cbrkit 1.2.0__tar.gz → 1.4.0__tar.gz

{cbrkit-1.2.0 → cbrkit-1.4.0}/PKG-INFO

@@ -1,16 +1,16 @@
-Metadata-Version: 2.3
+Metadata-Version: 2.4
 Name: cbrkit
-Version: 1.2.0
+Version: 1.4.0
 Summary: Customizable Case-Based Reasoning (CBR) toolkit for Python with a built-in API and CLI
 Keywords: cbr,case-based reasoning,api,similarity,nlp,retrieval,cli,tool,library
 Author: Mirko Lenz
 Author-email: Mirko Lenz <mirko@mirkolenz.com>
+License-Expression: MIT
 Classifier: Development Status :: 4 - Beta
 Classifier: Environment :: Console
 Classifier: Framework :: Pytest
 Classifier: Intended Audience :: Developers
 Classifier: Intended Audience :: Science/Research
-Classifier: License :: OSI Approved :: MIT License
 Classifier: Natural Language :: English
 Classifier: Operating System :: OS Independent
 Classifier: Programming Language :: Python :: 3.13
@@ -30,7 +30,7 @@ Requires-Dist: pyyaml>=6,<7
 Requires-Dist: rtoml>=0.12,<1
 Requires-Dist: scipy>=1,<2
 Requires-Dist: xmltodict>=1,<2
-Requires-Dist: cbrkit[anthropic,api,bm25,chromadb,chunking,cohere,eval,google,graphs,graphviz,instructor,lancedb,levenshtein,nltk,ollama,openai,openai-agents,pandas,pydantic-ai,spacy,sql,timeseries,transformers,voyageai,zvec] ; extra == 'all'
+Requires-Dist: cbrkit[anthropic,api,bm25,chromadb,chunking,cohere,eval,google,graphs,graphviz,instructor,lancedb,levenshtein,nltk,ollama,openai,openai-agents,pandas,pgvector,pydantic-ai,spacy,sql,sqlite-vec,timeseries,transformers,voyageai,zvec] ; extra == 'all'
 Requires-Dist: anthropic>=0.40,<1 ; extra == 'anthropic'
 Requires-Dist: cbrkit[cli] ; extra == 'api'
 Requires-Dist: fastapi>=0.100,<1 ; extra == 'api'
@@ -41,11 +41,11 @@ Requires-Dist: fastmcp>=3,<4 ; extra == 'api'
 Requires-Dist: bm25s[core,stem,indexing]>=0.3,<1 ; extra == 'bm25'
 Requires-Dist: chromadb>=1,<2 ; extra == 'chromadb'
 Requires-Dist: chonkie>=1,<2 ; extra == 'chunking'
-Requires-Dist: rich>=13,<15 ; extra == 'cli'
-Requires-Dist: typer>=0.9,<1 ; extra == 'cli'
-Requires-Dist: cohere>=5,<6 ; extra == 'cohere'
+Requires-Dist: rich>=14,<16 ; extra == 'cli'
+Requires-Dist: typer>=0.20,<1 ; extra == 'cli'
+Requires-Dist: cohere>=7,<8 ; extra == 'cohere'
 Requires-Dist: ranx>=0.3,<1 ; extra == 'eval'
-Requires-Dist: google-genai>=1,<2 ; extra == 'google'
+Requires-Dist: google-genai>=2,<3 ; extra == 'google'
 Requires-Dist: networkx>=3,<4 ; extra == 'graphs'
 Requires-Dist: rustworkx>=0.15,<1 ; extra == 'graphs'
 Requires-Dist: pygraphviz>=1,<2 ; extra == 'graphviz'
@@ -58,14 +58,20 @@ Requires-Dist: openai>=1,<3 ; extra == 'openai'
 Requires-Dist: tiktoken>=0.8,<1 ; extra == 'openai'
 Requires-Dist: openai-agents>=0.2,<1 ; extra == 'openai-agents'
 Requires-Dist: pandas>=2,<4 ; extra == 'pandas'
+Requires-Dist: pgvector>=0.4,<1 ; extra == 'pgvector'
+Requires-Dist: cbrkit[sql] ; extra == 'pgvector'
 Requires-Dist: pydantic-ai-slim>=1,<2 ; extra == 'pydantic-ai'
 Requires-Dist: spacy>=3.8,<4 ; extra == 'spacy'
-Requires-Dist: sqlalchemy>=2,<3 ; extra == 'sql'
+Requires-Dist: sqlalchemy[asyncio]>=2,<3 ; extra == 'sql'
+Requires-Dist: sqlite-vec>=0.1,<1 ; extra == 'sqlite-vec'
+Requires-Dist: aiosqlite>=0.20,<1 ; extra == 'sqlite-vec'
+Requires-Dist: cbrkit[sql] ; extra == 'sqlite-vec'
 Requires-Dist: minineedle>=3,<4 ; extra == 'timeseries'
 Requires-Dist: sentence-transformers>=4,<6 ; extra == 'transformers'
 Requires-Dist: torch>=2.5,<3 ; extra == 'transformers'
 Requires-Dist: transformers>=4,<6 ; extra == 'transformers'
 Requires-Dist: voyageai>=0.3,<1 ; extra == 'voyageai'
+Requires-Dist: zvec>=0.2,<1 ; extra == 'zvec'
 Requires-Python: >=3.13, <4
 Project-URL: Repository, https://github.com/wi2trier/cbrkit
 Project-URL: Documentation, https://wi2trier.github.io/cbrkit/
@@ -91,12 +97,15 @@ Provides-Extra: ollama
 Provides-Extra: openai
 Provides-Extra: openai-agents
 Provides-Extra: pandas
+Provides-Extra: pgvector
 Provides-Extra: pydantic-ai
 Provides-Extra: spacy
 Provides-Extra: sql
+Provides-Extra: sqlite-vec
 Provides-Extra: timeseries
 Provides-Extra: transformers
 Provides-Extra: voyageai
+Provides-Extra: zvec
 Description-Content-Type: text/markdown
 
