Skip to main content

TypePact

Python resource contracts, inspected by Rust.

PyPI CLI on PyPI Python Quality Documentation License: MIT

TypePact lets you declare how resources close and how APIs intend to access them, alongside ordinary Python types. A shared Rust inspector resolves those declarations into contract facts with source locations, provenance, and diagnostics.

0.1.0 is the first contract-inspection release. It records declared intent and reports conflicting declarations or unsupported analysis boundaries. Checking resource use across control flow, including use-after-close, is planned for 0.2. APIs and the JSON schema remain provisional during 0.x.

Install

Add TypePact to your Python project:

uv add typepact==0.1.0

This installs two matching distributions:

Distribution Contents
typepact Pure Python declaration helpers, typing metadata, and py.typed
typepact-cli The native Rust typepact executable

typepact pins typepact-cli to the same version. Importing the Python helpers does not start the analyzer or download an executable. Prebuilt CLI wheels require no Rust compiler; building the CLI from source requires Rust 1.96 or newer and a linker.

For the command alone:

uv tool install typepact-cli==0.1.0
typepact --version

Quick start

Save this as connection.py:

from typepact import resource
from typepact.rust import Borrowed, Owned


@resource(closed_by="close")
class Connection:
    def close(self) -> None:
        pass


def read(conn: Borrowed[Connection]) -> str:
    return "ready"


def consume(conn: Owned[Connection]) -> None:
    conn.close()

Inspect its declarations:

uv run typepact inspect-contracts connection.py

TypePact emits JSON containing these contract views:

Declaration Base type Lifecycle Access intent
Connection connection.Connection Closed by close Unspecified
read.conn connection.Connection Closed by close Shared
consume.conn connection.Connection Closed by close Owned

The resource decorator preserves the class identity and records lifecycle metadata. The inspector reads the file as text without importing it, evaluating annotations, or invoking decorators. In 0.1, the method bodies above are not checked: declaring closed_by="close" does not establish that close releases anything.

See the complete paired example and its JSON snapshot.

Annotation forms

The access aliases use typing.Annotated, so ordinary Python type checkers retain the underlying type:

Alias from typepact.rust Metadata from typepact.contracts Intent
Borrowed[T] Annotated[T, SharedBorrow()] Shared access
Owned[T] Annotated[T, OwnedAccess()] Owned access
BorrowedMut[T] Annotated[T, ExclusiveBorrow()] Exclusive access

For example, Borrowed[Connection] and Annotated[Connection, SharedBorrow()] produce the same shared-access facts. The aliases are also exported from typepact.contracts.

You can attach lifecycle and access metadata to an alias of an existing class:

from typing import Annotated

from typepact.contracts import Resource, SharedBorrow


ConnectionView = Annotated[
    Connection,
    Resource(closed_by="close"),
    SharedBorrow(),
]

Here, Connection is the class from the quick start. Metadata on an alias describes that view; it does not change every use of the underlying class. TypePact combines compatible lifecycle and access declarations and reports conflicting values with their source locations. These annotations express intent; they do not enforce ownership or borrowing at runtime.

Inspecting a project

Choose a source target explicitly when needed. The default is Python 3.11:

uv run typepact inspect-contracts --python-version 3.13 connection.py

For imports between files, pass every module that should participate in resolution. For a project with a src directory:

uv run typepact inspect-contracts --root src src/models.py src/client.py

--root determines logical paths and package-qualified module names. It does not discover files or follow imports automatically. Without it, each input file's stem supplies its module name. Absolute from module import Class imports and direct imports of supported assignment aliases resolve across supplied files.

Output and exit codes

JSON is the default output; --json is accepted explicitly. Each result includes:

  • contracts: normalized base symbols, lifecycle and access facts, source spans, and declaration provenance.
  • diagnostics: errors and unsupported-boundary findings, including related locations where applicable.
  • coverage: complete-for-supported-subset or incomplete, with reasons and work counts.
  • schema_version: 0 and assurance: "contract-inspection-only".

Contract provenance is declared-unverified. Line and column numbers are one-based; columns count UTF-8 bytes.

Exit code Meaning
0 Inspection completed with no declaration errors or incomplete coverage
1 Declaration errors, syntax errors, or unsupported/unresolved constructs were reported
2 An invocation, file access, input limit, or fatal execution error prevented inspection

A successful inspection establishes the recorded declarations within the supported subset. It does not prove resource safety.

