Record cassettes for embedder, ignore huggingface while recording

haiku_rag_slim/haiku/rag/reranking/cohere.py

@@ -12,7 +12,7 @@ except ImportError as e:
 class CohereReranker(RerankerBase):  # pragma: no cover
     def __init__(self):
         # Cohere SDK reads CO_API_KEY from environment by default
-        self._client = cohere.ClientV2()
+        self._client = cohere.AsyncClientV2()
 
     async def rerank(
         self, query: str, chunks: list[Chunk], top_n: int = 10
@@ -23,7 +23,7 @@ class CohereReranker(RerankerBase):  # pragma: no cover
         documents = [chunk.content for chunk in chunks]
 
         model_name = self._model or "rerank-v3.5"
-        response = self._client.rerank(
+        response = await self._client.rerank(
             model=model_name, query=query, documents=documents, top_n=top_n
         )
 

haiku_rag_slim/haiku/rag/reranking/zeroentropy.py

@@ -1,4 +1,4 @@
-from zeroentropy import ZeroEntropy
+from zeroentropy import AsyncZeroEntropy
 
 from haiku.rag.reranking.base import RerankerBase
 from haiku.rag.store.models.chunk import Chunk
@@ -15,7 +15,7 @@ class ZeroEntropyReranker(RerankerBase):  # pragma: no cover
         """
         self._model = model
         # Zero Entropy SDK reads ZEROENTROPY_API_KEY from environment by default
-        self._client = ZeroEntropy()
+        self._client = AsyncZeroEntropy()
 
     async def rerank(
         self, query: str, chunks: list[Chunk], top_n: int = 10
@@ -38,7 +38,7 @@ class ZeroEntropyReranker(RerankerBase):  # pragma: no cover
 
         # Call Zero Entropy reranking API
         model_name = self._model or "zerank-1"
-        response = self._client.models.rerank(
+        response = await self._client.models.rerank(
             model=model_name,
             query=query,
             documents=documents,

pyproject.toml

@@ -71,6 +71,10 @@ dev = [
     "mkdocs>=1.6.1",
     "mkdocs-material>=9.7.0",
     "pre-commit>=4.5.0",
+    "pydantic-ai-slim[anthropic]",
+    "pydantic-ai-slim[bedrock]",
+    "pydantic-ai-slim[google]",
+    "pydantic-ai-slim[groq]",
     "pyright>=1.1.407",
     "pytest>=9.0.2",
     "pytest-asyncio>=1.3.0",

tests/cassettes/test_converters/TestDoclingServeConverterIntegration.test_convert_code_file_real_service.yaml

