Crate ferrify_context 

Repository modeling and working-context selection.

agent-context is the read-only exploration layer in Ferrify. It inspects a repository, records structural facts such as workspace members and toolchain files, and produces a bounded working set that later stages can use without carrying the entire repo into memory.

The crate follows a structural-first read order. It looks at root manifests, toolchain configuration, CI files, repository policy, and only then expands into nearby code. That order matters because Ferrify treats current repository evidence as stronger than remembered conventions.

Examples

use agent_context::RepoModeler;
use std::path::Path;

let model = RepoModeler::scan(Path::new("."))?;
assert!(!model.crates.is_empty());

Structs

ApiBoundary

A discovered public API boundary.

ContextBudget

Limits for context selection before compaction.

ContextBuilder

Builds a compact working set from a RepoModel.

ContextSnapshot

The compacted context snapshot that survives between stages.

CrateFacts

Facts collected for one Cargo crate.

OpenQuestion

An unresolved question that should remain visible to later stages.

RepoFact

A repository fact preserved across compaction.

RepoModel

The repository model built before planning.

RepoModeler

Scans a repository root and builds a RepoModel.

ToolchainFacts

Toolchain files and CI entry points discovered at the root.

WorkingSet

The compact working set handed to planning.

Enums

AsyncModel

The async runtime posture inferred from dependencies.

CliStyle

The CLI implementation style inferred from dependencies.

ContextError

Errors produced while scanning repository context.

ErrorStyle

The error handling style inferred from dependencies.

LoggingStyle

The logging style inferred from dependencies.

TestStyle

The testing style inferred from dependencies.

WorkspaceKind

The broad workspace shape discovered during scanning.