01 · Roasts
Notebook monoculture
95% Jupyter Notebook means the GPU is going brrr, but the language portfolio is mostly one long cell execution.
CI wanted
Klaudia proves you can wire CircleCI; tunnel-engine, klaudia-core, and mcp-gsheets are still waiting for the invite.
Shipyard, not user base
146 public repos and 68 total stars: the release cadence is louder than the adoption evidence.
Tested islands
Tachikoma-Observatory's 101 tests and docx-compressor's end-to-end suite make the untested MCP repos look especially exposed.
Built using
Zoral
Shadows one worker for a week, then takes over their job with zero extra setup. Behaves exactly like the original.
zoral.ai
02 · Category breakdown
- Impact25% weight62C
- Consistency20% weight65C
- Quality20% weight69C
- Depth15% weight58D
- Breadth10% weight55D
- Community10% weight55D
03 · Stats
365-day commit heatmap
88 active days
Language distribution
- Jupyter Notebook95%
- Python4%
- TypeScript0%
- HTML0%
- Jinja0%
- Shell0%
- Other1%
04 · Numbers
Owned repos
non-fork
56
Commits
last 12 months
479
Followers
23
Joined GitHub
Dec 2022
05 · Top repos
Yudhyy /
Klaudia
Klaudia is a substantial self-hosted AI accounting system with FastAPI, LangGraph/MCP agents, PostgreSQL-backed ledgers, OCR ingestion, deterministic numeric verification, approval gates, and automated CI tests, but has only 7 stars and no license.
Yudhyy /
Tachikoma-Observatory
A documented Python benchmark dashboard with a 15-scenario deterministic tool-calling suite, lockstep multi-model execution, SQLite history, and 101-test coverage; adoption remains limited at 2 stars.
Yudhyy /
tunnel-engine
A substantial, documented Python LLM gateway with vLLM, LMCache, LiteLLM routing, guardrails, observability, orchestration, and broad unit coverage, but only 4 stars and no demonstrated external adoption.
Yudhyy /
docx-compressor
A focused, locally-run DOCX image compressor with a polished README, modular src layout, and meaningful end-to-end tests, but only 2 stars and a short one-week development burst limit demonstrated adoption and longevity.
Yudhyy /
klaudia-dev
A typed Expo/React Native ledger-chat app with authentication, SSE streaming, attachment handling, and approval workflows; it has focused service tests but no visible adoption, README, CI, or license.
Yudhyy /
mcp-gsheets
A documented, modular Python MCP server exposing 16 Google Sheets read, write, and sheet-management tools, but with no demonstrated adoption, tests, CI, or license.
Yudhyy /
klaudia-core
A structured Python/LangGraph supervisor with provider abstraction, MCP integration, and careful sheet-coordinate safeguards, but currently has 0 stars, minimal README documentation, no tests or CI, no license, and no static typing.
Yudhyy /
mcp-sqlite
Small, undocumented-in-practice MCP SQLite server exposing asynchronous CRUD tools over FastMCP; modular Python layout is offset by absent tests, CI, license, and meaningful setup documentation.
Yudhyy /
pytorch-protocol
A same-day PyTorch learning notebook centered on one marimo file, with clear setup notes but no tests, CI, license, or demonstrated external adoption.
Yudhyy /
Yudhyy
A lightly adopted GitHub profile repository with 1 star and a README describing LLM/agentic-AI interests, but no fetched source files or evidence of a packaged, tested product.
Yudhyy /
relicbench
Relicbench is a newly created, zero-star benchmarking concept with an evocative README but no sampled implementation, tests, CI, license, or typed-language evidence.
Yudhyy /
TensorTonic-Solutions
TensorTonic-Solutions is a newly created, one-commit personal solutions repository with a README describing synchronization from TensorTonic but no fetched source files or implemented solutions yet.
06 · Timeline
- Dec 2, 2022Joined GitHub
- Sep 28, 2023Created Yudhyy
- Jan 13, 2026Created mcp-gsheets
- Apr 16, 2026Created klaudia-core
- Apr 16, 2026Created mcp-sqlite
- Apr 16, 2026Created Klaudia — Self-hosted AI accountant. Reads receipts / invoices, reconciles ledgers, checks its own numbers, and asks before anything irreversible. Open source, agentic, runs on your infrastr
- Apr 19, 2026Created klaudia-dev
- Jun 1, 2026Created Tachikoma-Observatory — A lightweight evaluation framework for testing Small Language Models (SLMs) on precise tool-calling capabilities to ensure they are production-ready.
- Jun 12, 2026Created relicbench — The last sanctuary of machine intelligence.
- Jun 12, 2026Created tunnel-engine — A production-grade LLM infra engine that uses vLLM for fast inference, LMCache for smart context caching, and LiteLM to handle multi-model routing and load balancing.
- Jul 20, 2026Created docx-compressor — Your private DOCX compression tool
- Jul 23, 2026Created TensorTonic-Solutions — My solutions to TensorTonic problems
- Aug 25, 2026Created pytorch-protocol — pytorch stuff
- Sep 3, 2026Most recent push to Yudhyy
07 · Compare
08 · Rubric
How this score was produced
Overall = Σ (category × weight) + gentle top-end curve
Tier thresholds
▸ How the pipeline works
- 01Scrape.Pull every non-fork repo pushed in the last 90 days, plus your contribution calendar, followers, and language byte counts — straight from GitHub's REST & GraphQL APIs.
- 02Triage.A small model reads every repo's file tree + README and picks the 20 files per repo that actually reveal how you code.
- 03Grade each repo. All repos run in parallel through a fast scoring model that reads the picked files and rates each one independently on Impact, Quality, and Depth — with evidence citations.
- 04Aggregate. A larger reasoning model combines the per-repo scores with server-computed stats (heatmap, commit cadence, language entropy, follower count) to produce the 6-dimension profile score + roasts.
- 05Correct.Deterministic server-side checks enforce anchor-scale floors (e.g. a profile with 2,000+ public commits can't score 30 Consistency) and recompute the final verdict.
~90 seconds per profile, ~$0.25 in compute. Total of ~240 files read across your top-12 repos. One rating per GitHub account per day.
▸ Data sources & caveats
- Heatmap & commit totals: GitHub GraphQL
contributionsCollection— covers the last 365 days, includes private repos when the user has opted in (default). - Language %: byte totals across the top 30 owned non-fork repos.
- Curve: a small upward nudge centered on raw score ≈ 70, capping at 100. Prevents specialists from being unfairly penalised for narrow breadth.
- Anchor corrections: when server-measured signals (e.g. privateWorkLikely, multiRepoVolume, follower count) mandate a minimum category score, the aggregation step enforces it. These are signal-conditional, not identity-based floors.