def test_pipeline(document_store, retriever):
documents = [
{
"name": "name_1",
"text": "text_1",
"embedding": np.random.rand(768).astype(np.float32)
},
{
"name": "name_2",
"text": "text_2",
"embedding": np.random.rand(768).astype(np.float32)
},
{
"name": "name_3",
"text": "text_3",
"embedding": np.random.rand(768).astype(np.float64)
},
{
"name": "name_4",
"text": "text_4",
"embedding": np.random.rand(768).astype(np.float32)
},
]
document_store.write_documents(documents)
pipeline = Pipeline()
pipeline.add_node(component=retriever, name="FAISS", inputs=["Query"])
output = pipeline.run(query="How to test this?", top_k_retriever=3)
assert len(output["documents"]) == 3
def test_load_yaml(document_store_with_docs):
# # test correct load from yaml
pipeline = Pipeline.load_from_yaml(Path("samples/pipeline/test_pipeline.yaml", pipeline_name="my_query"))
prediction = pipeline.run(query="Who lives in Berlin?", top_k_retriever=10, top_k_reader=3)
assert prediction["query"] == "Who lives in Berlin?"
assert prediction["answers"][0]["answer"] == "Carla"
# test invalid pipeline name
with pytest.raises(Exception):
Pipeline.load_from_yaml(path=Path("samples/pipeline/test_pipeline.yaml"), pipeline_name="invalid")
def test_graph_creation(reader, retriever_with_docs, document_store_with_docs):
pipeline = Pipeline()
pipeline.add_node(name="ES", component=retriever_with_docs, inputs=["Query"])
with pytest.raises(AssertionError):
pipeline.add_node(name="Reader", component=retriever_with_docs, inputs=["ES.output_2"])
with pytest.raises(AssertionError):
pipeline.add_node(name="Reader", component=retriever_with_docs, inputs=["ES.wrong_edge_label"])
with pytest.raises(Exception):
pipeline.add_node(name="Reader", component=retriever_with_docs, inputs=["InvalidNode"])
def test_load_yaml(document_store_with_docs):
# test correct load of indexing pipeline from yaml
pipeline = Pipeline.load_from_yaml(Path("samples/pipeline/test_pipeline.yaml"),
pipeline_name="test_indexing_pipeline")
pipeline.run(file_path=Path("samples/pdf/sample_pdf_1.pdf"), top_k_retriever=10, top_k_reader=3)
# test correct load of query pipeline from yaml
pipeline = Pipeline.load_from_yaml(Path("samples/pipeline/test_pipeline.yaml"), pipeline_name="test_query_pipeline")
prediction = pipeline.run(query="Who made the PDF specification?", top_k_retriever=10, top_k_reader=3)
assert prediction["query"] == "Who made the PDF specification?"
assert prediction["answers"][0]["answer"] == "Adobe Systems"
# test invalid pipeline name
with pytest.raises(Exception):
Pipeline.load_from_yaml(path=Path("samples/pipeline/test_pipeline.yaml"), pipeline_name="invalid")
def test_load_and_save_yaml(document_store_with_docs, tmp_path):
# test correct load of indexing pipeline from yaml
pipeline = Pipeline.load_from_yaml(Path("samples/pipeline/test_pipeline.yaml"), pipeline_name="indexing_pipeline")
pipeline.run(file_path=Path("samples/pdf/sample_pdf_1.pdf"), top_k_retriever=10, top_k_reader=3)
# test correct load of query pipeline from yaml
pipeline = Pipeline.load_from_yaml(Path("samples/pipeline/test_pipeline.yaml"), pipeline_name="query_pipeline")
prediction = pipeline.run(query="Who made the PDF specification?", top_k_retriever=10, top_k_reader=3)
assert prediction["query"] == "Who made the PDF specification?"
assert prediction["answers"][0]["answer"] == "Adobe Systems"
# test invalid pipeline name
with pytest.raises(Exception):
Pipeline.load_from_yaml(path=Path("samples/pipeline/test_pipeline.yaml"), pipeline_name="invalid")
# test config export
pipeline.save_to_yaml(tmp_path / "test.yaml")
with open(tmp_path/"test.yaml", "r", encoding='utf-8') as stream:
saved_yaml = stream.read()
expected_yaml = '''
components:
- name: ESRetriever
params:
document_store: ElasticsearchDocumentStore
type: ElasticsearchRetriever
- name: ElasticsearchDocumentStore
params:
index: haystack_test_document
label_index: haystack_test_label
type: ElasticsearchDocumentStore
- name: Reader
params:
model_name_or_path: deepset/roberta-base-squad2
no_ans_boost: -10
type: FARMReader
pipelines:
- name: query
nodes:
- inputs:
- Query
name: ESRetriever
- inputs:
- ESRetriever
name: Reader
type: Query
version: '0.8'
'''
assert saved_yaml.replace(" ", "").replace("\n", "") == expected_yaml.replace(" ", "").replace("\n", "")
def test_join_document_pipeline(document_store_with_docs, reader):
es = ElasticsearchRetriever(document_store=document_store_with_docs)
dpr = DensePassageRetriever(
document_store=document_store_with_docs,
query_embedding_model="facebook/dpr-question_encoder-single-nq-base",
passage_embedding_model="facebook/dpr-ctx_encoder-single-nq-base",
use_gpu=False,
)
document_store_with_docs.update_embeddings(dpr)
query = "Where does Carla lives?"
