Implementation:Run llama Llama index BaseImageRetriever

Overview

BaseImageRetriever is an abstract base class that defines the interface for image retrieval in LlamaIndex. It supports two retrieval modes -- text-to-image and image-to-image -- each with both synchronous and asynchronous variants. Subclasses must implement four abstract methods to provide the actual retrieval logic.

Source file: llama-index-core/llama_index/core/image_retriever.py (112 lines)

Class Hierarchy

PromptMixin, DispatcherSpanMixin
  └── BaseImageRetriever

BaseImageRetriever inherits from:

• PromptMixin -- provides prompt management capabilities
• DispatcherSpanMixin -- provides instrumentation and span tracking

Retrieval Modes

The class supports two distinct retrieval paradigms:

Mode	Description	Input
Text-to-Image	Retrieve images based on a text query	Text string or QueryBundle
Image-to-Image	Retrieve images similar to a given image	Image path string or QueryBundle

Public Methods

text_to_image_retrieve

def text_to_image_retrieve(
    self, str_or_query_bundle: QueryType
) -> List[NodeWithScore]:

Retrieves image nodes given a text query. Accepts either a plain string or a QueryBundle object. If a string is provided, it is automatically wrapped in a QueryBundle with the string set as query_str. Delegates to the abstract _text_to_image_retrieve method.

image_to_image_retrieve

def image_to_image_retrieve(
    self, str_or_query_bundle: QueryType
) -> List[NodeWithScore]:

Retrieves image nodes given an input image. Accepts either a file path string or a QueryBundle. If a string is provided, wraps it in a QueryBundle with query_str set to empty and image_path set to the provided string. Delegates to the abstract _image_to_image_retrieve method.

atext_to_image_retrieve

async def atext_to_image_retrieve(
    self, str_or_query_bundle: QueryType
) -> List[NodeWithScore]:

Async version of text_to_image_retrieve. Performs the same input normalization and delegates to _atext_to_image_retrieve.

aimage_to_image_retrieve

async def aimage_to_image_retrieve(
    self, str_or_query_bundle: QueryType
) -> List[NodeWithScore]:

Async version of image_to_image_retrieve. Performs the same input normalization and delegates to _aimage_to_image_retrieve.

Abstract Methods (Must Be Implemented by Subclasses)

Method	Signature	Description
_text_to_image_retrieve	(query_bundle: QueryBundle) -> List[NodeWithScore]	Sync text-to-image retrieval logic
_image_to_image_retrieve	(query_bundle: QueryBundle) -> List[NodeWithScore]	Sync image-to-image retrieval logic
_atext_to_image_retrieve	async (query_bundle: QueryBundle) -> List[NodeWithScore]	Async text-to-image retrieval logic
_aimage_to_image_retrieve	async (query_bundle: QueryBundle) -> List[NodeWithScore]	Async image-to-image retrieval logic

All four methods receive a normalized QueryBundle (never a raw string) and must return a list of NodeWithScore objects.

Input Handling Pattern

Both the sync and async public methods follow the same pattern:

1. Accept QueryType (which is Union[str, QueryBundle]).
2. If the input is a string, wrap it in an appropriate QueryBundle.
3. Delegate to the corresponding abstract method.

For text-to-image, the string is used as query_str. For image-to-image, the string is used as image_path with an empty query_str.

Dependencies

• llama_index.core.indices.query.schema -- provides QueryBundle and QueryType
• llama_index.core.schema -- provides NodeWithScore
• llama_index.core.instrumentation -- provides DispatcherSpanMixin
• llama_index.core.prompts.mixin -- provides PromptMixin