Full-stack developer focused on AI security and Buddhist digital humanities — securing LLM applications and building open-source tools that make ancient texts accessible to modern researchers.
-
FoJin 佛津
— The world's encyclopedic Buddhist digital text platform. 450+ sources, 30 languages, full-text reading, AI Q&A, knowledge graph, parallel reader. FastAPI + React + Elasticsearch.
-
llm-seclint
— Static security linter for LLM-powered applications. 8 detection rules (prompt injection, SSTI, XXE, RCE, SQLi). The Bandit for the AI era.
-
Buddhist AI Translator
— AI translation for Buddhist texts across Sanskrit, Pali, Tibetan, and Classical Chinese.
Merged
| Project | Stars | PR | Description |
|---|---|---|---|
| Dify |  |
#33769 | fix: remove legacy z-index overrides on model config popup |
| Dify | #33767 | fix(tests): correct keyword arguments in tool provider test constructors | |
| LiteLLM |  |
#24070 | fix: thinking blocks dropped when thinking field is null |
| SurfSense |  |
#886 | fix: use asyncio.to_thread for embedding calls in search endpoints |
| gstack |  |
#128 | fix: eliminate duplicate command sets in chain, improve flush perf |
| buddhist-uni |  |
#636 | Added Node.js unit tests for core JavaScript utilities |
| buddhist-uni | #637 | Split search_index.js into pure JS logic and Liquid data template | |
| buddhist-uni | #634 | Add help text to Archive.org borrowable links |
In Review
| Project | Stars | PR | Description |
|---|---|---|---|
| Dify | |
#33986 | fix: constant-time API key comparison + prevent IDOR in DataSourceOauthBinding |
| vllm |  |
#37939 | fix(security): replace eval() with safe math in tool examples |
| crewAI |  |
#5005 | fix(security): prevent XXE attacks with defusedxml |
| MCP Servers |  |
#3663 | fix(sqlite): prevent SQL injection in describe_table |
| LiteLLM | |
#24458 | fix(security): use Jinja2 SandboxedEnvironment to prevent SSTI |
| LiteLLM | #24455 | fix(security): add AST validation for custom code exec() | |
| LiteLLM | #24346 | fix(security): fix prompt injection detection — async heuristics | |
| FastChat |  |
#3820 | fix(security): sanitize HTML to prevent XSS injection |
If you're interested in AI security, Buddhist studies, digital humanities, or NLP for historical texts — open an issue or start a discussion on any of my repos.