phoenix-oss/llama-stack
1
0
Fork 0
forked from phoenix-oss/llama-stack-mirror
Code Pull requests Releases Packages 2 Activity Actions

feat: update default tool use prompt (#1803)

Browse source 
# What does this PR do?
User reports in
https://github.com/meta-llama/llama-stack/issues/1769#issuecomment-2755564632
that Agent uses tool even on a prompt 'Hello'.

Updated the default prompt. Also move the instruction part out of
`function_description` so that user can override it if desired.

## Test Plan
<img width="1344" alt="image"
src="https://github.com/user-attachments/assets/c606d65d-071f-4211-a719-b4742676acda"
/>

Also performance on 100 hotpotqa questions are similar to the current
prompt.
This commit is contained in:
ehhuang 2025-04-12 11:54:22 -07:00 • committed by GitHub
parent f34f22f8c7
commit 1e5bf6c19d
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 5 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

9
llama_stack/models/llama/llama3/prompt_templates/system_prompts.py
View file

@ -229,6 +229,11 @@ class PythonListCustomToolGenerator(PromptTemplateGeneratorBase): # noqa: N801
You are an expert in composing functions. You are given a question and a set of possible functions.
Based on the question, you may or may not need to make one function/tool call to achieve the purpose.
If you decide to invoke any of the function(s), you MUST put it in the format of [func_name1(params_name1=params_value1, params_name2=params_value2...), func_name2(params)]
If you decide to invoke a function, you SHOULD NOT include any other text in the response. besides the function call in the above format.
For a boolean parameter, be sure to use `True` or `False` (capitalized) for the value.
{{ function_description }}
""".strip("\n")
)
@ -243,10 +248,6 @@ class PythonListCustomToolGenerator(PromptTemplateGeneratorBase): # noqa: N801
def _gen_function_description(self, custom_tools: List[ToolDefinition]) -> PromptTemplate:
template_str = textwrap.dedent(
"""
If you decide to invoke any of the function(s), you MUST put it in the format of [func_name1(params_name1=params_value1, params_name2=params_value2...), func_name2(params)]
For a boolean parameter, be sure to use `True` or `False` (capitalized) for the value.
You SHOULD NOT include any other text in the response.
Here is a list of functions in JSON format that you can invoke.
[