Skip to content

Quickstart: 5-Minute Tour

Get up and running with TLDRGraph in your repository.

Step 1: Initialize the Graph with Your Agent

Agent-Driven Architecture

Do not run tldrgraph init manually in a terminal. TLDRGraph is designed to be agent-driven: your coding agent inspects the repository's symbols and designs architectural layers tailored specifically to your project.

Trigger the initialization command within your coding agent:

/tldrgraph-init

Open /skills and select tldrgraph-init, or run: 

$tldrgraph-init

Ask your agent: 

Please run the /tldrgraph-init workflow as specified in AGENTS.md.

What Happens Behind the Scenes:

  1. AST Extraction: Parses Python, TypeScript, and JavaScript source files into an AST symbol graph.
  2. Dynamic Layer Design: Your agent reads the extracted symbols and writes .tldrgraph/layers.config.yaml with customized layer boundaries.
  3. Approval & Enrichment: Asks once before spending LLM tokens to summarize key bottleneck symbols in 200-node chunks.
  4. Vector Indexing: Automatically builds local FastEmbed ONNX dense embeddings.

Step 2: Explore the Interactive Visualizer

Launch the self-contained visualizer:

tldrgraph ui --serve

This opens .tldrgraph/TLDRGRAPH_VISUALIZER.html in your default web browser: - Architecture Map: Zoom out to see multi-layer module clusters; zoom in to see classes, functions, and cross-layer connection lines. - Workflows Explorer: Click on the Workflows tab to view step-by-step decision journeys with inputs, outputs, and branch conditions. - Node Inspector: Click any node to isolate upstream callers, downstream callees, and read live file content.

Step 3: Query Execution Flows via CLI

Ask natural language queries about how features work:

tldrgraph query "pension calculation and approval flow"

TLDRGraph retrieves top execution flows spanning across UI, API, Service, and DB layers, formatting them as compact Markdown tables.

Step 4: Trace Exact Call Paths

Trace the shortest cross-layer dependency path between any two symbols:

tldrgraph trace "ApplicationsController" "PensionCalculator"

Output: 

Trace Path: ApplicationsController -> PensionService -> PensionCalculator
  Layer 1: ApplicationsController [API Layer]
    └── calls PensionService.calculate() (tldrgraph/service.py:42)
  Layer 2: PensionService [Business Logic]
    └── calls PensionCalculator.evaluate() (tldrgraph/calculator.py:18)

Step 5: Review Dead Code Candidates

Inspect unreferenced symbols that may be safe to refactor or remove:

tldrgraph dead-code

Surfaces orphaned functions, unused database models, and unreachable handlers with forward reachability confidence scores.