Acta

Acta is an Event Sourcing library for .NET 10. It provides durable, append-only event storage, projections, and multi-pod coordination backed by database guarantees — so multiple application instances can safely share one event store without a separate coordinator.

Status: early scaffold — Phase 0 (Bootstrap). The public API and implementation are being built out across the roadmap. This repository currently contains the solution skeleton, supply-chain hardening, and the CI security gates.

What it does

• Event store — an append-only log with two interchangeable backends: an in-memory backend for fast tests and a PostgreSQL backend for production.
• Aggregates & repositories — load/replay aggregates from their event streams, enforcing the Owner.Target ownership invariant.
• Projections — build read models both inline and via an asynchronous projection daemon.
• Snapshots & correlation — bounded replay via snapshots and end-to-end correlation of events.
• Multi-pod coordination — N application pods plus PostgreSQL, using the database's own guarantees for ordering and coordination; a hexagonal (ports & adapters) design keeps the core independent of the backend.

The feature set is layered in tiers: event model & serialization → event store → aggregates → projections → snapshots → upcasting, outbox, idempotency and observability → crypto-shredding and multi-tenancy.

Packages

Package	Purpose
Acta.Abstractions	Public ports and contracts (10 port groups)
Acta	Core implementation
Acta.Postgres	PostgreSQL backend adapter
Acta.Testing	Testing helpers for consumers

Tech stack

• Target framework: net10.0 · Language: C# 14 · SDK: 10.0.301
• Testing: xUnit v3, Testcontainers, Stryker.NET (≥ 80% mutation score on src/Acta), CsCheck, BenchmarkDotNet

Build

dotnet build Acta.slnx

CI & security gates

Every push and pull request runs the full gate set: build + tests, architecture tests, mutation testing (Stryker.NET), SAST (CodeQL + Roslyn security analyzers), dependency scanning (plus Dependabot), and secret scanning with push protection. Supply-chain integrity is enforced from the first commit via a pinned NuGet source and a committed packages.lock.json lock file (RestoreLockedMode in CI).

Mutation testing

dotnet tool restore

# What CI runs. On a PR the >=80% threshold applies to changed code only.
dotnet stryker
dotnet stryker --since:origin/main

# Local, with per-test coverage analysis: reports NoCoverage separately from
# Survived, so you can see which code no test touches at all.
dotnet stryker -f stryker-config.coverage.json

stryker-config.json sets "coverage-analysis": "off", which makes Stryker run the whole test suite against every mutant instead of trusting its per-test coverage map. This is the stricter setting, not the looser one: a mutant that no test kills is still reported as Survived and still counts against the 80% threshold. It is required because Stryker 4.15.0's per-test coverage capture returns a near-empty map under --since on the MTP runner (which xUnit v3 requires), marking mutants NoCoverage that the suite demonstrably kills — a combination that pins the diff-scoped score near 8% regardless of test quality. The suite runs in well under a second, so the cost of off is a few minutes.

Use stryker-config.coverage.json when you want the NoCoverage vs Survived distinction back — it is the only thing off gives up. Run it without --since; per-test coverage capture is reliable on a full run.

License

MIT © 2026 Wizard Software (Artur Sawicki)