|
from typing import Any, Dict, List |
|
|
|
from haystack import Document |
|
|
|
from fastrag.embedders import ( |
|
IPEXSentenceTransformersDocumentEmbedder, |
|
IPEXSentenceTransformersTextEmbedder, |
|
) |
|
|
|
|
|
class EndpointHandler: |
|
def __init__(self, path=""): |
|
model_id = "Intel/bge-small-en-v1.5-rag-int8-static" |
|
|
|
self.query_embedder = IPEXSentenceTransformersTextEmbedder(model=model_id) |
|
self.document_embedder = IPEXSentenceTransformersDocumentEmbedder(model=model_id) |
|
|
|
self.query_embedder.warm_up() |
|
self.document_embedder.warm_up() |
|
|
|
def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]: |
|
text = data.get("text", None) |
|
if text is not None: |
|
assert isinstance(text, str), "Expected text to be a string" |
|
return self.query_embedder.run(text) |
|
|
|
documents = data.get("documents", None) |
|
if documents is not None: |
|
assert isinstance(documents, list), "Expected documents to be a list" |
|
assert all( |
|
isinstance(document, dict) for document in documents |
|
), "Expected each document in documents to be a dictionary" |
|
|
|
documents = [Document(**document) for document in documents] |
|
return self.document_embedder.run(documents) |
|
|
|
raise ValueError("Expected either text or documents") |
|
|