Skip to content

Latest commit

Β 

History

History
51 lines (42 loc) Β· 4.47 KB

File metadata and controls

51 lines (42 loc) Β· 4.47 KB

Repo hub β€” generation recipes (RECIPE layer)

Generated from data/stack.yml + data/frontier.yml by scripts/repos.py. See the design in HUB_ARCHITECTURE.md.

22 cited repos (20 on GitHub). This is the recipe, not the artifacts: run a repo's commands on demand and cache the result by the repo's commit SHA β€” deep knowledge graphs and skills are made just-in-time, never committed stale. You never pay for a repo you don't use.

The recipe (per repo)

# 1. knowledge graph
understand-anything <clone-dir>     # GitHub codebases (architecture KG)
graphify <url>                      # model cards / docs / papers (clustered KG)
# 2. agentic tooling (skill / plugin / workflow)
anyagent reverse <url> --markdown  # β†’ blueprint, then:
anyagent build/refine "<blueprint>" --target-score 80   # β†’ a gated skill/workflow
# 3. serve to any client
anyagent mcp                       # expose the generated tools over MCP

Quality gate: stamp every generated artifact with the source repo + commit SHA (reproducible + staleness-detectable); a skill ships only after refine clears its score.

The repos

Repo Kind Host KG engine Tooling
BixBench benchmark github understand-anything anyagent reverse Future-House/BixBench
LAB-Bench benchmark github understand-anything anyagent reverse Future-House/LAB-Bench
Open Problems in Single-Cell benchmark github understand-anything anyagent reverse openproblems-bio/openproblems
DunedinPACE clock github understand-anything anyagent reverse danbelsky/DunedinPACE
scAge clock github understand-anything anyagent reverse alex-trapp/scAge
Marinka Zitnik frontier-work github understand-anything anyagent reverse mims-harvard/ATHENA
Morgan Levine frontier-work github understand-anything anyagent reverse HigginsChenLab/methylCIPHER
cognee graph github understand-anything anyagent reverse topoteretes/cognee
Graphiti graph github understand-anything anyagent reverse getzep/graphiti
PyTorch Geometric Temporal graph github understand-anything anyagent reverse benedekrozemberczki/pytorch_geometric_temporal
TGB (Temporal Graph Benchmark) graph github understand-anything anyagent reverse shenyangHuang/TGB
AltumAge model github understand-anything anyagent reverse rsinghlab/AltumAge
ESM-2 model github understand-anything anyagent reverse facebookresearch/esm
Geneformer model huggingface graphify anyagent reverse <url>
scGPT model github understand-anything anyagent reverse bowang-lab/scGPT
TxGNN model github understand-anything anyagent reverse mims-harvard/TxGNN
AlphaFold tool github understand-anything anyagent reverse google-deepmind/alphafold
Biolearn tool web graphify anyagent reverse <url>
CZ CELLxGENE Census tool github understand-anything anyagent reverse chanzuckerberg/cellxgene-census
pyaging tool github understand-anything anyagent reverse rsinghlab/pyaging
scanpy tool github understand-anything anyagent reverse scverse/scanpy
ToolUniverse tool github understand-anything anyagent reverse mims-harvard/ToolUniverse

KG engine split: understand-anything for the 20 GitHub codebases, graphify for model cards / docs. PR a repo into data/stack.yml and it joins the hub automatically.