Skip to content

prompt

Hooks to modify the prompts.

Here is a collection of methods to hook the prompts components that instruct the Agent.

agent_prompt_instructions(instructions, cat)

Hook the instruction prompt.

Allows to edit the instructions that the Cat feeds to the Agent.

The instructions are then composed with two other prompt components, i.e. agent_prompt_prefix and agent_prompt_suffix.

Parameters:

Name Type Description Default
cat CheshireCat

Cheshire Cat instance.

 required

Returns:

Name Type Description
instructions str

The string with the set of instructions informing the Agent on how to format its reasoning to select a proper tool for the task at hand.
Notes

This prompt explains the Agent how to format its chain of reasoning when deciding when and which tool to use. Default prompt splits the reasoning in::

- Thought: Yes/No answer to the question "Do I need to use a tool?";

- Action: a tool chosen among the available ones;

- Action Input: input to be passed to the tool. This is inferred as explained in the tool docstring;

- Observation: description of the result (which is the output of the @tool decorated function found in plugins).
Source code in cat/mad_hatter/core_plugin/hooks/prompt.py 
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
@hook(priority=0)
def agent_prompt_instructions(instructions: str, cat) -> str:
    """Hook the instruction prompt.

    Allows to edit the instructions that the Cat feeds to the *Agent*.

    The instructions are then composed with two other prompt components, i.e. `agent_prompt_prefix`
    and `agent_prompt_suffix`.

    Parameters
    ----------
    cat : CheshireCat
        Cheshire Cat instance.

    Returns
    -------
    instructions : str
        The string with the set of instructions informing the *Agent* on how to format its reasoning to select a
        proper tool for the task at hand.

    Notes
    -----
    This prompt explains the *Agent* how to format its chain of reasoning when deciding when and which tool to use.
    Default prompt splits the reasoning in::

        - Thought: Yes/No answer to the question "Do I need to use a tool?";

        - Action: a tool chosen among the available ones;

        - Action Input: input to be passed to the tool. This is inferred as explained in the tool docstring;

        - Observation: description of the result (which is the output of the @tool decorated function found in plugins).

    """

    return instructions

agent_prompt_prefix(prefix, cat)

Hook the main prompt prefix.

Allows to edit the prefix of the Main Prompt that the Cat feeds to the Agent.

The prefix is then composed with two other prompts components, i.e. the agent_prompt_instructions and the agent_prompt_suffix.

Parameters:

Name Type Description Default
cat CheshireCat

Cheshire Cat instance.

 required

Returns:

Name Type Description
prefix str

The prefix string to be composed with the other two components to make up the Main Prompt.
Notes

The default prefix describe who the AI is and how it is expected to answer the Human. The next part of the prompt (generated form the Agent) contains the list of available Tools.

Source code in cat/mad_hatter/core_plugin/hooks/prompt.py 
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
@hook(priority=0)
def agent_prompt_prefix(prefix, cat) -> str:
    """Hook the main prompt prefix.

    Allows to edit the prefix of the *Main Prompt* that the Cat feeds to the *Agent*.

    The prefix is then composed with two other prompts components, i.e. the `agent_prompt_instructions`
    and the `agent_prompt_suffix`.

    Parameters
    ----------
    cat : CheshireCat
        Cheshire Cat instance.

    Returns
    -------
    prefix : str
        The prefix string to be composed with the other two components to make up the *Main Prompt*.

    Notes
    -----
    The default prefix describe who the AI is and how it is expected to answer the Human.
    The next part of the prompt (generated form the *Agent*) contains the list of available Tools.

    """

    return prefix

agent_prompt_suffix(prompt_suffix, cat)

Hook the main prompt suffix.

Allows to edit the suffix of the Main Prompt that the Cat feeds to the Agent.

The suffix is then composed with two other prompts components, i.e. the agent_prompt_prefix and the agent_prompt_instructions.

Parameters:

Name Type Description Default
cat CheshireCat

Cheshire Cat instance.

 required

Returns:

Name Type Description
suffix str

The suffix string to be composed with the other two components that make up the Main Prompt.
Notes

The default suffix has a few placeholders: - {episodic_memory} provides memories retrieved from episodic memory (past conversations) - {declarative_memory} provides memories retrieved from declarative memory (uploaded documents) - {chat_history} provides the Agent the recent conversation history - {input} provides the last user's input - {agent_scratchpad} is where the Agent can concatenate tools use and multiple calls to the LLM.

Source code in cat/mad_hatter/core_plugin/hooks/prompt.py 
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
@hook(priority=0)
def agent_prompt_suffix(prompt_suffix: str, cat) -> str:
    """Hook the main prompt suffix.

    Allows to edit the suffix of the *Main Prompt* that the Cat feeds to the *Agent*.

    The suffix is then composed with two other prompts components, i.e. the `agent_prompt_prefix`
    and the `agent_prompt_instructions`.

    Parameters
    ----------
    cat : CheshireCat
        Cheshire Cat instance.

    Returns
    -------
    suffix : str
        The suffix string to be composed with the other two components that make up the *Main Prompt*.

    Notes
    -----
    The default suffix has a few placeholders:
    - {episodic_memory} provides memories retrieved from *episodic* memory (past conversations)
    - {declarative_memory} provides memories retrieved from *declarative* memory (uploaded documents)
    - {chat_history} provides the *Agent* the recent conversation history
    - {input} provides the last user's input
    - {agent_scratchpad} is where the *Agent* can concatenate tools use and multiple calls to the LLM.

    """

    return prompt_suffix