diff --git a/CHANGELOG.md b/CHANGELOG.md
index 84674d9..94586c3 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -13,6 +13,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/), and this
 - Soft-delete chunks: delete button with confirmation dialog, chunks hidden from UI but preserved in data
 - Vector index metadata schema: `IndexedChunk` domain model, OpenSearch mapping builder, configurable embedding dimension
 - `VectorStore` port (Protocol): `ensure_index`, `index_chunks`, `search_similar`, `get_chunks`, `delete_document`
+- OpenSearch adapter (`OpenSearchStore`): kNN vector search, full-text search, bulk indexing, document CRUD
 
 ### Fixed
 
diff --git a/document-parser/domain/ports.py b/document-parser/domain/ports.py
index aa852cc..0375da5 100644
--- a/document-parser/domain/ports.py
+++ b/document-parser/domain/ports.py
@@ -6,7 +6,7 @@ Infrastructure adapters (local Docling, Docling Serve, etc.) implement these.
 
 from __future__ import annotations
 
-from typing import TYPE_CHECKING, Protocol
+from typing import TYPE_CHECKING, Protocol, runtime_checkable
 
 if TYPE_CHECKING:
     from domain.models import AnalysisJob, Document
@@ -82,6 +82,7 @@ class AnalysisRepository(Protocol):
     async def delete_by_document(self, document_id: str) -> int: ...
 
 
+@runtime_checkable
 class VectorStore(Protocol):
     """Port for vector storage and retrieval.
 
diff --git a/document-parser/infra/opensearch_store.py b/document-parser/infra/opensearch_store.py
new file mode 100644
index 0000000..c4a97f9
--- /dev/null
+++ b/document-parser/infra/opensearch_store.py
@@ -0,0 +1,204 @@
+"""OpenSearch adapter implementing the VectorStore port.
+
+Uses the opensearch-py client for kNN vector search, full-text search,
+and document CRUD against an OpenSearch cluster.
+"""
+
+from __future__ import annotations
+
+import logging
+from typing import Any
+
+from opensearchpy import AsyncOpenSearch, NotFoundError
+
+from domain.vector_schema import (
+    ChunkBboxEntry,
+    ChunkOrigin,
+    DocItemRef,
+    IndexedChunk,
+    SearchResult,
+)
+
+logger = logging.getLogger(__name__)
+
+
+def _hit_to_indexed_chunk(hit: dict[str, Any]) -> IndexedChunk:
+    """Reconstruct an IndexedChunk from an OpenSearch _source document."""
+    src = hit["_source"]
+    origin_raw = src.get("origin")
+    origin = (
+        ChunkOrigin(binary_hash=origin_raw["binary_hash"], filename=origin_raw["filename"])
+        if origin_raw
+        else None
+    )
+    return IndexedChunk(
+        doc_id=src["doc_id"],
+        filename=src["filename"],
+        content=src["content"],
+        embedding=src.get("embedding", []),
+        chunk_index=src["chunk_index"],
+        chunk_type=src["chunk_type"],
+        page_number=src["page_number"],
+        bboxes=[
+            ChunkBboxEntry(page=b["page"], x=b["x"], y=b["y"], w=b["w"], h=b["h"])
+            for b in src.get("bboxes", [])
+        ],
+        headings=src.get("headings", []),
+        doc_items=[
+            DocItemRef(self_ref=d["self_ref"], label=d["label"]) for d in src.get("doc_items", [])
+        ],
+        origin=origin,
+    )
+
+
+def _hit_to_result(hit: dict[str, Any]) -> SearchResult:
+    """Convert an OpenSearch hit to a SearchResult."""
+    return SearchResult(
+        chunk=_hit_to_indexed_chunk(hit),
+        score=hit.get("_score", 0.0),
+    )
+
+
+class OpenSearchStore:
+    """Concrete VectorStore adapter backed by OpenSearch.
+
+    Satisfies the ``VectorStore`` Protocol defined in ``domain.ports``.
+
+    Args:
+        url: OpenSearch cluster URL (e.g. ``http://localhost:9200``).
+        verify_certs: Whether to verify TLS certificates.
+    """
+
+    def __init__(self, url: str, *, verify_certs: bool = False) -> None:
+        self._client = AsyncOpenSearch(
+            hosts=[url],
+            use_ssl=url.startswith("https"),
+            verify_certs=verify_certs,
+            ssl_show_warn=False,
+        )
+
+    # -- lifecycle -------------------------------------------------------------
+
+    async def close(self) -> None:
+        """Close the underlying HTTP connection pool."""