# test merge without weights
join_node = JoinDocuments(join_mode="merge")
p = Pipeline()
p.add_node(component=es, name="R1", inputs=["Query"])
p.add_node(component=dpr, name="R2", inputs=["Query"])
p.add_node(component=join_node, name="Join", inputs=["R1", "R2"])
results = p.run(query=query)
assert len(results["documents"]) == 3
# test merge with weights
join_node = JoinDocuments(join_mode="merge",
weights=[1000, 1],
top_k_join=2)
p = Pipeline()
p.add_node(component=es, name="R1", inputs=["Query"])
p.add_node(component=dpr, name="R2", inputs=["Query"])
p.add_node(component=join_node, name="Join", inputs=["R1", "R2"])
results = p.run(query=query)
assert results["documents"][0].score > 1000
assert len(results["documents"]) == 2
# test concatenate
join_node = JoinDocuments(join_mode="concatenate")
p = Pipeline()
p.add_node(component=es, name="R1", inputs=["Query"])
p.add_node(component=dpr, name="R2", inputs=["Query"])
p.add_node(component=join_node, name="Join", inputs=["R1", "R2"])
results = p.run(query=query)
assert len(results["documents"]) == 3
# test join_node with reader
join_node = JoinDocuments()
p = Pipeline()
p.add_node(component=es, name="R1", inputs=["Query"])
p.add_node(component=dpr, name="R2", inputs=["Query"])
p.add_node(component=join_node, name="Join", inputs=["R1", "R2"])
p.add_node(component=reader, name="Reader", inputs=["Join"])
results = p.run(query=query)
assert results["answers"][0]["answer"] == "Berlin"
def test_query_keyword_statement_classifier():
class KeywordOutput(RootNode):
outgoing_edges = 2
def run(self, **kwargs):
kwargs["output"] = "keyword"
return kwargs, "output_1"
class QuestionOutput(RootNode):
outgoing_edges = 2
def run(self, **kwargs):
kwargs["output"] = "question"
return kwargs, "output_2"
pipeline = Pipeline()
pipeline.add_node(
name="SkQueryKeywordQuestionClassifier",
component=SklearnQueryClassifier(),
inputs=["Query"],
)
pipeline.add_node(
name="KeywordNode",
component=KeywordOutput(),
inputs=["SkQueryKeywordQuestionClassifier.output_2"],
)
pipeline.add_node(
name="QuestionNode",
component=QuestionOutput(),
inputs=["SkQueryKeywordQuestionClassifier.output_1"],
)
output = pipeline.run(query="morse code")
assert output["output"] == "keyword"
output = pipeline.run(query="How old is John?")
assert output["output"] == "question"
pipeline = Pipeline()
pipeline.add_node(
name="TfQueryKeywordQuestionClassifier",
component=TransformersQueryClassifier(),
inputs=["Query"],
)
pipeline.add_node(
name="KeywordNode",
component=KeywordOutput(),
inputs=["TfQueryKeywordQuestionClassifier.output_2"],
)
pipeline.add_node(
name="QuestionNode",
component=QuestionOutput(),
inputs=["TfQueryKeywordQuestionClassifier.output_1"],
)
output = pipeline.run(query="morse code")
assert output["output"] == "keyword"
output = pipeline.run(query="How old is John?")
