Skip to content

teff.rag.stores.milvus

teff.rag.stores.milvus

Milvus vector store — requires pymilvus.

Classes:

Name Description
MilvusVectorStore

Vector store backed by Milvus.

MilvusVectorStore

Bases: VectorStore

Vector store backed by Milvus.

Requires the pymilvus package. Uses cosine distance (scores are similarity-like, higher = more similar). Works against a Milvus server or a local Milvus Lite file when the uri is a local path (pymilvus[milvus_lite]).

Source code in teff/rag/stores/milvus.py
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
class MilvusVectorStore(VectorStore):
    """Vector store backed by Milvus.

    Requires the ``pymilvus`` package. Uses cosine distance (scores are
    similarity-like, higher = more similar). Works against a Milvus server
    or a local Milvus Lite file when the ``uri`` is a local path
    (``pymilvus[milvus_lite]``).
    """

    def __init__(
        self,
        uri: str = "./milvus.db",
        token: str = "",
        collection: str = "teff",
        dim: int | None = None,
    ):
        from pymilvus import MilvusClient

        self._client = MilvusClient(uri=uri, token=token or None)
        self.collection = collection
        self.dim = dim
        self._created = False

    def _ensure_collection(self, dim: int) -> None:
        if self._created:
            return
        from pymilvus import DataType, MilvusClient

        if self._client.has_collection(self.collection):
            self._created = True
            return
        schema = MilvusClient.create_schema(auto_id=False, enable_dynamic_field=True)
        schema.add_field(
            "id", datatype=DataType.VARCHAR, is_primary=True, max_length=512
        )
        schema.add_field("vector", datatype=DataType.FLOAT_VECTOR, dim=dim)
        index_params = MilvusClient.prepare_index_params()
        index_params.add_index(
            field_name="vector", index_type="AUTOINDEX", metric_type="COSINE"
        )
        self._client.create_collection(
            self.collection, schema=schema, index_params=index_params
        )
        self.dim = dim
        self._created = True

    @staticmethod
    def _meta_from(entity: dict) -> dict:
        return {k: v for k, v in entity.items() if k not in ("id", "vector")}

    async def add(self, vectors: list[tuple[str, list[float], dict]]) -> None:
        if not vectors:
            return
        self._ensure_collection(len(vectors[0][1]))
        rows = [{"id": vid, "vector": vec, **meta} for vid, vec, meta in vectors]
        self._client.insert(self.collection, rows)

    async def search(
        self,
        query: list[float],
        k: int = 10,
        filter: dict | None = None,
        hybrid: bool = False,
        query_text: str | None = None,
    ) -> list[tuple[str, float, dict]]:
        if not self._client.has_collection(self.collection):
            return []
        limit = max(k, k * 4) if hybrid else k
        res = self._client.search(
            self.collection,
            data=[query],
            limit=limit,
            output_fields=["*"],
            filter=_to_milvus_expr(filter),
        )
        if not res:
            return []
        candidates = [
            (hit["id"], float(hit["distance"]), self._meta_from(hit.get("entity", {})))
            for hit in res[0]
        ]
        return finalize_results(
            candidates, k, filter=None, hybrid=hybrid, query_text=query_text
        )

    async def delete(self, ids: list[str]) -> None:
        if not self._client.has_collection(self.collection):
            return
        vals = ", ".join(json.dumps(i) for i in ids)
        self._client.delete(self.collection, filter=f"id in [{vals}]")

    async def count(self) -> int:
        if not self._client.has_collection(self.collection):
            return 0
        return int(self._client.get_collection_stats(self.collection)["row_count"])

    async def entries(
        self, limit: int = 100, offset: int = 0
    ) -> list[tuple[str, dict]]:
        if not self._client.has_collection(self.collection):
            return []
        res = self._client.query(
            self.collection, filter="", output_fields=["*"], limit=limit, offset=offset
        )
        return [(row["id"], self._meta_from(dict(row))) for row in res]

    async def get(self, ids: list[str]) -> list[tuple[str, dict]]:
        if not ids or not self._client.has_collection(self.collection):
            return []
        vals = ", ".join(json.dumps(i) for i in ids)
        res = self._client.query(
            self.collection, filter=f"id in [{vals}]", output_fields=["*"]
        )
        return [(row["id"], self._meta_from(dict(row))) for row in res]

    async def update_metadata(self, id: str, metadata: dict) -> None:
        if not self._client.has_collection(self.collection):
            return
        res = self._client.query(
            self.collection, filter=f"id == {json.dumps(id)}", output_fields=["*"]
        )
        rows = list(res)
        if not rows:
            return
        row = dict(rows[0])
        vector = row.get("vector")
        merged = {**self._meta_from(row), **metadata}
        self._client.upsert(self.collection, [{"id": id, "vector": vector, **merged}])

    async def clear(self) -> None:
        if self._client.has_collection(self.collection):
            self._client.drop_collection(self.collection)
            self._created = False