 <!-- markdownlint-disable MD033 MD041 -->
@@ -229,12 +238,14 @@ df = pl.read_csv("path/to/cases.csv")
 casebase = cbrkit.loaders.polars(df)
 ```
 
-For database access, CBRkit provides `sqlite` and `sqlalchemy` loaders (the latter requires the `sql` extra):
+For ad-hoc SQLite loading, CBRkit ships a stdlib-based loader:
 
 ```python
 casebase = cbrkit.loaders.sqlite("path/to/database.db", "SELECT * FROM cases")
 ```
 
+For richer relational backends (filters, upserts, vector/FTS search via pgvector on PostgreSQL or sqlite-vec on SQLite), see `cbrkit.indexable.sqlalchemy`, `cbrkit.indexable.pgvector`, and `cbrkit.indexable.sqlite_vec`.
+
 **Tip:** You can validate a loaded casebase against a Pydantic model using `cbrkit.loaders.validate()`:
 
 ```python
@@ -680,8 +691,7 @@ The result contains `similarities` with quality assessment scores for each case.
 ## Retain
 
 The retain phase decides whether and how to integrate new cases into the casebase.
-The `cbrkit.retain` module provides utility functions for this purpose.
-You build a retain pipeline by specifying an assessment function and a storage function:
+Build a retain pipeline from an assessment function and a storage function:
 
 ```python
 retainer = cbrkit.retain.build(
@@ -693,27 +703,9 @@ retainer = cbrkit.retain.build(
 )
 ```
 
-CBRkit provides several built-in storage functions:
-
-- `static`: Generates keys from a fixed reference casebase to avoid collisions.
-- `indexable`: Keeps an `IndexableFunc`'s index in sync with the casebase.
-
-You can filter retained cases based on their assessment scores using the `dropout` wrapper:
-
-```python
-retainer = cbrkit.retain.dropout(
-    retainer_func=cbrkit.retain.build(...),
-    min_similarity=0.5,
-)
-```
-
-The retainer can be applied to a revise result:
-
-```python
-result = cbrkit.retain.apply_result(revise_result, retainer)
-```
-
-The result contains `similarities` with fitness scores and `casebase` with the updated cases.
+The built-in storage functions are `static` (generates collision-free keys from a reference casebase) and `indexable` (keeps an `IndexableFunc`'s index in sync with the casebase).
+Wrap a retainer with `dropout` to filter by assessment score (e.g. `min_similarity=0.5`), then apply it to a revise result via `cbrkit.retain.apply_result(revise_result, retainer)`.
+The result exposes `similarities` (fitness scores) and `casebase` (updated cases).
 