+        await self._client.close()
+
+    # -- VectorStore protocol methods ------------------------------------------
+
+    async def ensure_index(self, index_name: str, mapping: dict) -> None:
+        """Create the index if it does not exist. No-op if it already exists."""
+        exists = await self._client.indices.exists(index=index_name)
+        if not exists:
+            await self._client.indices.create(index=index_name, body=mapping)
+            logger.info("Created OpenSearch index '%s'", index_name)
+        else:
+            logger.debug("Index '%s' already exists — skipping creation", index_name)
+
+    async def index_chunks(self, index_name: str, chunks: list[IndexedChunk]) -> int:
+        """Bulk-index a list of chunks. Returns the number successfully indexed."""
+        if not chunks:
+            return 0
+
+        body: list[dict[str, Any]] = []
+        for chunk in chunks:
+            doc_id = f"{chunk.doc_id}_{chunk.chunk_index}"
+            body.append({"index": {"_index": index_name, "_id": doc_id}})
+            body.append(chunk.to_dict())
+
+        resp = await self._client.bulk(body=body, refresh="wait_for")
+
+        errors = sum(1 for item in resp["items"] if item["index"].get("error"))
+        indexed = len(chunks) - errors
+        if errors:
+            logger.warning("Bulk index to '%s': %d/%d failed", index_name, errors, len(chunks))
+        return indexed
+
+    async def search_similar(
+        self,
+        index_name: str,
+        embedding: list[float],
+        *,
+        k: int = 10,
+        doc_id: str | None = None,
+    ) -> list[SearchResult]:
+        """kNN search for the k nearest chunks by embedding similarity."""
+        knn_query: dict[str, Any] = {
+            "knn": {
+                "embedding": {
+                    "vector": embedding,
+                    "k": k,
+                },
+            },
+        }
+        if doc_id:
+            knn_query["knn"]["embedding"]["filter"] = {
+                "term": {"doc_id": doc_id},
+            }
+
+        resp = await self._client.search(
+            index=index_name,
+            body={"size": k, "query": knn_query},
+            _source_excludes=["embedding"],
+        )
+        return [_hit_to_result(hit) for hit in resp["hits"]["hits"]]
+
+    async def get_chunks(
+        self,
+        index_name: str,
+        doc_id: str,
+        *,
+        limit: int = 1000,
+    ) -> list[SearchResult]:
+        """Retrieve all indexed chunks for a document, ordered by chunk_index."""
+        resp = await self._client.search(
+            index=index_name,
+            body={
+                "size": limit,
+                "query": {"term": {"doc_id": doc_id}},
+                "sort": [{"chunk_index": {"order": "asc"}}],
+            },
+            _source_excludes=["embedding"],
+        )
+        return [_hit_to_result(hit) for hit in resp["hits"]["hits"]]
+
+    async def delete_document(self, index_name: str, doc_id: str) -> int:
+        """Delete all chunks for a document. Returns the number deleted."""
+        try:
+            resp = await self._client.delete_by_query(
+                index=index_name,
+                body={"query": {"term": {"doc_id": doc_id}}},
+                refresh=True,
+            )
+            deleted: int = resp.get("deleted", 0)
+            return deleted
+        except NotFoundError:
+            return 0
+
+    # -- full-text search (bonus from spec) ------------------------------------
+
+    async def search_fulltext(
+        self,
+        index_name: str,
+        query_text: str,
+        *,
+        k: int = 10,
+        doc_id: str | None = None,
+    ) -> list[SearchResult]:
+        """Full-text search on the content field.
+
+        This method is not part of the VectorStore protocol but is specified
+        in the issue acceptance criteria.
+        """
+        must: list[dict[str, Any]] = [{"match": {"content": query_text}}]
+        if doc_id:
+            must.append({"term": {"doc_id": doc_id}})
+
+        resp = await self._client.search(
+            index=index_name,
+            body={
+                "size": k,
+                "query": {"bool": {"must": must}},
+            },
+            _source_excludes=["embedding"],
+        )
+        return [_hit_to_result(hit) for hit in resp["hits"]["hits"]]
diff --git a/document-parser/requirements.txt b/document-parser/requirements.txt
index 67cae25..81f72a7 100644
--- a/document-parser/requirements.txt
+++ b/document-parser/requirements.txt
@@ -7,3 +7,4 @@ pillow>=10.0.0,<11.0.0
 aiosqlite>=0.20.0,<1.0.0
 httpx>=0.27.0,<1.0.0
 pypdfium2>=4.0.0,<5.0.0
+opensearch-py[async]>=2.6.0,<3.0.0
diff --git a/document-parser/tests/test_opensearch_store.py b/document-parser/tests/test_opensearch_store.py
new file mode 100644
index 0000000..e29bf0a
--- /dev/null
+++ b/document-parser/tests/test_opensearch_store.py
@@ -0,0 +1,320 @@
+"""Tests for the OpenSearch adapter (infra.opensearch_store).
