Implementation:Run llama Llama index PromptMixin

Knowledge Sources	Run_llama_Llama_index
Domains	Prompts, AbstractBaseClass, Architecture
Last Updated	2026-02-11 19:00 GMT

Overview

PromptMixin is an abstract base class that provides a standardized interface for getting and setting prompts across LlamaIndex modules and their nested sub-modules.

Description

PromptMixin defines a hierarchical prompt management protocol used throughout LlamaIndex by query engines, response synthesizers, postprocessors, and other modules. It provides:

• get_prompts() - Retrieves all prompts from the current module and recursively from all sub-modules. Sub-module prompts are namespaced using a colon separator (e.g., response_synthesizer:text_qa_template). Returns a deep copy to prevent accidental mutation.
• update_prompts(prompts_dict) - Updates prompts at both the current level and in nested sub-modules. Keys containing a colon are routed to the appropriate sub-module by splitting on the first colon.
• _validate_prompts() - Internal validation that ensures prompt keys and module names do not contain the reserved colon character, which is used as the namespace separator.

Subclasses must implement three abstract methods:

• _get_prompts() - Return the module's own prompts as a Dict[str, BasePromptTemplate].
• _get_prompt_modules() - Return a dictionary of sub-modules that also implement PromptMixin, enabling recursive prompt discovery.
• _update_prompts(prompts_dict) - Apply prompt updates to the module's own prompts.

The module also defines three type aliases used across the codebase:

• HasPromptType = Union[PromptMixin, BasePromptTemplate]
• PromptDictType = Dict[str, BasePromptTemplate]
• PromptMixinType = Dict[str, PromptMixin]

Usage

Use PromptMixin as a base class for any LlamaIndex component that manages prompts. It enables users to inspect and customize prompts at any level of a nested pipeline (e.g., changing the text QA prompt inside a response synthesizer that is inside a query engine) through a unified interface.

Code Reference

Source Location

• Repository: Run_llama_Llama_index
• File: llama-index-core/llama_index/core/prompts/mixin.py

Signature

class PromptMixin(ABC):
    def get_prompts(self) -> Dict[str, BasePromptTemplate]:
        ...

    def update_prompts(self, prompts_dict: Dict[str, BasePromptTemplate]) -> None:
        ...

    @abstractmethod
    def _get_prompts(self) -> PromptDictType:
        ...

    @abstractmethod
    def _get_prompt_modules(self) -> PromptMixinType:
        ...

    @abstractmethod
    def _update_prompts(self, prompts_dict: PromptDictType) -> None:
        ...

Import

from llama_index.core.prompts.mixin import PromptMixin, PromptDictType, PromptMixinType

I/O Contract

Inputs (get_prompts)

Name	Type	Required	Description
(none)	-	-	No parameters. Recursively collects prompts from self and all sub-modules.

Inputs (update_prompts)

Name	Type	Required	Description
prompts_dict	Dict[str, BasePromptTemplate]	Yes	Dictionary mapping prompt keys to prompt templates. Keys with colons are routed to sub-modules.

Outputs (get_prompts)

Name	Type	Description
prompts	Dict[str, BasePromptTemplate]	All prompts from the current module and sub-modules, with sub-module prompts namespaced by colon.

Usage Examples

from llama_index.core.prompts.mixin import PromptMixin, PromptDictType, PromptMixinType
from llama_index.core.prompts.base import BasePromptTemplate

# Inspect prompts in a query engine
query_engine = index.as_query_engine()
all_prompts = query_engine.get_prompts()
for key, prompt in all_prompts.items():
    print(f"{key}: {prompt.get_template()}")

# Update a specific prompt
from llama_index.core.prompts import PromptTemplate

new_qa_prompt = PromptTemplate(
    "Context: {context_str}\nQuestion: {query_str}\nAnswer concisely:"
)
query_engine.update_prompts({
    "response_synthesizer:text_qa_template": new_qa_prompt
})

Related Pages

• Environment:Run_llama_Llama_index_Python_LlamaIndex_Core
• Run_llama_Llama_index_BaseNodePostprocessor - Uses PromptMixin for prompt management
• Run_llama_Llama_index_LLMQuestionGenerator - Implements PromptMixin for question generation prompts