 ## Full CBR Cycle
 
@@ -846,37 +838,74 @@ result = cbrkit.retrieval.apply_query(casebase, query, (retriever, reranker))
 
 ### Indexed Retrieval
 
-Some retrievers like `bm25`, `embed`, and `lancedb` support **indexed retrieval**, where the casebase is pre-indexed once and then queried without passing the full casebase each time.
-This is useful for large casebases or when using external search backends.
+Indexed retrieval pre-indexes the casebase once and then queries it without passing the full casebase each time, which helps for large casebases or external search backends.
+Index maintenance lives on whichever object owns the index.
 
-To use indexed retrieval, first create a retriever and call its `index()` method:
+The self-contained `bm25` and `embed` retrievers own their index, so you call `put_index()` on the retriever:
 
 ```python
 from frozendict import frozendict
 
-bm25_func = cbrkit.sim.embed.bm25(language="en")
-retriever = cbrkit.retrieval.bm25(conversion_func=bm25_func)
-retriever.create_index(frozendict(casebase))
+retriever = cbrkit.retrieval.bm25(conversion_func=cbrkit.sim.embed.bm25(language="en"))
+retriever.put_index(frozendict(casebase))
 ```
 
-Then pass an empty casebase (`{}`) to signal that the retriever should use its pre-indexed data:
+The storage-backed `lancedb`, `chromadb`, `zvec`, `pgvector`, and `sqlite_vec` retrievers are pure query paths over a separate `cbrkit.indexable` storage that owns the index, so you index on the storage and wrap it for querying:
 
 ```python
-result = cbrkit.retrieval.apply_query({}, query, retriever)
+storage = cbrkit.indexable.lancedb(uri="./cases", table_name="cases")
+storage.put_index(frozendict(casebase))
+retriever = cbrkit.retrieval.lancedb(storage=storage, search_type="dense")
 ```
 
-As a convenience, CBRkit provides `apply_query_indexed` and `apply_queries_indexed` which handle the empty casebase automatically:
+Query a pre-indexed retriever with `apply_query_indexed` / `apply_queries_indexed` (or pass an empty casebase `{}` to `apply_query`); querying an un-indexed retriever raises `ValueError`:
 
 ```python
 result = cbrkit.retrieval.apply_query_indexed(query, retriever)
-# or for multiple queries:
-result = cbrkit.retrieval.apply_queries_indexed(queries, retriever)
 ```
 
