---

---

What is GraphRuntime?

GraphRuntime extracts the architectural graph of any software project or package, enables LLMs to reason on that graph, and generates a runtime.json that connects, modifies, or fuses multiple systems — in any language.

pip install graphruntime

---

The 4 Universal Questions

Every file in every project answers 4 questions:

→ What enters this file?      (imports, dependencies)
→ What exits this file?       (classes, functions, exports)
→ What does it call?          (internal dependencies)
→ Who calls it?               (inferred by graph inversion)

These 4 questions work on Python, TypeScript, Rust, Go, Java, C++, Terraform, SQL, GraphQL — 42 languages supported.

---

Quick Start

# Extract graph from a local repo
graphruntime extract ./my-project

# Pull a pre-analyzed graph from the registry
graphruntime pull flask
graphruntime pull numpy

# Inspect an architecture
graphruntime inspect graph.json

# Merge two architectures with an objective
graphruntime merge graph_flask.json graph_numpy.json \
  --objective "REST API that processes dataframes"

# Execute a runtime
graphruntime run runtime.json

# Let the AI choose everything
graphruntime goal "I want an API that transcribes audio files to PDF"

---

CLI Commands

Command	Description
extract <source>	Extract graph from local path, GitHub, PyPI, npm, cargo
pull <package>	Pull pre-analyzed graph from registry
inspect <graph>	Human-readable architecture summary
diff <graph_a> <graph_b>	Architectural diff between two versions
modify <repo>	Modify existing repo guided by graph
create <repo>	Create missing files identified by graph
rewire <repo>	Invert or reroute data flow between modules
merge <a> <b>	Fuse two architectures into a runtime
run <runtime>	Execute a runtime.json
watch <repo>	Live graph rebuild on file changes
explain <graph>	LLM explains the architecture
validate <repo>	Validate architectural integrity
goal "<objective>"	AI selects libs, builds and runs everything
publish <graph>	Contribute a graph to the registry

---

Supported Sources

graphruntime extract ./local-path
graphruntime extract github:user/repo
graphruntime extract pip:flask
graphruntime extract npm:react
graphruntime extract cargo:tokio
graphruntime extract go:gin-gonic/gin

---

LLM Providers

graphruntime config set provider groq      # default
graphruntime config set provider openai
graphruntime config set provider anthropic
graphruntime config set provider ollama    # local

---

Registry

The GraphRuntime Registry contains pre-analyzed graph.json files for the most critical packages in the world. Browse at registry/ or pull directly:

graphruntime pull pandas
graphruntime pull kubernetes
graphruntime pull react

Contribute a new graph:

graphruntime publish my_graph.json --package mypackage --version 1.0.0

---

SKILL.md for AI Agents

Any LLM or AI agent can understand and pilot GraphRuntime by reading SKILL.md.

---

Contributors

---

License

MIT © GraphRuntime Contributors