15 projects
trainlens
Open-source Jupyter AI training explainer.
lastlight
Low-power, offline retrieval for disaster and infrastructure-failure guidance.
matplotlib-dark
Automatic dark themes for Matplotlib, with one-line activation
vlm-prior-probe
Black-box behavioral probes for measuring visual-evidence reliance and learned-prior persistence in Vision-Language Models.
evidenceflow
Analyze how retrieved and tool-generated evidence propagates through AI-agent execution traces.
context-dedup
Detect and remove redundant context before sending it to an LLM.
text-to-music-prompt-structurer
Turn free-form musical ideas into structured prompts for text-to-music workflows.
vlm-occlusion
Find which image regions influence a VLM claim through grid-based occlusion testing.
parametricbench
Catch LLM and VLM regressions with lightweight, parameterized benchmarks for any Python callable.
visual-patch-audit
Compare and audit image patches for VLM and computer vision workflows.
embedding-drift-lite
Compare two embedding sets and detect drift between versions.
rag-chunk-audit
Find common quality and safety issues in RAG chunks before indexing.
metaclean-vlm
Remove metadata from image datasets before VLM training or ingestion to reduce hidden prompt injection risk.
promptshield-llm
Redact sensitive data from LLM prompts before sending them to a model.
openai-image-token-counter
Calculate token costs for images in OpenAI Vision API