-If a retriever receives an empty casebase but has not been indexed yet, a `ValueError` is raised with a message to call `index()` first.
+The `System` class also defaults its casebase to `{}`, so a system of pre-indexed retrievers needs no casebase at query time.
+
+#### Typed Values and the Retain Caveat
+
+Each backend has one text-field knob — `value_column` (`value_field` for `zvec`/`chromadb`) — naming the embeddable text, and the value type `V` follows the schema source:
+
+- **Plain text** (`V = str`, the default) — the bare string is stored under the text knob and read back as a string.
+- **Typed model** (`V = YourModel`) — pass a `model`: a dataclass or Pydantic model for `lancedb`/`zvec`/`chromadb` (fields become columns), or a SQLAlchemy mapped class for `sqlalchemy`/`pgvector`/`sqlite_vec` (its `__table__` defines the schema). Reads reconstruct model instances.
+- **Mapping** (`V = Mapping[str, Any]`) — `sqlalchemy`/`pgvector`/`sqlite_vec` only, via a host-supplied `table` or `reflect=True`.
+
+```python
+# plain strings — cbrkit builds the table
+store = cbrkit.indexable.pgvector[str, str](
+    url=..., value_column="body", pgvector_dim=384, conversion_func=embed,
+)
+
+# typed rows — a SQLAlchemy mapped class defines the schema
+class Car(Base):
+    __tablename__ = "cars"
+    key: Mapped[str] = mapped_column(primary_key=True)
+    desc: Mapped[str] = mapped_column()
+    _pgvec: Mapped[Any] = mapped_column(cbrkit.indexable.PGVECTOR(384), nullable=False)
+
+store = cbrkit.indexable.pgvector[str, Car](url=..., model=Car, value_column="desc", ...)
+```
+
+Pass `vector_type="halfvec"` for half-precision storage (~2x smaller, negligible recall loss); for a typed model, declare the column with the re-exported `cbrkit.indexable.HALFVEC` instead of `PGVECTOR`.
+
+For a self-contained, file-based store, `sqlite_vec` offers the same dense/sparse/hybrid API on SQLite via the [`sqlite-vec`](https://github.com/asg017/sqlite-vec) extension (loaded automatically). Dense KNN uses a `vec0` virtual table (so the backend inherits future `vec0` capabilities such as approximate search, and supports quantized `vector_type="int8"` storage today), sparse search uses built-in FTS5, and `Filter` `WHERE` clauses compose by joining matches back to the main table:
+
+```python
+store = cbrkit.indexable.sqlite_vec[str, str](
+    url="sqlite+aiosqlite:///cases.db",
+    value_column="body", vector_dim=384, index_type="hybrid", conversion_func=embed,
+)
+store.put_index(frozendict(casebase))
+retriever = cbrkit.retrieval.indexable.sqlite_vec(storage=store, search_type="hybrid")
+```
 
-The `System` class also supports indexed retrieval by defaulting the casebase to an empty dict.
-This allows creating a system where all retrievers are pre-indexed and no casebase needs to be provided at query time.
+**Retain caveat:** the storage-backed retrievers search by the text column and always return `Casebase[K, str]`, projecting richer values down to their text.
+A retrieve → retain round-trip via `cbrkit.retrieval.indexable.*` + `cbrkit.retain.indexable` therefore lines up cleanly only when `V = str`.
+For model or mapping stores, either use the backend as a typed store (read `storage.index` as `Casebase[K, V]` and retrieve with a value-based retriever like `cbrkit.retrieval.build(...)`), or re-hydrate full rows by key from `storage.index` after text retrieval.
 
 ## Evaluation
 

{cbrkit-1.2.0 → cbrkit-1.4.0}/README.md

@@ -128,12 +128,14 @@ df = pl.read_csv("path/to/cases.csv")
 casebase = cbrkit.loaders.polars(df)
 ```
 
-For database access, CBRkit provides `sqlite` and `sqlalchemy` loaders (the latter requires the `sql` extra):
+For ad-hoc SQLite loading, CBRkit ships a stdlib-based loader:
 
 ```python
 casebase = cbrkit.loaders.sqlite("path/to/database.db", "SELECT * FROM cases")
 ```
 
+For richer relational backends (filters, upserts, vector/FTS search via pgvector on PostgreSQL or sqlite-vec on SQLite), see `cbrkit.indexable.sqlalchemy`, `cbrkit.indexable.pgvector`, and `cbrkit.indexable.sqlite_vec`.
+
 **Tip:** You can validate a loaded casebase against a Pydantic model using `cbrkit.loaders.validate()`:
 
 ```python
@@ -579,8 +581,7 @@ The result contains `similarities` with quality assessment scores for each case.
 ## Retain
 
 The retain phase decides whether and how to integrate new cases into the casebase.
-The `cbrkit.retain` module provides utility functions for this purpose.
-You build a retain pipeline by specifying an assessment function and a storage function:
+Build a retain pipeline from an assessment function and a storage function:
 
 ```python
 retainer = cbrkit.retain.build(
@@ -592,27 +593,9 @@ retainer = cbrkit.retain.build(
 )
 ```
 
