dchisholm125

dchisholm125 / graph-oriented-generation

Public

Graph-Oriented Generation (GOG)

51
7
100% credibility
Found Mar 10, 2026 at 47 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

This repository implements and benchmarks Graph-Oriented Generation, a method using project dependency graphs to provide precise context for AI code generation tasks, outperforming traditional vector retrieval in token efficiency and accuracy.

How It Works

1
🔍 Discover GOG

You stumble upon this clever research tool that helps AI assistants understand code projects better by following real connections between files.

2
🏗️ Build Sample Project

You create a pretend website project packed with files and tricky paths to test how well AI can navigate it.

3
🗺️ Map the Connections

You prepare smart maps showing exactly how files link together in your sample project.

4
Pick Your AI Helper
☁️
Online AI

Get fast, powerful results from a cloud helper.

💻
Local AI

Run everything on your computer for full privacy.

5
🎯 Choose Challenge

Pick an easy, medium, or hard coding task like adding a login feature.

6
⚡ Run the Tests

Watch as the tool compares old fuzzy searches with precise path-following, showing huge savings in effort and perfect matches.

🎉 See the Wins

Celebrate how the new graph method delivers spot-on code changes with way less confusion and fewer words needed.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 47 to 51 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is graph-oriented-generation?

Graph-Oriented Generation (GOG) is a Python tool for goal-oriented graph generation that boosts LLM-assisted code tasks by traversing your project's import dependency graph instead of fuzzy vector search. It seeds relevant files from prompt similarity, pulls the minimal subgraph of connected code, and feeds a tiny, precise context to any LLM—cutting tokens 88-91% on benchmarks like adding auth features across stores and views. Users get CLI benchmarks for cloud APIs or local models like Ollama, plus a patch layer ensuring imports stay within bounds.

Why is it gaining traction?

GOG delivers deterministic context isolation that RAG can't match, proven by side-by-side timings on a 100-file Vue/TS maze with red herrings—vanilla graph traversal alone shrinks payloads massively, and the membrane fixes hallucinations zero-shot. Developers notice faster gen times on GPU, offline compatibility, and structural correctness scores without retries. The hook: run `benchmark_cloud_cli.py` or local version, see tables proving it works now.

Who should use this?

Fullstack devs wiring multi-file changes in TypeScript/Vue apps via LLMs, frustrated by RAG noise pulling irrelevant chunks. AI engineers prototyping neuro-symbolic agents for graph-oriented tasks like transmission expansion planning. Teams evaluating retrieval for codebases before scaling to production prompts.

Verdict

Grab it if you're in Python LLM tooling—reproducible benchmarks and easy extension make it practical despite 45 stars and 1.0% credibility score. Active prototype with strong docs; low maturity means test thoroughly, but contribute parsers to push it forward.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.