references

references

Baseline references you already pointed to

Framing / impetus

• Nick Spisak tweet: https://x.com/NickSpisak_/status/2041012360668750229
• Nick Spisak follow-up: https://x.com/NickSpisak_/status/2040448463540830705

Core pattern

• Andrej Karpathy gist: https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f

Concrete implementation reference

• atomicmemory / llm-wiki-compiler: https://github.com/atomicmemory/llm-wiki-compiler

Strongest architectural references after deeper GitHub research

These are the most relevant public repos for traccia, even though none is an exact match for an evidence-backed personal skill graph.

Best conceptual seed

• atomicmemory / llm-wiki-compiler https://github.com/atomicmemory/llm-wiki-compiler
  • closest match for immutable raw input, incremental compilation, and Obsidian-friendly derived artifacts

Best heavy-duty donor

• swarmclawai / swarmvault https://github.com/swarmclawai/swarmvault
  • strongest donor for provenance tagging, approvals, graph exports, and local-first knowledge compilation

Best repo-structured agent workflow reference

• SamurAIGPT / llm-wiki-agent https://github.com/SamurAIGPT/llm-wiki-agent
  • strongest reference for schema-driven repo maintenance, append-only log patterns, and Obsidian-compatible outputs

Best polished product/UI reference

• nashsu / llm_wiki https://github.com/nashsu/llm_wiki
  • strongest reference for ingest queue UX, review flow, and local graph browsing

Best graph/report donors

• safishamsi / graphify https://github.com/safishamsi/graphify
  • strongest reference for graph packaging, cache/update behavior, and EXTRACTED / INFERRED / AMBIGUOUS edge semantics
• Lum1104 / Understand-Anything https://github.com/Lum1104/Understand-Anything
  • strongest reference for interactive graph browsing and explanation-oriented graph UI

Useful boundary and caution references

Personal knowledge graph / personal data ingestion systems

• hanig / engram https://github.com/hanig/engram
• MurtazaPlumber68 / GraphVault https://github.com/MurtazaPlumber68/GraphVault
• saxenauts / persona https://github.com/saxenauts/persona
  • useful as warnings about how quickly whole-person archive systems become much broader than a skill graph

Skill extraction / taxonomy helpers

• KonstantinosPetrakis / esco-skill-extractor https://github.com/KonstantinosPetrakis/esco-skill-extractor

Skill-tree UI references

• RoninATX / Human-Skill-Tree https://github.com/RoninATX/Human-Skill-Tree
• nshadov / personal-skill-tree https://github.com/nshadov/personal-skill-tree
  • useful for visual motif only, not as architectural bases

Negative references for traccia

These repos are relevant mostly because they show the wrong default framing for this project.

Resume / CV -> skill graph

• NickSaulnier / SkillGraph https://github.com/NickSaulnier/SkillGraph
• brianmarti1994 / AI-Skill-Graph https://github.com/brianmarti1994/AI-Skill-Graph
• Rishet11 / SkillGraph https://github.com/Rishet11/SkillGraph
• LohitDamarla / SkillGraph-AI https://github.com/LohitDamarla/SkillGraph-AI
  • mostly job-fit, resume, onboarding, or gap-analysis systems rather than reflective self-archiving

Why they matter

• Karpathy gist plus llm-wiki-compiler are the closest match for the raw -> compiled knowledge pattern.
• swarmvault is the richest donor for provenance, reviewability, and graph/report ideas.
• llm-wiki-agent and nashsu/llm_wiki are the best references for repo ergonomics and user-facing browsing.
• graphify and Understand-Anything are useful references for graph outputs and interaction patterns.
• resume-driven skill-graph repos are mostly anti-references because they bias the product toward hiring and gap analysis instead of reflective archival.

Current conclusion

The best path is to build traccia from scratch while borrowing patterns, not to fork an existing project.

Recommended stance:

• borrow architecture heavily from atomicmemory/llm-wiki-compiler
• borrow provenance and review patterns selectively from swarmvault
• borrow repo and export conventions from llm-wiki-agent
• borrow browsing and graph-output ideas from nashsu/llm_wiki, graphify, and Understand-Anything
• do not build directly on the job-search-oriented skill-graph repos