-CBRkit provides several built-in storage functions:
-
-- `static`: Generates keys from a fixed reference casebase to avoid collisions.
-- `indexable`: Keeps an `IndexableFunc`'s index in sync with the casebase.
-
-You can filter retained cases based on their assessment scores using the `dropout` wrapper:
-
-```python
-retainer = cbrkit.retain.dropout(
-    retainer_func=cbrkit.retain.build(...),
-    min_similarity=0.5,
-)
-```
-
-The retainer can be applied to a revise result:
-
-```python
-result = cbrkit.retain.apply_result(revise_result, retainer)
-```
-
-The result contains `similarities` with fitness scores and `casebase` with the updated cases.
+The built-in storage functions are `static` (generates collision-free keys from a reference casebase) and `indexable` (keeps an `IndexableFunc`'s index in sync with the casebase).
+Wrap a retainer with `dropout` to filter by assessment score (e.g. `min_similarity=0.5`), then apply it to a revise result via `cbrkit.retain.apply_result(revise_result, retainer)`.
+The result exposes `similarities` (fitness scores) and `casebase` (updated cases).
 
 ## Full CBR Cycle
 
@@ -745,37 +728,74 @@ result = cbrkit.retrieval.apply_query(casebase, query, (retriever, reranker))
 
 ### Indexed Retrieval
 
-Some retrievers like `bm25`, `embed`, and `lancedb` support **indexed retrieval**, where the casebase is pre-indexed once and then queried without passing the full casebase each time.
-This is useful for large casebases or when using external search backends.
+Indexed retrieval pre-indexes the casebase once and then queries it without passing the full casebase each time, which helps for large casebases or external search backends.
+Index maintenance lives on whichever object owns the index.
 
-To use indexed retrieval, first create a retriever and call its `index()` method:
+The self-contained `bm25` and `embed` retrievers own their index, so you call `put_index()` on the retriever:
 
 ```python
 from frozendict import frozendict
 
-bm25_func = cbrkit.sim.embed.bm25(language="en")
-retriever = cbrkit.retrieval.bm25(conversion_func=bm25_func)
-retriever.create_index(frozendict(casebase))
+retriever = cbrkit.retrieval.bm25(conversion_func=cbrkit.sim.embed.bm25(language="en"))
+retriever.put_index(frozendict(casebase))
 ```
 
-Then pass an empty casebase (`{}`) to signal that the retriever should use its pre-indexed data:
+The storage-backed `lancedb`, `chromadb`, `zvec`, `pgvector`, and `sqlite_vec` retrievers are pure query paths over a separate `cbrkit.indexable` storage that owns the index, so you index on the storage and wrap it for querying:
 
 ```python
-result = cbrkit.retrieval.apply_query({}, query, retriever)
+storage = cbrkit.indexable.lancedb(uri="./cases", table_name="cases")
+storage.put_index(frozendict(casebase))
+retriever = cbrkit.retrieval.lancedb(storage=storage, search_type="dense")
 ```
 
-As a convenience, CBRkit provides `apply_query_indexed` and `apply_queries_indexed` which handle the empty casebase automatically:
+Query a pre-indexed retriever with `apply_query_indexed` / `apply_queries_indexed` (or pass an empty casebase `{}` to `apply_query`); querying an un-indexed retriever raises `ValueError`:
 
 ```python
 result = cbrkit.retrieval.apply_query_indexed(query, retriever)
-# or for multiple queries:
-result = cbrkit.retrieval.apply_queries_indexed(queries, retriever)
 ```
 