@@ -1,13 +1,13 @@
 interactions:
 - request:
-    body: "--4379b32700c5b8d95062406eb16d4651\r\nContent-Disposition: form-data; name=\"to_formats\"\r\n\r\njson\r\n--4379b32700c5b8d95062406eb16d4651\r\nContent-Disposition:
+    body: "--313fb28cff30a333a81f610f9c6bdaf3\r\nContent-Disposition: form-data; name=\"to_formats\"\r\n\r\njson\r\n--313fb28cff30a333a81f610f9c6bdaf3\r\nContent-Disposition:
-      form-data; name=\"do_ocr\"\r\n\r\ntrue\r\n--4379b32700c5b8d95062406eb16d4651\r\nContent-Disposition: form-data; name=\"force_ocr\"\r\n\r\nfalse\r\n--4379b32700c5b8d95062406eb16d4651\r\nContent-Disposition:
+      form-data; name=\"do_ocr\"\r\n\r\ntrue\r\n--313fb28cff30a333a81f610f9c6bdaf3\r\nContent-Disposition: form-data; name=\"force_ocr\"\r\n\r\nfalse\r\n--313fb28cff30a333a81f610f9c6bdaf3\r\nContent-Disposition:
-      form-data; name=\"do_table_structure\"\r\n\r\ntrue\r\n--4379b32700c5b8d95062406eb16d4651\r\nContent-Disposition: form-data;
+      form-data; name=\"do_table_structure\"\r\n\r\ntrue\r\n--313fb28cff30a333a81f610f9c6bdaf3\r\nContent-Disposition: form-data;
-      name=\"table_mode\"\r\n\r\naccurate\r\n--4379b32700c5b8d95062406eb16d4651\r\nContent-Disposition: form-data; name=\"table_cell_matching\"\r\n\r\ntrue\r\n--4379b32700c5b8d95062406eb16d4651\r\nContent-Disposition:
+      name=\"table_mode\"\r\n\r\naccurate\r\n--313fb28cff30a333a81f610f9c6bdaf3\r\nContent-Disposition: form-data; name=\"table_cell_matching\"\r\n\r\ntrue\r\n--313fb28cff30a333a81f610f9c6bdaf3\r\nContent-Disposition:
-      form-data; name=\"images_scale\"\r\n\r\n2.0\r\n--4379b32700c5b8d95062406eb16d4651\r\nContent-Disposition: form-data;
+      form-data; name=\"images_scale\"\r\n\r\n2.0\r\n--313fb28cff30a333a81f610f9c6bdaf3\r\nContent-Disposition: form-data;
-      name=\"generate_picture_images\"\r\n\r\nfalse\r\n--4379b32700c5b8d95062406eb16d4651\r\nContent-Disposition: form-data;
+      name=\"generate_picture_images\"\r\n\r\nfalse\r\n--313fb28cff30a333a81f610f9c6bdaf3\r\nContent-Disposition: form-data;
-      name=\"files\"; filename=\"tmpeojgnbxt.md\"\r\nContent-Type: text/markdown\r\n\r\n```python\ndef test():\n    return
+      name=\"files\"; filename=\"tmpgc9za8mg.md\"\r\nContent-Type: text/markdown\r\n\r\n```python\ndef test():\n    return
-      42\n```\r\n--4379b32700c5b8d95062406eb16d4651--\r\n"
+      42\n```\r\n--313fb28cff30a333a81f610f9c6bdaf3--\r\n"
     headers:
       accept:
       - '*/*'
@@ -18,7 +18,7 @@ interactions:
       content-length:
       - '1011'
       content-type:
-      - multipart/form-data; boundary=4379b32700c5b8d95062406eb16d4651
+      - multipart/form-data; boundary=313fb28cff30a333a81f610f9c6bdaf3
       host:
       - localhost:5001
     method: POST
@@ -26,13 +26,13 @@ interactions:
   response:
     headers:
       content-length:
-      - '1114'
+      - '1113'
       content-type:
       - application/json
     parsed_body:
       document:
         doctags_content: null
-        filename: tmpeojgnbxt.md
+        filename: tmpgc9za8mg.md
         html_content: null
         json_content:
           body:
@@ -55,10 +55,10 @@ interactions:
             self_ref: '#/furniture'
           groups: []
           key_value_items: []
-          name: tmpeojgnbxt
+          name: tmpgc9za8mg
           origin:
             binary_hash: 9008975733065065710
-            filename: tmpeojgnbxt.md
+            filename: tmpgc9za8mg.md
             mimetype: text/markdown
             uri: null
           pages: {}
@@ -91,7 +91,7 @@ interactions:
         md_content: null
         text_content: null
       errors: []
-      processing_time: 0.0017275830000471615
+      processing_time: 0.002258999999980915
       status: success
       timings: {}
     status:

tests/cassettes/test_converters/TestDoclingServeConverterIntegration.test_convert_text_real_service.yaml

@@ -1,12 +1,12 @@
 interactions:
 - request:
-    body: "--798e57b119ed24827326455c645311ae\r\nContent-Disposition: form-data; name=\"to_formats\"\r\n\r\njson\r\n--798e57b119ed24827326455c645311ae\r\nContent-Disposition:
+    body: "--a906caa5bbe5d30e17b44fb63c456240\r\nContent-Disposition: form-data; name=\"to_formats\"\r\n\r\njson\r\n--a906caa5bbe5d30e17b44fb63c456240\r\nContent-Disposition:
-      form-data; name=\"do_ocr\"\r\n\r\ntrue\r\n--798e57b119ed24827326455c645311ae\r\nContent-Disposition: form-data; name=\"force_ocr\"\r\n\r\nfalse\r\n--798e57b119ed24827326455c645311ae\r\nContent-Disposition:
+      form-data; name=\"do_ocr\"\r\n\r\ntrue\r\n--a906caa5bbe5d30e17b44fb63c456240\r\nContent-Disposition: form-data; name=\"force_ocr\"\r\n\r\nfalse\r\n--a906caa5bbe5d30e17b44fb63c456240\r\nContent-Disposition:
-      form-data; name=\"do_table_structure\"\r\n\r\ntrue\r\n--798e57b119ed24827326455c645311ae\r\nContent-Disposition: form-data;
+      form-data; name=\"do_table_structure\"\r\n\r\ntrue\r\n--a906caa5bbe5d30e17b44fb63c456240\r\nContent-Disposition: form-data;
-      name=\"table_mode\"\r\n\r\naccurate\r\n--798e57b119ed24827326455c645311ae\r\nContent-Disposition: form-data; name=\"table_cell_matching\"\r\n\r\ntrue\r\n--798e57b119ed24827326455c645311ae\r\nContent-Disposition:
+      name=\"table_mode\"\r\n\r\naccurate\r\n--a906caa5bbe5d30e17b44fb63c456240\r\nContent-Disposition: form-data; name=\"table_cell_matching\"\r\n\r\ntrue\r\n--a906caa5bbe5d30e17b44fb63c456240\r\nContent-Disposition:
-      form-data; name=\"images_scale\"\r\n\r\n2.0\r\n--798e57b119ed24827326455c645311ae\r\nContent-Disposition: form-data;
+      form-data; name=\"images_scale\"\r\n\r\n2.0\r\n--a906caa5bbe5d30e17b44fb63c456240\r\nContent-Disposition: form-data;
-      name=\"generate_picture_images\"\r\n\r\nfalse\r\n--798e57b119ed24827326455c645311ae\r\nContent-Disposition: form-data;
+      name=\"generate_picture_images\"\r\n\r\nfalse\r\n--a906caa5bbe5d30e17b44fb63c456240\r\nContent-Disposition: form-data;
-      name=\"files\"; filename=\"content.md\"\r\nContent-Type: text/markdown\r\n\r\n# Test Document\n\nThis is a test.\r\n--798e57b119ed24827326455c645311ae--\r\n"
+      name=\"files\"; filename=\"content.md\"\r\nContent-Type: text/markdown\r\n\r\n# Test Document\n\nThis is a test.\r\n--a906caa5bbe5d30e17b44fb63c456240--\r\n"
     headers:
       accept:
       - '*/*'
@@ -17,7 +17,7 @@ interactions:
       content-length:
       - '1000'
       content-type:
-      - multipart/form-data; boundary=798e57b119ed24827326455c645311ae
+      - multipart/form-data; boundary=a906caa5bbe5d30e17b44fb63c456240
       host:
       - localhost:5001
     method: POST
@@ -94,7 +94,7 @@ interactions:
         md_content: null
         text_content: null
       errors: []
-      processing_time: 0.002875374999803171
+      processing_time: 0.004331874999934371
       status: success
       timings: {}
     status:

