TLDRGraph builds a source-backed feature catalog and Workflow Explorer. It does not construct an architecture graph and does not launch an AI process itself.
Run tldrgraph init to create a catalog. Run tldrgraph refresh to update an
existing catalog after source changes. A missing, invalid, or stale catalog returns
needs_feature_workflows and prints the current source_hash. The active coding
agent identifies feature outcomes and writes the catalog index first. It then
delegates each indexed outcome to a separate source-reading subagent. Each
worker writes only its assigned workflow file; the active agent does not collect
worker objects or write workflows:
.tldrgraph/features.yaml
.tldrgraph/workflows/<feature_id>.yaml
Each final artifact must use the reported source hash. Run the command that reported the status again; TLDRGraph validates the direct artifacts against the current source inventory and generates the visualizer.
Every feature and every displayed workflow step must have at least one evidence record:
file: relative/path/to/source.py
symbol: verified_symbol
line: 12
code_start: 12
code_end: 28Paths must be repository-relative and present in the current source inventory. Line ranges must exist in the current file. Never invent evidence.
When one workflow step has mutually exclusive paths, modes, or choices, add an
options list to that step. Each option becomes its own flow-chart node and must
include its own title, text, and verified evidence. For example, a runtime
bootstrap step with Docker and Kubernetes paths should use two options, not one
step with both paths hidden as references.
- Areas have a
productortechnicalperspective and an order within it. - Features describe user, admin, developer, or operator outcomes—not code symbols.
- A
generatedworkflow proves the complete journey through its final response or visible result. - A
partialworkflow contains only proven steps and explains the absent portion inmissing_coverage. - A
pendingworkflow has no steps and explains what could not be established. - User-facing flows begin at the initiating action and include the client request, backend work, response, and UI update when those stages exist.
- Alternate paths are represented as step
optionsso the explorer can draw separate branch nodes that reconnect to the following step.
features.yaml uses schema tldrgraph/features@4, generator
feature-catalog-agent@4, the reported source_hash, areas, and lightweight
feature index metadata. Each worker-owned workflow uses schema
tldrgraph/feature-workflow@4, generator feature-workflow-subagent@4, the
reported source hash, and all status and evidence. Continue until tldrgraph init or tldrgraph refresh reports done, then run tldrgraph ui --serve.