+
+These tests mock the AsyncOpenSearch client to validate adapter logic
+without requiring a running OpenSearch instance.
+"""
+
+from __future__ import annotations
+
+from unittest.mock import AsyncMock
+
+import pytest
+
+from domain.ports import VectorStore
+from domain.vector_schema import (
+    ChunkBboxEntry,
+    ChunkOrigin,
+    DocItemRef,
+    IndexedChunk,
+    SearchResult,
+    build_index_mapping,
+)
+from infra.opensearch_store import OpenSearchStore, _hit_to_indexed_chunk, _hit_to_result
+
+# -- Fixtures -----------------------------------------------------------------
+
+
+def _make_chunk(
+    doc_id: str = "doc-1",
+    chunk_index: int = 0,
+    content: str = "hello world",
+    embedding: list[float] | None = None,
+) -> IndexedChunk:
+    return IndexedChunk(
+        doc_id=doc_id,
+        filename="test.pdf",
+        content=content,
+        embedding=embedding or [0.1, 0.2, 0.3],
+        chunk_index=chunk_index,
+        chunk_type="text",
+        page_number=1,
+        bboxes=[ChunkBboxEntry(page=1, x=0.0, y=0.0, w=100.0, h=50.0)],
+        headings=["Chapter 1"],
+        doc_items=[DocItemRef(self_ref="#/texts/0", label="text")],
+        origin=ChunkOrigin(binary_hash="abc123", filename="test.pdf"),
+    )
+
+
+def _make_hit(
+    doc_id: str = "doc-1",
+    chunk_index: int = 0,
+    score: float = 0.95,
+    content: str = "hello world",
+) -> dict:
+    return {
+        "_id": f"{doc_id}_{chunk_index}",
+        "_score": score,
+        "_source": {
+            "doc_id": doc_id,
+            "filename": "test.pdf",
+            "content": content,
+            "chunk_index": chunk_index,
+            "chunk_type": "text",
+            "page_number": 1,
+            "bboxes": [{"page": 1, "x": 0.0, "y": 0.0, "w": 100.0, "h": 50.0}],
+            "headings": ["Chapter 1"],
+            "doc_items": [{"self_ref": "#/texts/0", "label": "text"}],
+            "origin": {"binary_hash": "abc123", "filename": "test.pdf"},
+        },
+    }
+
+
+@pytest.fixture
+def store() -> OpenSearchStore:
+    return OpenSearchStore("http://localhost:9200")
+
+
+@pytest.fixture
+def mock_client(store: OpenSearchStore) -> AsyncMock:
+    client = AsyncMock()
+    store._client = client
+    return client
+
+
+# -- Protocol satisfaction -----------------------------------------------------
+
+
+class TestProtocolSatisfaction:
+    def test_satisfies_vector_store_protocol(self) -> None:
+        """OpenSearchStore structurally satisfies VectorStore Protocol."""