tests/cassettes/test_reranker/test_cohere_reranker.yaml

@@ -0,0 +1,70 @@
+interactions:
+- request:
+    headers:
+      accept:
+      - '*/*'
+      accept-encoding:
+      - gzip, deflate, zstd
+      connection:
+      - keep-alive
+      content-length:
+      - '1261'
+      content-type:
+      - application/json
+      host:
+      - api.cohere.com
+    method: POST
+    parsed_body:
+      documents:
+      - To Kill a Mockingbird is a novel by Harper Lee published in 1960. It was immediately successful, winning the Pulitzer
+        Prize, and has become a classic of modern American literature.
+      - The novel Moby-Dick was written by Herman Melville and first published in 1851. It is considered a masterpiece of
+        American literature and deals with complex themes of obsession, revenge, and the conflict between good and evil.
+      - Harper Lee, an American novelist widely known for her novel To Kill a Mockingbird, was born in 1926 in Monroeville,
+        Alabama. She received the Pulitzer Prize for Fiction in 1961.
+      - Jane Austen was an English novelist known primarily for her six major novels, which interpret, critique and comment
+        upon the British landed gentry at the end of the 18th century.
+      - The Harry Potter series, which consists of seven fantasy novels written by British author J.K. Rowling, is among the
+        most popular and critically acclaimed books of the modern era.
+      - The Great Gatsby, a novel written by American author F. Scott Fitzgerald, was published in 1925. The story is set
+        in the Jazz Age and follows the life of millionaire Jay Gatsby and his pursuit of Daisy Buchanan.
+      model: rerank-v3.5
+      query: Who wrote 'To Kill a Mockingbird'?
+      top_n: 2
+    uri: https://api.cohere.com/v2/rerank
+  response:
+    headers:
+      access-control-expose-headers:
+      - X-Debug-Trace-ID
+      alt-svc:
+      - h3=":443"; ma=2592000,h3-29=":443"; ma=2592000
+      cache-control:
+      - no-cache, no-store, no-transform, must-revalidate, private, max-age=0
+      content-length:
+      - '210'
+      content-type:
+      - application/json
+      expires:
+      - Thu, 01 Jan 1970 00:00:00 GMT
+      pragma:
+      - no-cache
+      transfer-encoding:
+      - chunked
+      vary:
+      - Origin,Accept-Encoding
+    parsed_body:
+      id: 56d8804a-d476-4c6a-ade8-69deb3fc8c64
+      meta:
+        api_version:
+          version: '2'
+        billed_units:
+          search_units: 1
+      results:
+      - index: 0
+        relevance_score: 0.9438932
+      - index: 2
+        relevance_score: 0.8455478
+    status:
+      code: 200
+      message: OK
+version: 1