-If a retriever receives an empty casebase but has not been indexed yet, a `ValueError` is raised with a message to call `index()` first.
+The `System` class also defaults its casebase to `{}`, so a system of pre-indexed retrievers needs no casebase at query time.
+
+#### Typed Values and the Retain Caveat
+
+Each backend has one text-field knob — `value_column` (`value_field` for `zvec`/`chromadb`) — naming the embeddable text, and the value type `V` follows the schema source:
+
+- **Plain text** (`V = str`, the default) — the bare string is stored under the text knob and read back as a string.
+- **Typed model** (`V = YourModel`) — pass a `model`: a dataclass or Pydantic model for `lancedb`/`zvec`/`chromadb` (fields become columns), or a SQLAlchemy mapped class for `sqlalchemy`/`pgvector`/`sqlite_vec` (its `__table__` defines the schema). Reads reconstruct model instances.
+- **Mapping** (`V = Mapping[str, Any]`) — `sqlalchemy`/`pgvector`/`sqlite_vec` only, via a host-supplied `table` or `reflect=True`.
+
+```python
+# plain strings — cbrkit builds the table
+store = cbrkit.indexable.pgvector[str, str](
+    url=..., value_column="body", pgvector_dim=384, conversion_func=embed,
+)
+
+# typed rows — a SQLAlchemy mapped class defines the schema
+class Car(Base):
+    __tablename__ = "cars"
+    key: Mapped[str] = mapped_column(primary_key=True)
+    desc: Mapped[str] = mapped_column()
+    _pgvec: Mapped[Any] = mapped_column(cbrkit.indexable.PGVECTOR(384), nullable=False)
+
+store = cbrkit.indexable.pgvector[str, Car](url=..., model=Car, value_column="desc", ...)
+```
+
+Pass `vector_type="halfvec"` for half-precision storage (~2x smaller, negligible recall loss); for a typed model, declare the column with the re-exported `cbrkit.indexable.HALFVEC` instead of `PGVECTOR`.
+
+For a self-contained, file-based store, `sqlite_vec` offers the same dense/sparse/hybrid API on SQLite via the [`sqlite-vec`](https://github.com/asg017/sqlite-vec) extension (loaded automatically). Dense KNN uses a `vec0` virtual table (so the backend inherits future `vec0` capabilities such as approximate search, and supports quantized `vector_type="int8"` storage today), sparse search uses built-in FTS5, and `Filter` `WHERE` clauses compose by joining matches back to the main table:
+
+```python
+store = cbrkit.indexable.sqlite_vec[str, str](
+    url="sqlite+aiosqlite:///cases.db",
+    value_column="body", vector_dim=384, index_type="hybrid", conversion_func=embed,
+)
+store.put_index(frozendict(casebase))
+retriever = cbrkit.retrieval.indexable.sqlite_vec(storage=store, search_type="hybrid")
+```
 
-The `System` class also supports indexed retrieval by defaulting the casebase to an empty dict.
-This allows creating a system where all retrievers are pre-indexed and no casebase needs to be provided at query time.
+**Retain caveat:** the storage-backed retrievers search by the text column and always return `Casebase[K, str]`, projecting richer values down to their text.
+A retrieve → retain round-trip via `cbrkit.retrieval.indexable.*` + `cbrkit.retain.indexable` therefore lines up cleanly only when `V = str`.
+For model or mapping stores, either use the backend as a typed store (read `storage.index` as `Casebase[K, V]` and retrieve with a value-based retriever like `cbrkit.retrieval.build(...)`), or re-hydrate full rows by key from `storage.index` after text retrieval.
 
