input/Observation can repeat N times)
Thought: I now know the final answer
Final Answer: the final answer to the original input question
Please follow these guidelines:

1. Ensure only fill in Thought, Plan, Action, Action Input. Do not fill in Observation.
2. Ensure separate each section of  Thought, Plan, Action, Action Input with a new line
3. Ensure start each section with the exactly keyword "Thought: ", "Plan: ",  "Action: ", "Action Input: ".
4. Ensure to start sql query with "SELECT" immediately after "Action Input: ". Do not add any further characters like
5. Ensure to wait for Observation after Action Input.
6. Ensure not to fill in Observation yourself, it will be filled in automatically.
7. When the thought is "I now know the final answer", start a new line with exact key word "Final Answer:" and provide the answer. Do not add any further "Thought" or "Action" sections.

"""
SUFFIX = """Begin!

Question: {input}
Thought:{agent_scratchpad}"""