tests/cassettes/test_reranker/test_zeroentropy_reranker.yaml

@@ -0,0 +1,59 @@
+interactions:
+- request:
+    headers:
+      accept:
+      - application/json
+      accept-encoding:
+      - gzip, deflate, zstd
+      connection:
+      - keep-alive
+      content-length:
+      - '1248'
+      content-type:
+      - application/json
+      host:
+      - api.zeroentropy.dev
+    method: POST
+    parsed_body:
+      documents:
+      - To Kill a Mockingbird is a novel by Harper Lee published in 1960. It was immediately successful, winning the Pulitzer
+        Prize, and has become a classic of modern American literature.
+      - The novel Moby-Dick was written by Herman Melville and first published in 1851. It is considered a masterpiece of
+        American literature and deals with complex themes of obsession, revenge, and the conflict between good and evil.
+      - Harper Lee, an American novelist widely known for her novel To Kill a Mockingbird, was born in 1926 in Monroeville,
+        Alabama. She received the Pulitzer Prize for Fiction in 1961.
+      - Jane Austen was an English novelist known primarily for her six major novels, which interpret, critique and comment
+        upon the British landed gentry at the end of the 18th century.
+      - The Harry Potter series, which consists of seven fantasy novels written by British author J.K. Rowling, is among the
+        most popular and critically acclaimed books of the modern era.
+      - The Great Gatsby, a novel written by American author F. Scott Fitzgerald, was published in 1925. The story is set
+        in the Jazz Age and follows the life of millionaire Jay Gatsby and his pursuit of Daisy Buchanan.
+      model: zerank-1
+      query: Who wrote 'To Kill a Mockingbird'?
+    uri: https://api.zeroentropy.dev/v1/models/rerank
+  response:
+    headers:
+      connection:
+      - keep-alive
+      content-length:
+      - '309'
+      content-type:
+      - application/json
+    parsed_body:
+      results:
+      - index: 2
+        relevance_score: 0.9392035026199311
+      - index: 0
+        relevance_score: 0.918713920386427
+      - index: 5
+        relevance_score: 0.0860773208330919
+      - index: 4
+        relevance_score: 0.08035746882220708
+      - index: 3
+        relevance_score: 0.07673913563377131
+      - index: 1
+        relevance_score: 0.07498651627727318
+    status:
+      code: 200
+      message: OK
+version: 1

tests/conftest.py

@@ -5,11 +5,17 @@ from pathlib import Path
 from typing import TYPE_CHECKING, Any
 
 # Prevent tests from loading user's local haiku.rag.yaml by setting env var
-# to an empty config file BEFORE any haiku.rag imports.
+# to a test config file BEFORE any haiku.rag imports.
-# This ensures tests always use default config values.
+# Uses Ollama for embeddings - HTTP calls are recorded/replayed via VCR.
 _test_config_dir = tempfile.mkdtemp()
 _test_config_path = Path(_test_config_dir) / "test-defaults.yaml"
-_test_config_path.write_text("{}")  # Empty YAML = use all defaults
+_test_config_path.write_text("""
+embeddings:
+  model:
+    provider: ollama
+    name: qwen3-embedding:4b
+    vector_dim: 2560
+""")
 os.environ["HAIKU_RAG_CONFIG_PATH"] = str(_test_config_path)
 
 import pydantic_ai.models  # noqa: E402
@@ -101,6 +107,12 @@ def set_mock_api_keys(monkeypatch):
         monkeypatch.setenv("ZEROENTROPY_API_KEY", "mock-ze-key-for-vcr-playback")
     if not os.getenv("VOYAGE_API_KEY"):
         monkeypatch.setenv("VOYAGE_API_KEY", "mock-voyage-key-for-vcr-playback")
+    if not os.getenv("GROQ_API_KEY"):
+        monkeypatch.setenv("GROQ_API_KEY", "mock-groq-key-for-vcr-playback")
+    if not os.getenv("GOOGLE_API_KEY"):
+        monkeypatch.setenv("GOOGLE_API_KEY", "mock-google-key-for-vcr-playback")
+    if not os.getenv("AWS_DEFAULT_REGION"):
+        monkeypatch.setenv("AWS_DEFAULT_REGION", "us-east-1")
 
 
 def pytest_recording_configure(config: Any, vcr: "VCR"):
@@ -113,6 +125,7 @@ def pytest_recording_configure(config: Any, vcr: "VCR"):
 def vcr_config():
     return {
         "ignore_localhost": False,
+        "ignore_hosts": ["huggingface.co"],
         "filter_headers": ["authorization", "x-api-key"],
         "decode_compressed_response": True,
     }

tests/graph/test_search_filter.py

@@ -33,7 +33,7 @@ async def client_with_docs(temp_db_path):
     client.close()
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_search_filter_restricts_results(client_with_docs):
     """Test that search_filter restricts search to specified documents."""
     client, doc1_id, doc2_id = client_with_docs

tests/store/test_read_only.py

@@ -246,7 +246,7 @@ class TestClientReadOnly:
         assert client.is_read_only is True
         client.close()
 
