OpenRAG is a comprehensive, single package Retrieval-Augmented Generation platform built on Langflow, Docling, and Opensearch.
Python
"""Writes that change what is searchable must drop the corpus-dependent caches."""
import pytest
from services import search_cache
from services.document_index_writer import (
DocumentIndexChunk,
DocumentIndexContext,
DocumentIndexWriter,
)
from utils.opensearch_delete import delete_document_ids
def _prime_caches() -> None:
search_cache.SUGGEST_CACHE.set("g", {"files": ["old.pdf"]})
search_cache.SPACES_CACHE.set("s", ["space"])
search_cache.QUERY_EMBEDDING_CACHE.set("q", [0.1])
def _assert_corpus_caches_dropped() -> None:
assert search_cache.SUGGEST_CACHE.get("g") is None
assert search_cache.SPACES_CACHE.get("s") is None
assert search_cache.QUERY_EMBEDDING_CACHE.get("q") == [0.1]
class _Indices:
async def exists(self, *, index):
return True
async def get_mapping(self, *, index):
return {}
async def refresh(self, *, index):
return None
class _WriteClient:
def __init__(self, bulk_errors=False):
self.indices = _Indices()
self.bulk_errors = bulk_errors
async def bulk(self, *, body, refresh):
items = [
{"index": {"_id": body[i]["index"]["_id"], "status": 500, "error": {"type": "boom"}}}
if self.bulk_errors
else {"index": {"_id": body[i]["index"]["_id"], "status": 200}}
for i in range(0, len(body), 2)
]
return {"errors": self.bulk_errors, "items": items}
async def delete_by_query(self, **_kwargs):
return {"deleted": 2}
def _writer(monkeypatch, client):
writer = DocumentIndexWriter(opensearch_client=client)
async def no_op_ensure(*_args, **_kwargs):
return "embedding_field"
monkeypatch.setattr(writer, "_ensure_index_and_embedding_field", no_op_ensure)
monkeypatch.setattr("config.settings.get_index_name", lambda: "documents")
return writer
def _context():
return DocumentIndexContext(
document_id="doc",
filename="report.pdf",
mimetype="application/pdf",
embedding_model="test-model",
owner="user-a",
)
_CHUNKS = [DocumentIndexChunk(chunk_id="doc_0", text="t", vector=[0.1, 0.2])]
@pytest.mark.asyncio
async def test_indexing_chunks_drops_corpus_caches(monkeypatch):
writer = _writer(monkeypatch, _WriteClient())
_prime_caches()
await writer.index_chunks(_context(), _CHUNKS)
_assert_corpus_caches_dropped()
@pytest.mark.asyncio
async def test_a_failed_batch_still_drops_corpus_caches(monkeypatch):
writer = _writer(monkeypatch, _WriteClient(bulk_errors=True))
monkeypatch.setattr("services.document_index_writer._BULK_RETRY_DELAY_SECONDS", 0)
_prime_caches()
with pytest.raises(Exception): # noqa: B017 - any bulk failure type
await writer.index_chunks(_context(), _CHUNKS)
_assert_corpus_caches_dropped()
@pytest.mark.asyncio
async def test_deleting_an_ingest_run_drops_corpus_caches(monkeypatch):
writer = _writer(monkeypatch, _WriteClient())
_prime_caches()
await writer.delete_ingest_run("run-1")
_assert_corpus_caches_dropped()
class _DeleteClient:
def __init__(self, results):
self.results = results
async def delete(self, *, index, id, refresh):
result = self.results[id]
if isinstance(result, Exception):
raise result
return {"result": result}
@pytest.mark.asyncio
async def test_deleting_documents_drops_corpus_caches():
_prime_caches()
deleted = await delete_document_ids(
_DeleteClient({"a": "deleted"}), index="documents", document_ids=["a"]
)
assert deleted == 1
_assert_corpus_caches_dropped()
@pytest.mark.asyncio
async def test_a_delete_that_removed_nothing_keeps_the_caches():
_prime_caches()
await delete_document_ids(
_DeleteClient({"a": "not_found"}), index="documents", document_ids=["a"]
)
assert search_cache.SUGGEST_CACHE.get("g") is not None
@pytest.mark.asyncio
async def test_a_delete_that_fails_midway_still_drops_caches():
_prime_caches()
with pytest.raises(RuntimeError):
await delete_document_ids(
_DeleteClient({"a": "deleted", "b": RuntimeError("down")}),
index="documents",
document_ids=["a", "b"],
)
_assert_corpus_caches_dropped()