Haystack (deepset) integration

Astrocyte as Retriever and Writer pipeline components for Haystack.

Module: astrocyte.integrations.haystack Pattern: Pipeline components — arun() returning {"documents": [...]} or {"written": N} Framework dependency: haystack-ai (optional)

Install

pip install astrocyte haystack-ai

Usage — Retriever

from astrocyte import Astrocyte
from astrocyte.integrations.haystack import AstrocyteRetriever

brain = Astrocyte.from_config("astrocyte.yaml")
retriever = AstrocyteRetriever(brain, bank_id="knowledge-base", top_k=10)

# Use in a Haystack pipeline
from haystack import Pipeline
pipe = Pipeline()
pipe.add_component("retriever", retriever)
pipe.add_component("reader", reader)
pipe.connect("retriever.documents", "reader.documents")

result = pipe.run({"retriever": {"query": "What is dark mode?"}})

# Or use standalone
docs = await retriever.arun("dark mode", top_k=5)
# → {"documents": [AstrocyteDocument(content="...", score=0.9, meta={...}), ...]}

Usage — Writer

from astrocyte.integrations.haystack import AstrocyteWriter, AstrocyteDocument

writer = AstrocyteWriter(brain, bank_id="knowledge-base")

# Write documents to memory
await writer.arun([
    AstrocyteDocument(content="New finding", meta={"source": "research"}, score=1.0, id="d1"),
    {"content": "Dict format also works", "meta": {"source": "upload"}},
])
# → {"written": 2}

Integration pattern

Component	Method	Input	Output
AstrocyteRetriever	arun(query)	Query string	{"documents": list[AstrocyteDocument]}
AstrocyteRetriever	run(query)	Query string (sync)	Same
AstrocyteWriter	arun(documents)	List of docs/dicts	{"written": int}

AstrocyteDocument

Compatible with haystack.Document:

Field	Type	Description
content	str	Document text
meta	dict	Metadata
score	float	Relevance score
id	str	Document/memory ID

API reference

AstrocyteRetriever(brain, bank_id, *, top_k=10)

AstrocyteWriter(brain, bank_id)