assert output["output"] == "question"
def test_parallel_paths_in_pipeline_graph_with_branching():
class AWithOutput1(RootNode):
outgoing_edges = 2
def run(self, **kwargs):
kwargs["output"] = "A"
return kwargs, "output_1"
class AWithOutput2(RootNode):
outgoing_edges = 2
def run(self, **kwargs):
kwargs["output"] = "A"
return kwargs, "output_2"
class AWithOutputAll(RootNode):
outgoing_edges = 2
def run(self, **kwargs):
kwargs["output"] = "A"
return kwargs, "output_all"
class B(RootNode):
def run(self, **kwargs):
kwargs["output"] += "B"
return kwargs, "output_1"
class C(RootNode):
def run(self, **kwargs):
kwargs["output"] += "C"
return kwargs, "output_1"
class D(RootNode):
def run(self, **kwargs):
kwargs["output"] += "D"
return kwargs, "output_1"
class E(RootNode):
def run(self, **kwargs):
kwargs["output"] += "E"
return kwargs, "output_1"
class JoinNode(RootNode):
def run(self, **kwargs):
if kwargs.get("inputs"):
kwargs["output"] = ""
for input_dict in kwargs["inputs"]:
kwargs["output"] += input_dict["output"]
return kwargs, "output_1"
pipeline = Pipeline()
pipeline.add_node(name="A", component=AWithOutput1(), inputs=["Query"])
pipeline.add_node(name="B", component=B(), inputs=["A.output_1"])
pipeline.add_node(name="C", component=C(), inputs=["A.output_2"])
pipeline.add_node(name="D", component=E(), inputs=["B"])
pipeline.add_node(name="E", component=D(), inputs=["B"])
pipeline.add_node(name="F", component=JoinNode(), inputs=["D", "E", "C"])
output = pipeline.run(query="test")
assert output["output"] == "ABEABD"
pipeline = Pipeline()
pipeline.add_node(name="A", component=AWithOutput2(), inputs=["Query"])
pipeline.add_node(name="B", component=B(), inputs=["A.output_1"])
pipeline.add_node(name="C", component=C(), inputs=["A.output_2"])
pipeline.add_node(name="D", component=E(), inputs=["B"])
pipeline.add_node(name="E", component=D(), inputs=["B"])
pipeline.add_node(name="F", component=JoinNode(), inputs=["D", "E", "C"])
output = pipeline.run(query="test")
assert output["output"] == "AC"
pipeline = Pipeline()
pipeline.add_node(name="A", component=AWithOutputAll(), inputs=["Query"])
pipeline.add_node(name="B", component=B(), inputs=["A.output_1"])
pipeline.add_node(name="C", component=C(), inputs=["A.output_2"])
pipeline.add_node(name="D", component=E(), inputs=["B"])
pipeline.add_node(name="E", component=D(), inputs=["B"])
pipeline.add_node(name="F", component=JoinNode(), inputs=["D", "E", "C"])
output = pipeline.run(query="test")
assert output["output"] == "ACABEABD"
def test_parallel_paths_in_pipeline_graph():
class A(RootNode):
def run(self, **kwargs):
kwargs["output"] = "A"
return kwargs, "output_1"
class B(RootNode):
def run(self, **kwargs):
kwargs["output"] += "B"
return kwargs, "output_1"
class C(RootNode):
def run(self, **kwargs):
kwargs["output"] += "C"
return kwargs, "output_1"
class D(RootNode):
def run(self, **kwargs):
kwargs["output"] += "D"
return kwargs, "output_1"
class E(RootNode):
def run(self, **kwargs):
kwargs["output"] += "E"
return kwargs, "output_1"
class JoinNode(RootNode):
def run(self, **kwargs):
kwargs["output"] = (kwargs["inputs"][0]["output"] +
kwargs["inputs"][1]["output"])
return kwargs, "output_1"
pipeline = Pipeline()
pipeline.add_node(name="A", component=A(), inputs=["Query"])
pipeline.add_node(name="B", component=B(), inputs=["A"])
pipeline.add_node(name="C", component=C(), inputs=["B"])
pipeline.add_node(name="E", component=E(), inputs=["C"])
pipeline.add_node(name="D", component=D(), inputs=["B"])
pipeline.add_node(name="F", component=JoinNode(), inputs=["D", "E"])
output = pipeline.run(query="test")
assert output["output"] == "ABDABCE"
pipeline = Pipeline()
pipeline.add_node(name="A", component=A(), inputs=["Query"])
pipeline.add_node(name="B", component=B(), inputs=["A"])
pipeline.add_node(name="C", component=C(), inputs=["B"])
pipeline.add_node(name="D", component=D(), inputs=["B"])
pipeline.add_node(name="E", component=JoinNode(), inputs=["C", "D"])
output = pipeline.run(query="test")
assert output["output"] == "ABCABD"
import os
import shutil
import uuid
from pathlib import Path
from typing import Optional, List
from fastapi import APIRouter, UploadFile, File, Form, HTTPException
from haystack.pipeline import Pipeline
from rest_api.config import PIPELINE_YAML_PATH, FILE_UPLOAD_PATH, INDEXING_PIPELINE_NAME
logger = logging.getLogger(__name__)
router = APIRouter()
try:
INDEXING_PIPELINE = Pipeline.load_from_yaml(
Path(PIPELINE_YAML_PATH), pipeline_name=INDEXING_PIPELINE_NAME)
except KeyError:
INDEXING_PIPELINE = None
logger.info(
"Indexing Pipeline not found in the YAML configuration. File Upload API will not be available."
)
os.makedirs(FILE_UPLOAD_PATH,
exist_ok=True) # create directory for uploading files
@router.post("/file-upload")
def file_upload(
files: List[UploadFile] = File(...),
meta: Optional[str] = Form("null"), # JSON serialized string
remove_numeric_tables: Optional[bool] = Form(None),