-    @pytest.mark.asyncio
+    @pytest.mark.vcr()
     async def test_client_create_document_raises_when_read_only(self, temp_db_path):
         """Client.create_document() raises ReadOnlyError when read_only=True."""
         client = HaikuRAG(temp_db_path, create=True)
@@ -256,7 +256,7 @@ class TestClientReadOnly:
             with pytest.raises(ReadOnlyError):
                 await client.create_document("test content")
 
-    @pytest.mark.asyncio
+    @pytest.mark.vcr()
     async def test_client_delete_document_raises_when_read_only(self, temp_db_path):
         """Client.delete_document() raises ReadOnlyError when read_only=True."""
         async with HaikuRAG(temp_db_path, create=True) as client:
@@ -268,7 +268,7 @@ class TestClientReadOnly:
             with pytest.raises(ReadOnlyError):
                 await client.delete_document(doc_id)
 
-    @pytest.mark.asyncio
+    @pytest.mark.vcr()
     async def test_client_search_works_when_read_only(self, temp_db_path):
         """Client.search() works in read-only mode."""
         async with HaikuRAG(temp_db_path, create=True) as client:
@@ -278,7 +278,7 @@ class TestClientReadOnly:
             results = await client.search("cats")
             assert len(results) > 0
 
-    @pytest.mark.asyncio
+    @pytest.mark.vcr()
     async def test_client_list_documents_works_when_read_only(self, temp_db_path):
         """Client.list_documents() works in read-only mode."""
         async with HaikuRAG(temp_db_path, create=True) as client:

tests/test_chunk.py

@@ -6,7 +6,7 @@ from haiku.rag.config import Config
 from haiku.rag.store.models.chunk import Chunk, ChunkMetadata, SearchResult
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_chunk_repository_operations(qa_corpus: Dataset, temp_db_path):
     """Test ChunkRepository operations."""
     # Create client
@@ -47,7 +47,7 @@ async def test_chunk_repository_operations(qa_corpus: Dataset, temp_db_path):
     client.close()
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_chunk_repository_pagination(qa_corpus: Dataset, temp_db_path):
     """Test ChunkRepository pagination with get_by_document_id and count_by_document_id."""
     async with HaikuRAG(db_path=temp_db_path, config=Config, create=True) as client:
@@ -96,7 +96,7 @@ async def test_chunk_repository_pagination(qa_corpus: Dataset, temp_db_path):
         assert len(empty_batch) == 0
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_chunking_pipeline(qa_corpus: Dataset, temp_db_path):
     """Test document chunking using client primitives."""
     from haiku.rag.client import HaikuRAG

