Download the PHP package dndark/laravel-logic-map without Composer

Laravel Logic Map 2.1

Laravel Logic Map builds an evidence-backed semantic graph of a Laravel application. It answers workflow and change-impact questions that a basic call graph cannot: which route reaches a service, where branches and transactions occur, which jobs/events/listeners continue the flow, which tables or external systems are affected, and which modules/tests should be reviewed after a change.

What 2.1 answers

• If function A changes, which callers, workflows, modules, shared resources, external contracts, and tests are affected?
• What is the full flow from a route/command/job through validation, authorization, business logic, branches, transactions, async boundaries, persistence, cache, notifications, and external calls?
• Which relationships are proven by AST/Laravel boot facts, which are probable/possible, and which were observed at runtime?
• Where did analysis stop because a target was dynamic or ambiguous?

The index is deterministic and stored in lm_* tables on the application's configured database connection. Runtime collection is optional and never replaces static evidence.

Requirements

• PHP 8.2+
• Laravel 10, 11, or 12
• ext-pdo and the PDO driver used by the application's database connection
• Composer

Installation

logic-map-assets is the only public asset tag. The package does not register the broad laravel-assets alias, so Composer discovery never publishes files by itself.

The Composer package name remains lowercase (dndark/laravel-logic-map), as required by Composer. PHP namespaces use the case-sensitive PSR-4 prefix DNDark\LogicMap.

Indexing

Configuration is flat under logic-map.*; there is no logic-map.v2.* wrapper. The defaults scan app, routes, database, config, and tests, then store the active snapshot in lm_* tables on the application's database connection (created by the package migrations — run php artisan migrate; override the connection with logic-map.storage.connection / LOGIC_MAP_DB_CONNECTION). The tests root is required for impact reports to select related tests; remove it only when that trade-off is intentional.

Open the protected viewer at /logic-map after publishing assets and creating an index.

Automatic domain grouping

The index groups every eligible route, class, method, async symbol, persistence resource, process, and test into exactly one primary module. Results are classified as:

• Business — a coherent domain inferred from routes, symbols, persistence, and graph relations;
• Shared — infrastructure reached by independent business entrypoints;
• Technical — namespace/directory/fallback grouping without enough business evidence.

Configured explicit mappings and app/Modules / app/Domain directory roots remain authoritative. Semantic assignments are probable, deterministic, and explainable through bounded source/evidence IDs on the member_of_module relation. Query endpoints only read the persisted result; they never infer modules during a request.

After upgrading to 2.1, run php artisan logic-map:index --force. To temporarily restore legacy namespace/directory/fallback behavior, set modules.semantic.enabled to false and force a new index.

Workflow examples

Use a canonical node ID or an exact qualified name:

Workflow output includes ordered steps, decisions, terminal paths, transaction segments, sync/async boundaries, effects, gaps, truncation metadata, and evidence IDs.

Impact examples

Analyze one symbol or a Git range:

Impact is reason-grouped rather than a flat dependency list. Each affected symbol contains its ordered node/edge chain and evidence IDs. Categories include hard dependency, workflow, async, shared state, external contract, module, uncertainty, and test scope.

AI documentation export

Export the active V2 snapshot as module and workflow dossiers:

The command writes overview.md, modules/*.md, and workflows/*.md. Module dossiers use the module workflow projection rather than treating every module member as a changed symbol. Workflow dossiers embed Mermaid diagrams together with ordered steps, effects, gaps, and evidence.

Export a machine-readable, impact-weighted bundle for AI agents:

The command writes graph.json (full node/edge/module inventory), impact/<symbol-slug>.json (per-symbol affected-symbol list with an explainable score/band/factors weight), modules/<slug>.json (membership and entrypoints), llms.txt (agent preamble and weight legend), and index.md. The bundle is a pure function of the active snapshot — byte-identical across repeated exports — so it is safe to commit or serve as a build artifact.

HTTP API

Canonical IDs in URL path segments use base64url encoding. Complete envelopes and examples are published on the developer docs site.

Evidence and certainty

Every graph relation references evidence records with an origin (static_ast, laravel_boot, runtime, or git_diff), detector, certainty (certain, probable, or possible), source location when available, and bounded attributes. Missing evidence remains unknown; it is never reported as proof that code did not execute.

HTTP viewer protection

The viewer/API are enabled only in local and testing by default and run through the configured middleware:

If production access is required, explicitly add the environment and an authentication/authorization middleware owned by the consumer application.

Runtime evidence (opt-in)

Runtime collection is disabled by default. Enable it only in an environment where bounded observations are appropriate:

Only observations with stable source and target node IDs, a compatible relation kind, and the same snapshot ID can overlay static queries. Coverage uses only these phrases: Observed in N selected runtime sessions, Not observed in selected sessions, and No runtime data available.

Known static-analysis limits

• Highly dynamic container resolution, reflection, runtime-generated routes, dynamic method names, and opaque package internals may remain unresolved.
• Runtime proximity alone is not treated as causality.
• Static non-observation does not prove a path is impossible; runtime non-observation does not prove it never executes.
• Bounded query limits can truncate a graph; always inspect meta/truncation and frontier data.

Commands

V1 cache snapshots, routes, reports, and UI are intentionally removed. The export-docs name is reused by a V2-native implementation; no V1 graph/report compatibility layer is loaded. See UPGRADING.md.

License

MIT. See LICENSE.