weavatrix-rust

Turn a repository into deterministic evidence your Rust code can trust.

weavatrix-rust maps a codebase into a typed, revision-bound evidence graph with exact source provenance. It answers impact, architecture, API, health, history, search, semantic, and temporal-memory questions without executing the repository it analyzes.

Use it to:

• embed repository analysis in a Rust application;
• serialize a Snapshot for CI, indexing, or review;
• run 43 bounded read-only operations in the default full build;
• enforce the current v1 architecture contract foundation;
• power the separate weavatrix MCP product.

This crate is an engine, not an MCP server. Protocol transport, npm packaging, profiles, and filesystem watching live in weavatrix.

Architecture Firewall

Architecture Firewall evaluates .weavatrix/architecture.json against the repository evidence graph. It supports direct and transitive component forbids, direct dependency allow-lists, required direct or transitive dependencies, unresolved-import policy, relation and coupling filters, runtime-cycle and source-size budgets, stable fingerprints, baselines, exceptions, capability verification, and change preflight.

Transitive violations include the deterministic shortest file path that crossed the declared boundary. Unknown rule actions, reachability modes, and relation kinds are rejected rather than silently skipped.

See Architecture Firewall for the contract, rule semantics, budgets, ratchet behavior, and operation reference.

Quick start

Use the default native engine:

[dependencies]
weavatrix-rust = "2.6.0"

use std::path::Path;
use weavatrix_rust::{Analyzer, AnalyzerConfig};

let snapshot = Analyzer::new(AnalyzerConfig::default())
    .analyze(Path::new("."))?;

println!("{} nodes, {} edges", snapshot.nodes.len(), snapshot.edges.len());
# Ok::<(), weavatrix_rust::Error>(())

Call the bounded operation catalog used by the CLI and MCP adapter:

use weavatrix_rust::{Weavatrix, operations};

let mut engine = Weavatrix::open(".")?;
let report = operations::call(
    &mut engine,
    "verify_architecture",
    blazingly_json::json!({}),
)?;

println!("{}", blazingly_json::to_string_pretty(&report)?);
# Ok::<(), Box<dyn std::error::Error>>(())

Minimal builds keep the analyzer, lossless parser, graph, snapshot model, and standalone CLI:

[dependencies]
weavatrix-rust = { version = "2.6.0", default-features = false }

MCP product

The canonical MCP distribution wraps this engine with stdio, profiles, incremental refresh, and native filesystem watching:

npx -y weavatrix mcp . --profile=code

Codex configuration:

[mcp_servers.weavatrix]
command = "npx"
args = ["-y", "weavatrix", "mcp", ".", "--profile=code"]

The adapter delegates its catalog and operations to this engine. The crate itself remains protocol-independent and does not open stdio or start a watcher.

Architecture

repository path or SourceInput[]
             |
             v
 language and contract adapters
             |
             v
      analysis pipeline
             |
             v
 evidence model and Snapshot
             |
             v
      repository engine
             |
             v
 read-only operations --> Rust facade / CLI / adapters

Focused first-party crates provide the reusable foundations:

Feature selection

The default is full + lang-rust. Disabled capabilities disappear from the operation catalog instead of being advertised as unavailable stubs.

Evidence and supported surfaces

Relationships can carry extractor identity, evidence class, confidence, source file and exact span, and extractor detail. Static reachability is never relabeled as measured coverage, and missing artifacts stay explicit.

The engine extracts evidence from Rust, JavaScript, TypeScript, Python, Go, Java, C#, C/C++, Bash, SQL, Solidity, Swift, HTML/CSS, Terraform, XML, Markdown-family sources, HTTP/GraphQL/gRPC APIs, common messaging systems, JSON/JSONC, YAML, Kubernetes, manifests, lockfiles, architecture contracts, and coverage artifacts.

See the evidence model and language support for exact interpretation limits.

Operations

The default full build exposes 43 operations:

The complete schemas live in the operation reference.

Standalone CLI

cargo install weavatrix-rust
weavatrix-rust analyze . --pretty
weavatrix-rust list-tools
weavatrix-rust tool verify_architecture .

Product boundary

This repository owns analysis, evidence, repository state, and read-only operations. MCP transport and npm packaging belong to weavatrix; source editing belongs to weavatrix-refactor, and licensed network workflows belong to weavatrix-online.

Safety boundary

• #![forbid(unsafe_code)] in the engine;
• no network implementation or application-source writes;
• no execution of analyzed repository code;
• no spawning Git, ripgrep, Node, Python, or language servers;
• canonical-path containment for repository reads;
• deterministic pagination and bounded results;
• explicit limitations instead of fabricated certainty.

Development gates

cargo fmt --all -- --check
cargo clippy --locked --all-targets --all-features -- -D warnings
cargo clippy --locked --all-targets --no-default-features -- -D warnings
cargo test --locked --no-default-features
cargo test --locked --test architecture_self
RUSTDOCFLAGS="-D warnings" cargo doc --locked --no-deps --all-features

Documentation

• Getting started
• Architecture Firewall
• Operation reference
• Evidence model
• Languages and repository surfaces
• Engine, CLI, and product boundary
• Architecture
• Dependencies and feature boundaries
• Benchmark methodology and evidence

License

MIT. Third-party crates retain their own licenses.