tests/test_client.py

@@ -17,7 +17,7 @@ def vcr_cassette_dir():
     return str(Path(__file__).parent / "cassettes" / "test_client")
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_document_crud(qa_corpus: Dataset, temp_db_path):
     """Test HaikuRAG CRUD operations for documents."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -84,7 +84,7 @@ async def test_client_document_crud(qa_corpus: Dataset, temp_db_path):
         assert deleted_again is False
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_update_document(qa_corpus: Dataset, temp_db_path):
     """Test updating document with individual parameters."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -155,7 +155,7 @@ async def test_client_update_document(qa_corpus: Dataset, temp_db_path):
         assert doc_chunks[1].content == "Custom chunk 2"
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_create_document_from_source(temp_db_path):
     """Test creating a document from a file source."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -186,7 +186,7 @@ async def test_client_create_document_from_source(temp_db_path):
             assert "md5" in doc2.metadata
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_create_document_from_source_with_title(temp_db_path):
     """Test creating a document from a file source with a title."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -203,7 +203,7 @@ async def test_client_create_document_from_source_with_title(temp_db_path):
             assert doc.title == "My Doc"
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_update_title_noop_behavior(temp_db_path):
     """When content is unchanged, updating title should update document without re-chunking."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -225,7 +225,7 @@ async def test_client_update_title_noop_behavior(temp_db_path):
             assert got.title == "Title B"
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_create_document_from_source_unsupported(temp_db_path):
     """Test creating a document from an unsupported file type."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -241,7 +241,7 @@ async def test_client_create_document_from_source_unsupported(temp_db_path):
                 await client.create_document_from_source(temp_path)
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_create_document_from_source_nonexistent(temp_db_path):
     """Test creating a document from a non-existent file."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -252,7 +252,7 @@ async def test_client_create_document_from_source_nonexistent(temp_db_path):
             await client.create_document_from_source(non_existent_path)
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_create_document_from_directory(temp_db_path):
     """Test creating documents from a directory recursively."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -287,7 +287,7 @@ async def test_client_create_document_from_directory(temp_db_path):
             assert not any("unsupported.xyz" in uri for uri in uris)
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_create_document_from_directory_with_filters(
     monkeypatch, temp_db_path
 ):
@@ -336,7 +336,7 @@ async def test_client_create_document_from_directory_with_filters(
             assert not any("debug.log" in uri for uri in uris)
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_create_document_from_url(temp_db_path):
     """Test creating a document from a URL."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -362,7 +362,7 @@ async def test_client_create_document_from_url(temp_db_path):
             assert doc.metadata["contentType"] == "text/html"
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_create_document_from_url_with_different_content_types(
     temp_db_path,
 ):
@@ -409,7 +409,7 @@ async def test_client_create_document_from_url_with_different_content_types(
             assert doc.metadata["contentType"] == "text/plain"
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_create_document_from_url_unsupported_content(temp_db_path):
     """Test creating a document from URL with unsupported content type."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -426,7 +426,7 @@ async def test_client_create_document_from_url_unsupported_content(temp_db_path)
                 )
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_create_document_from_url_http_error(temp_db_path):
     """Test handling HTTP errors when creating document from URL."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -443,7 +443,7 @@ async def test_client_create_document_from_url_http_error(temp_db_path):
                 )
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_get_extension_from_content_type_or_url(temp_db_path):
     """Test the helper method for determining file extensions."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -488,7 +488,7 @@ async def test_get_extension_from_content_type_or_url(temp_db_path):
         )
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_metadata_content_type_and_md5(temp_db_path):
     """Test that contentType and md5 metadata are correctly set."""
     import hashlib
@@ -523,7 +523,7 @@ async def test_client_metadata_content_type_and_md5(temp_db_path):
                 assert url_doc.metadata["md5"] == expected_md5
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_create_update_no_op_behavior(temp_db_path):
     """Test create/update/no-op behavior based on MD5 changes."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -562,7 +562,7 @@ async def test_client_create_update_no_op_behavior(temp_db_path):
             assert retrieved_doc.content == updated_content
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_unchanged_file_keeps_timestamp(temp_db_path):
     """Test that unchanged files don't update the updated_at timestamp."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -584,7 +584,7 @@ async def test_client_unchanged_file_keeps_timestamp(temp_db_path):
             assert doc2.updated_at == original_updated_at  # Timestamp should not change
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_url_create_update_no_op_behavior(temp_db_path):
     """Test create/update/no-op behavior for URLs based on MD5 changes."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -623,7 +623,7 @@ async def test_client_url_create_update_no_op_behavior(temp_db_path):
             assert doc3.content == updated_content.decode()  # Updated content
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_search(temp_db_path):
     """Test HaikuRAG search functionality."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -666,7 +666,7 @@ async def test_client_search(temp_db_path):
         assert len(limited_results) <= 1
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_async_context_manager(temp_db_path):
     """Test HaikuRAG as async context manager."""
 
@@ -691,7 +691,7 @@ async def test_client_async_context_manager(temp_db_path):
     # but the test passing means the context manager methods work correctly
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_import_document_with_custom_chunks(temp_db_path):
     """Test importing a document with pre-created chunks."""
     from docling_core.types.doc.document import DoclingDocument
@@ -762,7 +762,7 @@ async def test_client_ask(allow_model_requests, temp_db_path):
         assert isinstance(citations, list)
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_expand_context(temp_db_path):
     """Test that expand_context method exists and works with basic input."""
     from haiku.rag.store.models import SearchResult
@@ -779,7 +779,7 @@ async def test_client_expand_context(temp_db_path):
         assert expanded_results[0].score == 0.9
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_create_document_stores_docling_json(temp_db_path):
     """Test that create_document stores DoclingDocument JSON."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -801,7 +801,7 @@ async def test_client_create_document_stores_docling_json(temp_db_path):
         assert parsed["version"] == doc.docling_version
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_import_document_stores_docling_data(temp_db_path):
     """Test that import_document stores DoclingDocument data correctly."""
     from docling_core.types.doc.document import DoclingDocument
@@ -828,7 +828,7 @@ async def test_client_import_document_stores_docling_data(temp_db_path):
         assert doc.docling_version == docling_doc.version
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_create_document_from_file_stores_docling_json(temp_db_path):
     """Test that create_document_from_source stores DoclingDocument JSON for files."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -850,7 +850,7 @@ async def test_client_create_document_from_file_stores_docling_json(temp_db_path
             assert retrieved.docling_version == doc.docling_version
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_update_document_stores_docling_json(temp_db_path):
     """Test that update_document stores DoclingDocument JSON when content changes."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -870,7 +870,7 @@ async def test_client_update_document_stores_docling_json(temp_db_path):
         assert updated_doc.docling_document_json != original_json
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_update_document_with_custom_chunks_no_docling_json(
     temp_db_path,
 ):
@@ -891,7 +891,7 @@ async def test_client_update_document_with_custom_chunks_no_docling_json(
         assert updated_doc.docling_document_json == original_json
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_update_document_content_docling_mutually_exclusive(
     temp_db_path,
 ):
@@ -915,7 +915,7 @@ async def test_client_update_document_content_docling_mutually_exclusive(
             )
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_update_document_with_docling_rechunks(temp_db_path):
     """Test that providing docling_document without chunks triggers rechunk."""
     from docling_core.types.doc.document import DoclingDocument
@@ -952,7 +952,7 @@ async def test_client_update_document_with_docling_rechunks(temp_db_path):
         assert new_chunks[0].content != original_chunks[0].content
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_update_document_docling_with_chunks(temp_db_path):
     """Test that providing both docling_document and chunks stores both."""
     from docling_core.types.doc.document import DoclingDocument
@@ -990,7 +990,7 @@ async def test_client_update_document_docling_with_chunks(temp_db_path):
         assert chunks[1].content == "Custom chunk 2"
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_file_update_stores_docling_json(temp_db_path):
     """Test that updating a file re-stores DoclingDocument JSON."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -1018,7 +1018,7 @@ async def test_client_file_update_stores_docling_json(temp_db_path):
             assert doc2.docling_version == original_version  # Version stays same
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_visualize_chunk_no_document(temp_db_path):
     """Test visualize_chunk returns empty list when chunk has no document_id."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -1027,7 +1027,7 @@ async def test_client_visualize_chunk_no_document(temp_db_path):
         assert images == []
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_visualize_chunk_no_bounding_boxes(temp_db_path):
     """Test visualize_chunk returns empty list when chunk has no bounding boxes."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -1049,7 +1049,7 @@ async def test_client_visualize_chunk_no_bounding_boxes(temp_db_path):
         assert images == []
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_visualize_chunk_returns_list(temp_db_path):
     """Test visualize_chunk returns a list (empty or with images)."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -1082,7 +1082,7 @@ This is paragraph four about topic C.
 