Scope of 0.1

The inspector supports module-level classes, literal TypePact resource decorators, simple assignment aliases, parameter and return annotations, and direct absolute imports among explicitly supplied modules. It resolves supported API imports by identity, respects local rebinding, and preserves the class and metadata captured by earlier aliases.

Relevant unsupported constructs produce incomplete coverage. Current boundaries include:

  • String annotations and from __future__ import annotations.
  • Relative imports, star imports, import re-export chains, and supplied-module cycles.
  • Class inheritance and metaclass behavior, other class decorators, and decorated functions.
  • Computed contract expressions, dynamic decorator arguments, and eager named expressions that could rebind module names.
  • Module-level control flow and unsupported nested or generic declarations.

The inspector has explicit source and work limits, including a 1,000,000-byte total source budget per request. See the inspector guide for the detailed limits and Rust session API.

Lifecycle flow checks, cleanup obligations, alias analysis through fields and containers, and enforcement of ownership or exclusive access belong to later phases. The release roadmap describes that progression.

Compatibility

Surface Support
Python helper runtime Python 3.11+; CI covers CPython 3.11–3.14
Analyzed Python source targets 3.11, 3.12, and 3.13
Python typing Inline types and py.typed; compatibility fixtures for mypy and Pyright
Rust source builds Rust 1.96+

The native CLI wheel matrix covers eight platform targets:

Platform x86-64 ARM64
Linux glibc glibc 2.17+ glibc 2.28+
Linux musl musl 1.2+ musl 1.2+
macOS 10.13+ 11.0+
Windows Supported Supported

The helper wheel is universal pure Python. CLI wheels contain a standalone executable and do not depend on a particular CPython ABI. Identifier validation follows the Unicode rules of the selected source target. See compatibility and distribution details for the complete policy.

Architecture

The Python helpers supply declarations; the Rust engine performs inspection. Pinned Ruff parser and AST components provide structured syntax and source ranges. Salsa tracks parsing and module semantics so reusable sessions can retain unaffected work across edits.

The Rust library exposes inspect for a single request and AnalysisSession for repeated updates, removal, reset, cancellation, and revision-tagged results. Both use the same bounded query pipeline as the CLI. See the library entry point and analyzer design.

Development

Use a source checkout with uv, Rust 1.96+, rustfmt, Clippy, and Make. Install the pinned dependency auditors, then run the full local gate:

uv sync --locked
cargo install cargo-audit --version 0.22.2 --locked
cargo install cargo-deny --version 0.20.2 --locked
uv run --locked make check

make check covers Rust formatting, compilation, Clippy, tests, MSRV validation, dependency audits, strict Python typing and linting, identifier-table verification, package builds and clean installation checks, and the inspector's execution guard. GitHub Actions runs the Python checks on 3.11–3.14 alongside the Rust and dependency checks.

Run the repository example from the checkout:

uv run --locked typepact inspect-contracts --root . examples/contracts.py

See CONTRIBUTING.md for focused check targets and contribution guidance.

Documentation

License

TypePact is distributed under the MIT license.

Metadata

Release files for typepact 0.1.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for typepact 0.1.0
File Size Uploaded Provenance Actions
typepact-0.1.0.tar.gz 7.0 kB Without provenance Details for typepact-0.1.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for typepact 0.1.0
File Interpreter ABI Platform Actions
typepact-0.1.0-py3-none-any.whl Python 3 none any Details for typepact-0.1.0-py3-none-any.whl

Total release size: 14.8 kB

Release files / typepact-0.1.0.tar.gz

Download URL typepact-0.1.0.tar.gz
Size 7.0 kB
Tags Source
SHA-256 checksum
How to use checksums
59502ecea2bfd149c713501bd6ec26720002e01cc280f0de354fcc80130981fd
BLAKE2b-256 checksum
How to use checksums
4daf327adc27fb10d1678d856c60a1a27dd0ed2ba90011f705cd377a7fdbf218
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.13.16

Release files / typepact-0.1.0-py3-none-any.whl

Download URL typepact-0.1.0-py3-none-any.whl
Size 7.8 kB
Tags Python 3
SHA-256 checksum
How to use checksums
f051b88cb1824b312e7f720ca973de369f1e1464c133d12500f866c8b0d8e160
BLAKE2b-256 checksum
How to use checksums
e5f7ce2668660b8c8d158a4fcd89a2ea6933670b4d3f1d23d208fd145d964c15
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.13.16

Security for typepact 0.1.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release

0.1.0 This release

2 release files No provenance

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page