+        store = OpenSearchStore("http://localhost:9200")
+        assert isinstance(store, VectorStore)
+
+
+# -- Hit deserialization -------------------------------------------------------
+
+
+class TestHitDeserialization:
+    def test_hit_to_indexed_chunk(self) -> None:
+        hit = _make_hit()
+        chunk = _hit_to_indexed_chunk(hit)
+        assert isinstance(chunk, IndexedChunk)
+        assert chunk.doc_id == "doc-1"
+        assert chunk.content == "hello world"
+        assert chunk.chunk_index == 0
+        assert chunk.page_number == 1
+        assert len(chunk.bboxes) == 1
+        assert chunk.bboxes[0].w == 100.0
+        assert len(chunk.doc_items) == 1
+        assert chunk.doc_items[0].label == "text"
+        assert chunk.origin is not None
+        assert chunk.origin.binary_hash == "abc123"
+
+    def test_hit_to_indexed_chunk_no_origin(self) -> None:
+        hit = _make_hit()
+        hit["_source"]["origin"] = None
+        chunk = _hit_to_indexed_chunk(hit)
+        assert chunk.origin is None
+
+    def test_hit_to_indexed_chunk_missing_optional_fields(self) -> None:
+        hit = _make_hit()
+        del hit["_source"]["bboxes"]
+        del hit["_source"]["headings"]
+        del hit["_source"]["doc_items"]
+        del hit["_source"]["origin"]
+        chunk = _hit_to_indexed_chunk(hit)
+        assert chunk.bboxes == []
+        assert chunk.headings == []
+        assert chunk.doc_items == []
+        assert chunk.origin is None
+
+    def test_hit_to_result(self) -> None:
+        hit = _make_hit(score=0.87)
+        result = _hit_to_result(hit)
+        assert isinstance(result, SearchResult)
+        assert result.score == 0.87
+        assert result.chunk.doc_id == "doc-1"
+
+
+# -- ensure_index --------------------------------------------------------------
+
+
+class TestEnsureIndex:
+    async def test_creates_index_when_not_exists(
+        self, store: OpenSearchStore, mock_client: AsyncMock
+    ) -> None:
+        mock_client.indices.exists.return_value = False
+        mapping = build_index_mapping()
+        await store.ensure_index("test-index", mapping)
+        mock_client.indices.create.assert_awaited_once_with(index="test-index", body=mapping)
+
+    async def test_noop_when_index_exists(
+        self, store: OpenSearchStore, mock_client: AsyncMock
+    ) -> None:
+        mock_client.indices.exists.return_value = True
+        await store.ensure_index("test-index", {})
+        mock_client.indices.create.assert_not_awaited()
+
+
+# -- index_chunks --------------------------------------------------------------
+
+
+class TestIndexChunks:
+    async def test_bulk_indexes_chunks(
+        self, store: OpenSearchStore, mock_client: AsyncMock
+    ) -> None:
+        chunks = [_make_chunk(chunk_index=0), _make_chunk(chunk_index=1)]
+        mock_client.bulk.return_value = {
+            "errors": False,
+            "items": [
+                {"index": {"_id": "doc-1_0", "status": 201}},
+                {"index": {"_id": "doc-1_1", "status": 201}},
+            ],
+        }
+        count = await store.index_chunks("test-index", chunks)
+        assert count == 2
+        mock_client.bulk.assert_awaited_once()
+
+    async def test_returns_zero_for_empty_list(
+        self, store: OpenSearchStore, mock_client: AsyncMock
+    ) -> None:
+        count = await store.index_chunks("test-index", [])
+        assert count == 0
+        mock_client.bulk.assert_not_awaited()
+
+    async def test_counts_partial_failures(
+        self, store: OpenSearchStore, mock_client: AsyncMock
+    ) -> None:
+        chunks = [_make_chunk(chunk_index=0), _make_chunk(chunk_index=1)]
+        mock_client.bulk.return_value = {
+            "errors": True,
+            "items": [
+                {"index": {"_id": "doc-1_0", "status": 201}},
+                {"index": {"_id": "doc-1_1", "error": {"reason": "mapping"}}},
+            ],
+        }
+        count = await store.index_chunks("test-index", chunks)
+        assert count == 1
+
+    async def test_bulk_body_structure(
+        self, store: OpenSearchStore, mock_client: AsyncMock
+    ) -> None:
+        chunk = _make_chunk(doc_id="d1", chunk_index=3)
+        mock_client.bulk.return_value = {
+            "errors": False,
+            "items": [{"index": {"_id": "d1_3", "status": 201}}],
+        }
+        await store.index_chunks("idx", [chunk])
+        call_body = mock_client.bulk.call_args[1]["body"]
+        assert call_body[0] == {"index": {"_index": "idx", "_id": "d1_3"}}
+        assert call_body[1]["doc_id"] == "d1"
+        assert call_body[1]["chunk_index"] == 3
+
+
+# -- search_similar ------------------------------------------------------------
+
+
+class TestSearchSimilar:
+    async def test_knn_search(self, store: OpenSearchStore, mock_client: AsyncMock) -> None:
+        mock_client.search.return_value = {"hits": {"hits": [_make_hit(score=0.99)]}}
+        results = await store.search_similar("idx", [0.1, 0.2, 0.3], k=5)
+        assert len(results) == 1
+        assert results[0].score == 0.99
+        call_body = mock_client.search.call_args[1]["body"]
+        assert call_body["query"]["knn"]["embedding"]["vector"] == [0.1, 0.2, 0.3]
+        assert call_body["query"]["knn"]["embedding"]["k"] == 5
+
+    async def test_knn_search_with_doc_filter(
+        self, store: OpenSearchStore, mock_client: AsyncMock
+    ) -> None:
+        mock_client.search.return_value = {"hits": {"hits": []}}
+        await store.search_similar("idx", [0.1], doc_id="doc-42")
+        call_body = mock_client.search.call_args[1]["body"]
+        assert call_body["query"]["knn"]["embedding"]["filter"] == {"term": {"doc_id": "doc-42"}}
+
+    async def test_knn_search_no_filter_by_default(
+        self, store: OpenSearchStore, mock_client: AsyncMock
+    ) -> None:
+        mock_client.search.return_value = {"hits": {"hits": []}}
+        await store.search_similar("idx", [0.1])
+        call_body = mock_client.search.call_args[1]["body"]
+        assert "filter" not in call_body["query"]["knn"]["embedding"]
+
+
+# -- get_chunks ----------------------------------------------------------------
+
+
+class TestGetChunks:
+    async def test_retrieves_by_doc_id(
+        self, store: OpenSearchStore, mock_client: AsyncMock
+    ) -> None:
+        mock_client.search.return_value = {
+            "hits": {"hits": [_make_hit(chunk_index=0), _make_hit(chunk_index=1)]}
+        }
+        results = await store.get_chunks("idx", "doc-1")
+        assert len(results) == 2
+        call_body = mock_client.search.call_args[1]["body"]
+        assert call_body["query"] == {"term": {"doc_id": "doc-1"}}
+        assert call_body["sort"] == [{"chunk_index": {"order": "asc"}}]
+
+    async def test_respects_limit(self, store: OpenSearchStore, mock_client: AsyncMock) -> None:
+        mock_client.search.return_value = {"hits": {"hits": []}}
+        await store.get_chunks("idx", "doc-1", limit=50)
+        call_body = mock_client.search.call_args[1]["body"]
+        assert call_body["size"] == 50
+
+
+# -- delete_document -----------------------------------------------------------
+
+
+class TestDeleteDocument:
+    async def test_deletes_by_doc_id(self, store: OpenSearchStore, mock_client: AsyncMock) -> None:
+        mock_client.delete_by_query.return_value = {"deleted": 5}
+        count = await store.delete_document("idx", "doc-1")
+        assert count == 5
+        call_body = mock_client.delete_by_query.call_args[1]["body"]
+        assert call_body["query"] == {"term": {"doc_id": "doc-1"}}
+
+    async def test_returns_zero_on_not_found(
+        self, store: OpenSearchStore, mock_client: AsyncMock
+    ) -> None:
+        from opensearchpy import NotFoundError
+
+        mock_client.delete_by_query.side_effect = NotFoundError(404, "index_not_found")
+        count = await store.delete_document("idx", "doc-1")
+        assert count == 0
+
+
+# -- search_fulltext -----------------------------------------------------------
+
+
+class TestSearchFulltext:
+    async def test_fulltext_search(self, store: OpenSearchStore, mock_client: AsyncMock) -> None:
+        mock_client.search.return_value = {
+            "hits": {"hits": [_make_hit(content="matching text", score=1.5)]}
+        }
+        results = await store.search_fulltext("idx", "matching")
+        assert len(results) == 1
+        assert results[0].chunk.content == "matching text"
+        call_body = mock_client.search.call_args[1]["body"]
+        assert {"match": {"content": "matching"}} in call_body["query"]["bool"]["must"]
+
+    async def test_fulltext_search_with_doc_filter(
+        self, store: OpenSearchStore, mock_client: AsyncMock
+    ) -> None:
+        mock_client.search.return_value = {"hits": {"hits": []}}
+        await store.search_fulltext("idx", "query", doc_id="doc-5")
+        call_body = mock_client.search.call_args[1]["body"]
+        must_clauses = call_body["query"]["bool"]["must"]
+        assert {"term": {"doc_id": "doc-5"}} in must_clauses
+
+
+# -- close ---------------------------------------------------------------------
+
+
+class TestClose:
+    async def test_close_delegates_to_client(
+        self, store: OpenSearchStore, mock_client: AsyncMock
+    ) -> None:
+        await store.close()
+        mock_client.close.assert_awaited_once()