 
 @pytest.mark.integration
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_visualize_chunk_with_pdf(temp_db_path):
     """Test visualize_chunk returns images with bounding boxes for PDF documents."""
     from PIL.Image import Image as PILImage
@@ -1120,7 +1120,7 @@ async def test_client_visualize_chunk_with_pdf(temp_db_path):
 # =============================================================================
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_convert_text(temp_db_path):
     """Test convert() with plain text content."""
     from docling_core.types.doc.document import DoclingDocument
@@ -1135,7 +1135,7 @@ async def test_client_convert_text(temp_db_path):
         assert "test content" in markdown
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_convert_file(temp_db_path):
     """Test convert() with a file path."""
     from docling_core.types.doc.document import DoclingDocument
@@ -1152,7 +1152,7 @@ async def test_client_convert_file(temp_db_path):
             assert "File content" in markdown
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_convert_file_not_found(temp_db_path):
     """Test convert() raises ValueError for non-existent file."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -1160,7 +1160,7 @@ async def test_client_convert_file_not_found(temp_db_path):
             await client.convert(Path("/nonexistent/path/file.txt"))
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_convert_unsupported_extension(temp_db_path):
     """Test convert() raises ValueError for unsupported file extension."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -1172,7 +1172,7 @@ async def test_client_convert_unsupported_extension(temp_db_path):
                 await client.convert(temp_path)
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_convert_file_uri(temp_db_path):
     """Test convert() with a file:// URI string."""
     from docling_core.types.doc.document import DoclingDocument
@@ -1195,7 +1195,7 @@ async def test_client_convert_file_uri(temp_db_path):
 # =============================================================================
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_chunk_basic(temp_db_path):
     """Test chunk() produces Chunk objects from DoclingDocument."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -1215,7 +1215,7 @@ async def test_client_chunk_basic(temp_db_path):
         assert all(c.document_id is None for c in chunks)
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_chunk_preserves_metadata(temp_db_path):
     """Test chunk() preserves structured metadata from DoclingDocument."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -1244,7 +1244,7 @@ This is a subsection.
         assert has_metadata, "Chunks should have structured metadata"
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_chunk_empty_document(temp_db_path):
     """Test chunk() with empty DoclingDocument."""
     from docling_core.types.doc.document import DoclingDocument
@@ -1259,7 +1259,7 @@ async def test_client_chunk_empty_document(temp_db_path):
         assert len(chunks) == 0
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_import_document_embeds_chunks_without_embeddings(temp_db_path):
     """Test that import_document embeds chunks that don't have embeddings."""
     from docling_core.types.doc.document import DoclingDocument
@@ -1295,7 +1295,7 @@ async def test_import_document_embeds_chunks_without_embeddings(temp_db_path):
         assert results[0].content == "First chunk without embedding"
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_update_document_embeds_chunks_without_embeddings(temp_db_path):
     """Test that update_document embeds chunks that don't have embeddings."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -1324,7 +1324,7 @@ async def test_update_document_embeds_chunks_without_embeddings(temp_db_path):
         assert results[0].content == "Updated chunk without embedding"
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_create_document_with_html_format(temp_db_path):
     """Test create_document with HTML format preserves document structure."""
     async with HaikuRAG(temp_db_path, create=True) as client:
@@ -1359,7 +1359,7 @@ async def test_client_create_document_with_html_format(temp_db_path):
         assert "list_item" in labels
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_client_convert_with_html_format(temp_db_path):
     """Test convert with HTML format."""
     async with HaikuRAG(temp_db_path, create=True) as client:

tests/test_context_enhancement.py

@@ -86,7 +86,7 @@ def small_chunk_config() -> AppConfig:
     return config
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_table_expansion_includes_split_rows(temp_db_path, small_chunk_config):
     """Verify that table expansion retrieves rows that were split into different chunks."""
     docling_doc = create_table_document()
@@ -129,7 +129,7 @@ async def test_table_expansion_includes_split_rows(temp_db_path, small_chunk_con
             )
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_list_expansion_includes_split_items(temp_db_path, small_chunk_config):
     """Verify that list expansion retrieves items that were split into different chunks."""
     docling_doc = create_list_document()
@@ -177,7 +177,7 @@ async def test_list_expansion_includes_split_items(temp_db_path, small_chunk_con
             )
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_code_expansion_includes_adjacent_blocks(
     temp_db_path, small_chunk_config
 ):
@@ -223,7 +223,7 @@ async def test_code_expansion_includes_adjacent_blocks(
             )
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_text_expansion_uses_radius(temp_db_path):
     """Text content expansion should use radius, not structural boundaries."""
     config = AppConfig()
@@ -265,7 +265,7 @@ async def test_text_expansion_uses_radius(temp_db_path):
         assert len(expanded[0].content) >= len(original.content)
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_expansion_preserves_metadata(temp_db_path, small_chunk_config):
     """Expansion should preserve document metadata."""
     docling_doc = create_table_document()
@@ -284,7 +284,7 @@ async def test_expansion_preserves_metadata(temp_db_path, small_chunk_config):
         assert expanded[0].document_id == results[0].document_id
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_format_for_agent_output(temp_db_path, small_chunk_config):
     """format_for_agent should include source, type, and content sections."""
     docling_doc = create_table_document()
@@ -307,7 +307,7 @@ async def test_format_for_agent_output(temp_db_path, small_chunk_config):
         assert "Content:" in formatted
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_max_items_limit_caps_expansion(temp_db_path):
     """Expansion should respect max_context_items limit."""
     config = AppConfig()
@@ -334,7 +334,7 @@ async def test_max_items_limit_caps_expansion(temp_db_path):
         assert item_count <= 2, f"Expected at most 2 items, got {item_count}"
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_search_result_get_primary_label():
     """Test _get_primary_label prioritizes structural labels correctly."""
     # Table should be prioritized
@@ -388,7 +388,7 @@ async def test_search_result_get_primary_label():
     assert result._get_primary_label() is None
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_expand_context_radius_zero(temp_db_path):
     """Test expand_context with radius 0 returns original results."""
     # Default config has context_radius=0
@@ -406,7 +406,7 @@ async def test_expand_context_radius_zero(temp_db_path):
         assert expanded_results[0].score == search_results[0].score
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_expand_context_multiple_documents(temp_db_path):
     """Test expand_context with results from multiple documents."""
     config = AppConfig()
@@ -467,7 +467,7 @@ async def test_expand_context_multiple_documents(temp_db_path):
         assert "Doc2 Section Y" in expanded2.content
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_expand_context_merges_overlapping_chunks(temp_db_path):
     """Test that overlapping expanded chunks are merged into one."""
     config = AppConfig()
@@ -522,7 +522,7 @@ async def test_expand_context_merges_overlapping_chunks(temp_db_path):
         assert merged.score == 0.8
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_expand_context_keeps_separate_non_overlapping(temp_db_path):
     """Test that non-overlapping expanded chunks remain separate."""
     config = AppConfig()
@@ -587,7 +587,7 @@ async def test_expand_context_keeps_separate_non_overlapping(temp_db_path):
         assert chunk5_expanded.score == 0.7
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_expand_context_with_docling_merges_overlapping(temp_db_path):
     """Test that expand_context with DoclingDocument merges overlapping results."""
     config = AppConfig()
@@ -642,7 +642,7 @@ This is paragraph four about topic C.
                 assert r.doc_item_refs is not None
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_expand_context_docling_merges_metadata(temp_db_path):
     """Test that expand_context properly merges metadata from multiple results."""
     config = AppConfig()

tests/test_database_autocreate.py

@@ -30,7 +30,7 @@ def test_database_created_with_create_flag():
         client.close()
 
 
-@pytest.mark.asyncio
+@pytest.mark.vcr()
 async def test_operations_work_after_database_created():
     """Test that operations work after DB is created."""
     with tempfile.TemporaryDirectory() as tmpdir: