sie-lancedb 0.1.10__tar.gz

sie_lancedb-0.1.10/.gitignore

@@ -0,0 +1,259 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[codz]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py.cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# UV
+#   Similar to Pipfile.lock, it is generally recommended to include uv.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#uv.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+#poetry.toml
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#   pdm recommends including project-wide configuration in pdm.toml, but excluding .pdm-python.
+#   https://pdm-project.org/en/latest/usage/project/#working-with-version-control
+#pdm.lock
+#pdm.toml
+.pdm-python
+.pdm-build/
+
+# pixi
+#   Similar to Pipfile.lock, it is generally recommended to include pixi.lock in version control.
+#pixi.lock
+#   Pixi creates a virtual environment in the .pixi directory, just like venv module creates one
+#   in the .venv directory. It is recommended not to include this directory in version control.
+.pixi
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/
+
+# Abstra
+# Abstra is an AI-powered process automation framework.
+# Ignore directories containing user credentials, local state, and settings.
+# Learn more at https://abstra.io/docs
+.abstra/
+
+# Visual Studio Code
+#  Visual Studio Code specific template is maintained in a separate VisualStudioCode.gitignore
+#  that can be found at https://github.com/github/gitignore/blob/main/Global/VisualStudioCode.gitignore
+#  and can be added to the global gitignore or merged into this file. However, if you prefer,
+#  you could uncomment the following to ignore the entire vscode folder
+# .vscode/
+
+# Ruff stuff:
+.ruff_cache/
+
+# PyPI configuration file
+.pypirc
+
+# Cursor
+#  Cursor is an AI-powered code editor. `.cursorignore` specifies files/directories to
+#  exclude from AI features like autocomplete and code analysis. Recommended for sensitive data
+#  refer to https://docs.cursor.com/context/ignore-files
+.cursorignore
+.cursorindexingignore
+
+# vcscode
+.vscode
+
+# Marimo
+marimo/_static/
+marimo/_lsp/
+__marimo__/
+
+# SIE specific
+# Model weights cache
+.cache/
+*.safetensors
+*.bin
+
+# Secrets (never commit)
+*.pem
+*.key
+credentials.json
+*-key.json
+
+# Terraform
+# Local .terraform directories (cached providers/modules)
+**/.terraform/*
+.terraform.lock*
+# State files (contain sensitive data)
+*.tfstate
+*.tfstate.*
+# Crash log files
+crash.log
+crash.*.log
+# Override files (local developer overrides)
+override.tf
+override.tf.json
+*_override.tf
+*_override.tf.json
+# tfvars files may contain secrets
+*.tfvars
+*.tfvars.json
+# Keep .terraform.lock.hcl for reproducibility (provider versions)
+
+# Node.js
+node_modules/
+
+# OS
+.DS_Store
+Thumbs.db
+
+# VIM
+*.swp
+
+# kilocode
+.kilo/
+
+# Worktree metadata
+.base-branch
+
+# Temporary files
+tmp/
+.tmp/
+.local/
+
+.requirements-modal.txt

sie_lancedb-0.1.10/PKG-INFO

@@ -0,0 +1,12 @@
+Metadata-Version: 2.4
+Name: sie-lancedb
+Version: 0.1.10
+Summary: SIE integration for LanceDB
+Requires-Python: >=3.12
+Requires-Dist: lancedb>=0.17
+Requires-Dist: pyarrow>=14.0
+Requires-Dist: pylance>=0.20
+Requires-Dist: sie-sdk>=0.1.0
+Provides-Extra: dev
+Requires-Dist: pytest-asyncio>=0.24; extra == 'dev'
+Requires-Dist: pytest>=8.0; extra == 'dev'

sie_lancedb-0.1.10/pyproject.toml

@@ -0,0 +1,32 @@
+[project]
+name = "sie-lancedb"
+version = "0.1.10"
+description = "SIE integration for LanceDB"
+requires-python = ">=3.12"
+dependencies = [
+    "sie-sdk>=0.1.0",
+    "lancedb>=0.17",
+    "pylance>=0.20",
+    "pyarrow>=14.0",
+]
+
+[project.optional-dependencies]
+dev = [
+    "pytest>=8.0",
+    "pytest-asyncio>=0.24",
+]
+
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[tool.hatch.build.targets.wheel]
+packages = ["src/sie_lancedb"]
+
+[tool.pytest.ini_options]
+testpaths = ["tests"]
+asyncio_mode = "auto"
+markers = [
+    "integration: marks tests as integration tests (require running SIE server and LanceDB)",
+]
+addopts = "-v --tb=short -m 'not integration'"

sie_lancedb-0.1.10/src/sie_lancedb/__init__.py

@@ -0,0 +1,70 @@
+"""SIE integration for LanceDB.
+
+This package provides LanceDB embedding functions, rerankers, and extractors
+that use SIE for inference:
+
+Embedding Functions (registered in LanceDB's embedding function registry):
+- SIEEmbeddingFunction: Dense embeddings via the "sie" registry name
+- SIEMultiVectorEmbeddingFunction: ColBERT/ColPali multi-vector via "sie-multivector"
+
+Rerankers:
+- SIEReranker: Cross-encoder reranking for hybrid search pipelines
+
+Extractors:
+- SIEExtractor: Entity extraction with table enrichment support
+
+Example usage with embedding function registry:
+
+    import lancedb
+    from lancedb.embeddings import get_registry
+    from lancedb.pydantic import LanceModel, Vector
+    import sie_lancedb  # registers "sie" and "sie-multivector"
+
+    sie = get_registry().get("sie").create(model="BAAI/bge-m3")
+
+    class Documents(LanceModel):
+        text: str = sie.SourceField()
+        vector: Vector(sie.ndims()) = sie.VectorField()
+
+    db = lancedb.connect("~/.lancedb")
+    table = db.create_table("docs", schema=Documents)
+    table.add([{"text": "hello world"}])  # auto-embeds
+    results = table.search("hello").limit(5).to_list()  # auto-embeds query
+
+Example usage with reranker:
+
+    from sie_lancedb import SIEReranker
+
+    table.create_fts_index("text")
+    results = (
+        table.search("hello", query_type="hybrid")
+        .rerank(SIEReranker(model="jinaai/jina-reranker-v2-base-multilingual"))
+        .limit(10)
+        .to_list()
+    )
+
+Example usage with extractor:
+
+    from sie_lancedb import SIEExtractor
+
+    extractor = SIEExtractor(model="urchade/gliner_multi-v2.1")
+    extractor.enrich_table(
+        table,
+        source_column="text",
+        target_column="entities",
+        labels=["person", "organization", "location"],
+        id_column="id",
+    )
+"""
+
+from sie_lancedb.embeddings import SIEEmbeddingFunction, SIEMultiVectorEmbeddingFunction
+from sie_lancedb.extractors import ENTITY_STRUCT, SIEExtractor
+from sie_lancedb.rerankers import SIEReranker
+
+__all__ = [
+    "ENTITY_STRUCT",
+    "SIEEmbeddingFunction",
+    "SIEExtractor",
+    "SIEMultiVectorEmbeddingFunction",
+    "SIEReranker",
+]

sie_lancedb-0.1.10/src/sie_lancedb/embeddings.py

@@ -0,0 +1,276 @@
+"""SIE embedding functions for LanceDB.
+
+Provides LanceDB-native embedding functions registered in the global registry:
+- SIEEmbeddingFunction: Dense text embeddings via "sie"
+- SIEMultiVectorEmbeddingFunction: ColBERT/ColPali multi-vector via "sie-multivector"
+
+These integrate with LanceDB's auto-embedding system: embeddings are computed
+automatically on table.add() and table.search() when the schema uses
+SourceField/VectorField annotations.
+"""
+
+from __future__ import annotations
+
+from typing import Any
+
+import numpy as np
+from lancedb.embeddings import EmbeddingFunction, TextEmbeddingFunction, register
+from pydantic import PrivateAttr
+
+
+@register("sie")
+class SIEEmbeddingFunction(TextEmbeddingFunction):
+    """Dense text embeddings via SIE, registered as "sie" in LanceDB.
+
+    Wraps SIEClient.encode() with output_types=["dense"] to implement
+    LanceDB's TextEmbeddingFunction interface. Embeddings are computed
+    automatically when using SourceField/VectorField annotations.
+
+    Example:
+        >>> from lancedb.embeddings import get_registry
+        >>> from lancedb.pydantic import LanceModel, Vector
+        >>> import sie_lancedb  # registers "sie"
+        >>>
+        >>> sie = (
+        ...     get_registry()
+        ...     .get("sie")
+        ...     .create(
+        ...         model="BAAI/bge-m3",
+        ...         base_url="http://localhost:8080",
+        ...     )
+        ... )
+        >>>
+        >>> class Documents(LanceModel):
+        ...     text: str = sie.SourceField()
+        ...     vector: Vector(sie.ndims()) = sie.VectorField()
+
+    Args:
+        base_url: URL of the SIE server.
+        model: Model name/ID to use for encoding.
+        instruction: Optional instruction prefix for embedding (model-dependent).
+        output_dtype: Output dtype: "float32" (default), "float16", "int8", "binary".
+        options: Runtime options dict passed to the model adapter.
+        gpu: Target GPU type for routing (e.g., "l4", "a100-80gb").
+        timeout_s: Request timeout in seconds.
+    """
+
+    base_url: str = "http://localhost:8080"
+    model: str = "BAAI/bge-m3"
+    instruction: str | None = None
+    output_dtype: str | None = None
+    options: dict[str, Any] | None = None
+    gpu: str | None = None
+    timeout_s: float = 180.0
+
+    _sie_client: Any = PrivateAttr(default=None)
+    _ndims: int | None = PrivateAttr(default=None)
+
+    @property
+    def client(self) -> Any:
+        """Lazily initialize the SIE client."""
+        if self._sie_client is None:
+            from sie_sdk import SIEClient
+
+            self._sie_client = SIEClient(
+                self.base_url,
+                timeout_s=self.timeout_s,
+                gpu=self.gpu,
+                options=self.options,
+            )
+        return self._sie_client
+
+    def ndims(self) -> int:
+        """Return embedding dimensionality.
+
+        Queries ``GET /v1/models/{model}`` on first call (lightweight,
+        reads from model config — no model loading or inference).
+        Cached after the first call per LanceDB's base class convention.
+        """
+        if self._ndims is None:
+            info = self.client.get_model(self.model)
+            dims = info.get("dims") or {}
+            if "dense" not in dims:
+                msg = f"Model '{self.model}' does not support dense embeddings"
+                raise ValueError(msg)
+            self._ndims = dims["dense"]
+        return self._ndims
+
+    def generate_embeddings(
+        self,
+        texts: list[str] | np.ndarray,
+    ) -> list[np.ndarray | None]:
+        """Generate dense embeddings for a list of texts.
+
+        Args:
+            texts: Texts to embed.
+
+        Returns:
+            List of embedding arrays, one per text.
+        """
+        if not len(texts):
+            return []
+
+        from sie_sdk.types import Item
+
+        items = [Item(text=t) for t in texts]
+        results = self.client.encode(
+            self.model,
+            items,
+            output_types=["dense"],
+            instruction=self.instruction,
+            output_dtype=self.output_dtype,
+        )
+
+        return [self._extract_dense(result) for result in results]
+
+    def _extract_dense(self, result: Any) -> list[float]:
+        """Extract dense embedding from encode result."""
+        dense = result.get("dense") if isinstance(result, dict) else getattr(result, "dense", None)
+        if dense is None:
+            msg = "Encode result missing dense embedding"
+            raise ValueError(msg)
+        return dense.tolist() if hasattr(dense, "tolist") else list(dense)
+
+
+@register("sie-multivector")
+class SIEMultiVectorEmbeddingFunction(EmbeddingFunction):
+    """Multi-vector embeddings via SIE, registered as "sie-multivector" in LanceDB.
+
+    Wraps SIEClient.encode() with output_types=["multivector"] for ColBERT
+    and ColPali models. Works with LanceDB's MultiVector type and native
+    MaxSim scoring.
+
+    Example:
+        >>> from lancedb.embeddings import get_registry
+        >>> from lancedb.pydantic import LanceModel, MultiVector
+        >>> import sie_lancedb  # registers "sie-multivector"
+        >>>
+        >>> sie_colbert = (
+        ...     get_registry()
+        ...     .get("sie-multivector")
+        ...     .create(
+        ...         model="jinaai/jina-colbert-v2",
+        ...         base_url="http://localhost:8080",
+        ...     )
+        ... )
+        >>>
+        >>> class Documents(LanceModel):
+        ...     text: str = sie_colbert.SourceField()
+        ...     vector: MultiVector(sie_colbert.ndims()) = sie_colbert.VectorField()
+
+    Args:
+        base_url: URL of the SIE server.
+        model: Multi-vector model name/ID (e.g., ColBERT, ColPali).
+        instruction: Optional instruction prefix for embedding.
+        options: Runtime options dict passed to the model adapter.
+        gpu: Target GPU type for routing (e.g., "l4", "a100-80gb").
+        timeout_s: Request timeout in seconds.
+    """
+
+    base_url: str = "http://localhost:8080"
+    model: str = "jinaai/jina-colbert-v2"
+    instruction: str | None = None
+    options: dict[str, Any] | None = None
+    gpu: str | None = None
+    timeout_s: float = 180.0
+
+    _sie_client: Any = PrivateAttr(default=None)
+    _ndims: int | None = PrivateAttr(default=None)
+
+    @property
+    def client(self) -> Any:
+        """Lazily initialize the SIE client."""
+        if self._sie_client is None:
+            from sie_sdk import SIEClient
+
+            self._sie_client = SIEClient(
+                self.base_url,
+                timeout_s=self.timeout_s,
+                gpu=self.gpu,
+                options=self.options,
+            )
+        return self._sie_client
+
+    def ndims(self) -> int:
+        """Return per-token embedding dimensionality.
+
+        Queries ``GET /v1/models/{model}`` on first call (lightweight,
+        reads from model config — no model loading or inference).
+        Cached after the first call per LanceDB's base class convention.
+        """
+        if self._ndims is None:
+            info = self.client.get_model(self.model)
+            dims = info.get("dims") or {}
+            if "multivector" not in dims:
+                msg = f"Model '{self.model}' does not support multivector embeddings"
+                raise ValueError(msg)
+            self._ndims = dims["multivector"]
+        return self._ndims
+
+    def compute_source_embeddings(
+        self,
+        *args: Any,
+        **kwargs: Any,  # noqa: ARG002
+    ) -> list[list[list[float]] | None]:
+        """Compute multi-vector embeddings for source documents.
+
+        Args:
+            *args: First positional arg is texts (str, list, pa.Array, etc.).
+
+        Returns:
+            List of 2D embeddings (tokens x dims), one per document.
+        """
+        texts = self.sanitize_input(args[0])
+
+        from sie_sdk.types import Item
+
+        items = [Item(text=t) for t in texts]
+        results = self.client.encode(
+            self.model,
+            items,
+            output_types=["multivector"],
+            instruction=self.instruction,
+        )
+
+        return [self._extract_multivector(result) for result in results]
+
+    def compute_query_embeddings(
+        self,
+        *args: Any,
+        **kwargs: Any,  # noqa: ARG002
+    ) -> list[list[list[float]] | None]:
+        """Compute multi-vector embeddings for queries.
+
+        Passes is_query=True for models that differentiate query/document encoding.
+
+        Args:
+            *args: First positional arg is query text(s).
+
+        Returns:
+            List of 2D embeddings (tokens x dims), one per query.
+        """
+        query = args[0]
+        if isinstance(query, str):
+            query = [query]
+        texts = self.sanitize_input(query)
+
+        from sie_sdk.types import Item
+
+        items = [Item(text=t) for t in texts]
+        results = self.client.encode(
+            self.model,
+            items,
+            output_types=["multivector"],
+            instruction=self.instruction,
+            options={"is_query": True},
+        )
+
+        return [self._extract_multivector(result) for result in results]
+
+    def _extract_multivector(self, result: Any) -> list[list[float]]:
+        """Extract multi-vector embedding from encode result."""
+        mv = result.get("multivector") if isinstance(result, dict) else getattr(result, "multivector", None)
+        if mv is None:
+            msg = "Encode result missing multivector embedding"
+            raise ValueError(msg)
+        return mv.tolist() if hasattr(mv, "tolist") else list(mv)