 ## Evaluation
 

{cbrkit-1.2.0 → cbrkit-1.4.0}/pyproject.toml

@@ -1,56 +1,56 @@
 [project]
 name = "cbrkit"
-version = "1.2.0"
+version = "1.4.0"
 description = "Customizable Case-Based Reasoning (CBR) toolkit for Python with a built-in API and CLI"
 authors = [{ name = "Mirko Lenz", email = "mirko@mirkolenz.com" }]
 readme = "README.md"
+license = "MIT"
 keywords = [
-    "cbr",
-    "case-based reasoning",
-    "api",
-    "similarity",
-    "nlp",
-    "retrieval",
-    "cli",
-    "tool",
-    "library",
+  "cbr",
+  "case-based reasoning",
+  "api",
+  "similarity",
+  "nlp",
+  "retrieval",
+  "cli",
+  "tool",
+  "library",
 ]
 classifiers = [
-    "Development Status :: 4 - Beta",
-    "Environment :: Console",
-    "Framework :: Pytest",
-    "Intended Audience :: Developers",
-    "Intended Audience :: Science/Research",
-    "License :: OSI Approved :: MIT License",
-    "Natural Language :: English",
-    "Operating System :: OS Independent",
-    "Programming Language :: Python :: 3.13",
-    "Programming Language :: Python :: 3.14",
-    "Programming Language :: Python :: 3",
-    "Topic :: Scientific/Engineering :: Artificial Intelligence",
-    "Topic :: Scientific/Engineering :: Information Analysis",
-    "Topic :: Software Development :: Libraries :: Python Modules",
-    "Topic :: Utilities",
-    "Typing :: Typed",
+  "Development Status :: 4 - Beta",
+  "Environment :: Console",
+  "Framework :: Pytest",
+  "Intended Audience :: Developers",
+  "Intended Audience :: Science/Research",
+  "Natural Language :: English",
+  "Operating System :: OS Independent",
+  "Programming Language :: Python :: 3.13",
+  "Programming Language :: Python :: 3.14",
+  "Programming Language :: Python :: 3",
+  "Topic :: Scientific/Engineering :: Artificial Intelligence",
+  "Topic :: Scientific/Engineering :: Information Analysis",
+  "Topic :: Software Development :: Libraries :: Python Modules",
+  "Topic :: Utilities",
+  "Typing :: Typed",
 ]
 requires-python = ">=3.13,<4"
 dependencies = [
-    "frozendict>=2,<3",
-    "numpy>=2,<3",
-    "orjson>=3,<4",
-    "polars>=1,<2",
-    "pydantic>=2,<3",
-    "pyyaml>=6,<7",
-    "rtoml>=0.12,<1",
-    "scipy>=1,<2",
-    "xmltodict>=1,<2",
+  "frozendict>=2,<3",
+  "numpy>=2,<3",
+  "orjson>=3,<4",
+  "polars>=1,<2",
+  "pydantic>=2,<3",
+  "pyyaml>=6,<7",
+  "rtoml>=0.12,<1",
+  "scipy>=1,<2",
+  "xmltodict>=1,<2",
 ]
 
 [project.optional-dependencies]
 # LLM providers
 anthropic = ["anthropic>=0.40,<1"]
-cohere = ["cohere>=5,<6"]
-google = ["google-genai>=1,<2"]
+cohere = ["cohere>=7,<8"]
+google = ["google-genai>=2,<3"]
 instructor = ["instructor>=1,<2"]
 ollama = ["ollama>=0.3,<1"]
 openai = ["openai>=1,<3", "tiktoken>=0.8,<1"]
@@ -76,26 +76,30 @@ graphviz = ["pygraphviz>=1,<2"]
 chromadb = ["chromadb>=1,<2"]
 lancedb = ["lancedb>=0.20,<1"]
 pandas = ["pandas>=2,<4"]
-sql = ["sqlalchemy>=2,<3"]
-# zvec = ["zvec>=0.2,<1"]
+pgvector = ["pgvector>=0.4,<1", "cbrkit[sql]"]
+sql = ["sqlalchemy[asyncio]>=2,<3"]
+sqlite-vec = ["sqlite-vec>=0.1,<1", "aiosqlite>=0.20,<1", "cbrkit[sql]"]
+zvec = ["zvec>=0.2,<1"]
 
 # Tools
-cli = ["rich>=13,<15", "typer>=0.9,<1"]
+cli = ["rich>=14,<16", "typer>=0.20,<1"]
 eval = ["ranx>=0.3,<1"]
 timeseries = ["minineedle>=3,<4"]
 
 # Entry points
 api = [
-    "cbrkit[cli]",
-    "fastapi>=0.100,<1",
-    "pydantic-settings>=2,<3",
-    "python-multipart>=0.0.15,<1",
-    "uvicorn[standard]>=0.30,<1",
-    "fastmcp>=3,<4",
+  "cbrkit[cli]",
+  "fastapi>=0.100,<1",
+  "pydantic-settings>=2,<3",
+  "python-multipart>=0.0.15,<1",
+  "uvicorn[standard]>=0.30,<1",
+  "fastmcp>=3,<4",
 ]
 
 # Bundle
-all = ["cbrkit[anthropic,api,bm25,chromadb,chunking,cohere,eval,google,graphs,graphviz,instructor,lancedb,levenshtein,nltk,ollama,openai,openai-agents,pandas,pydantic-ai,spacy,sql,timeseries,transformers,voyageai,zvec]"]
+all = [
+  "cbrkit[anthropic,api,bm25,chromadb,chunking,cohere,eval,google,graphs,graphviz,instructor,lancedb,levenshtein,nltk,ollama,openai,openai-agents,pandas,pgvector,pydantic-ai,spacy,sql,sqlite-vec,timeseries,transformers,voyageai,zvec]",
+]
 
 [project.urls]
 Repository = "https://github.com/wi2trier/cbrkit"
@@ -117,11 +121,23 @@ build-backend = "uv_build"
 
 [tool.pytest]
 testpaths = ["src", "tests"]
-addopts = ["--cov=src/cbrkit", "--cov-report=term-missing", "--doctest-modules", "--import-mode=importlib"]
+addopts = [
+  "--cov=src/cbrkit",
+  "--cov-report=term-missing",
+  "--doctest-modules",
+  "--import-mode=importlib",
+]
 doctest_optionflags = ["NORMALIZE_WHITESPACE", "IGNORE_EXCEPTION_DETAIL", "ELLIPSIS"]
 
 [tool.uv]
 default-groups = ["dev", "test", "docs"]
 
+[tool.uv.extra-build-dependencies]
+pygraphviz = ["setuptools"]
+cbor = ["setuptools"]
+warc3-wet-clueweb09 = ["setuptools"]
+zlib-state = ["setuptools"]
+pystemmer = ["setuptools", "cython"]
+
 [tool.ruff.lint.pydocstyle]
 convention = "google"

{cbrkit-1.2.0 → cbrkit-1.4.0}/src/cbrkit/api.py

@@ -189,7 +189,7 @@ def synthesize(
     )
 
 
-def openapi_generator():
+def openapi_generator() -> dict[str, Any]:
     """Generate and cache the OpenAPI schema for the CBRKit API."""
     if not app.openapi_schema:
         app.openapi_schema = get_openapi(
@@ -203,4 +203,4 @@ def openapi_generator():
     return app.openapi_schema
 
 
-app.openapi = openapi_generator  # type: ignore[assignment]
+app.openapi = openapi_generator  # type: ignore[assignment]  # ty: ignore[invalid-assignment]

{cbrkit-1.2.0 → cbrkit-1.4.0}/src/cbrkit/eval/common.py

@@ -7,7 +7,7 @@ from typing import Literal, cast
 from ..helpers import (
     get_logger,
     normalize_and_scale,
-    round,
+    round_int,
     sim_map2ranking,
     unpack_float,
     unpack_floats,
@@ -487,15 +487,18 @@ def generate_metrics(
         >>> generate_metrics(["precision", "recall"], ks=5)
         ['precision@5', 'recall@5']
     """
-    if not isinstance(ks, Iterable):
-        ks = [ks]
-
-    if not isinstance(relevance_levels, Iterable):
-        relevance_levels = [relevance_levels]
+    ks_list: list[int | None] = [ks] if ks is None or isinstance(ks, int) else list(ks)
+    relevance_levels_list: list[int | None] = (
+        [relevance_levels]
+        if relevance_levels is None or isinstance(relevance_levels, int)
+        else list(relevance_levels)
+    )
 
     return [
-        generate_metric(*args)
-        for args in itertools.product(metrics, ks, relevance_levels)
+        generate_metric(metric, k, relevance_level)
+        for metric, k, relevance_level in itertools.product(
+            metrics, ks_list, relevance_levels_list
+        )
     ]
 
 
@@ -528,7 +531,7 @@ def similarities_to_qrels[Q, C](
 
     return {
         query: {
-            case: round(
+            case: round_int(
                 normalize_and_scale(sim, min_sim, max_sim, min_qrel, max_qrel),
                 round_mode,
             )