hello guys, i have a problem acccessing

aamamr

hello guys, i have a problem acccessing agents prompts and changing them

Attachment

26 comments

LLogan M

you probably dont want to mess around too much with the react prompt

LLogan M

instead, you can add some context to it

LLogan M

ReActAgent.from_tools(..., context="Some extra context")

LLogan M

if you really want to change the react prompt, it looks like you are on the right track, it will be somewhere in that dict

aamamr

all i wanna do is change the system prompt from it being so vague like that ,give it more personality and instructions

Attachment

aamamr

ill try that context arg thank you

aamamr

the thing for agents , that dict almost like hidden or private , theyre not accessible as u can see

LLogan M

yea if you follow the guide, theres an update_prompts method

i.e.

LLogan M

Plain Text

 query_engine.update_prompts(
    {"response_synthesizer:summary_template": new_summary_tmpl}
)

aamamr

that one doesnt work for agents

aamamr

worked for query engines and not the agents

LLogan M

works fine for me

LLogan M

>>> from llama_index.core.agent import ReActAgent
>>> agent = ReActAgent.from_tools([])

>>> agent.get_prompts()
{'agent_worker:system_prompt': PromptTemplate(metadata={'prompt_type': <PromptType.CUSTOM: 'custom'>}, template_vars=['tool_desc', 'tool_names'], kwargs={}, output_parser=None, template_var_mappings=None, function_mappings=None, template='You are designed to help with a variety of tasks, from answering questions to providing summaries to other types of analyses.\n\n## Tools\n\nYou have access to a wide variety of tools. You are re .... l you have enough information to answer the question without using any more tools. At that point, you MUST respond in the one of the following two formats:\n\n
Plain Text
\nThought: I can answer without using any more tools. I\'ll use the user\'s language to answer\nAnswer: [your answer here (In the same language as the user\'s question)]\n
\n\n
Plain Text
\nThought: I cannot answer the question with the provided tools.\nAnswer: [your answer here (In the same language as the user\'s question)]\n
\n\n## Current Conversation\n\nBelow is the current conversation consisting of interleaving human and assistant messages.\n')}

>>> prompt = agent.get_prompts()['agent_worker:system_prompt']
>>> agent.update_prompts({'agent_worker:system_prompt': prompt})
>>>