Analytics & Monitoring Agent Skills for Zed

Reduce LLM size and accelerate inference using pruning techniques like Wanda and SparseGPT.

Curate LLM training data: dedupe, filter, PII redaction.

OpenAI Images API: batches, prompt sampler, gallery.

Autonomously optimize code for performance using CodSpeed benchmarks, flamegraph analysis, and iterative improvement.

Outlines: structured JSON/regex/Pydantic LLM generation.

Add and manage evaluation results in Hugging Face model cards.

Build and run evaluators for AI/LLM applications using Phoenix.

Clean training loops with built-in distributed support.

SAM: zero-shot image segmentation via points, boxes, masks.

Run the full Semantica semantic extraction pipeline on a file or selected text — NER, relations, events, coreference resolution, triplets…

Trace data lineage, source attribution, audit trails, and provenance assertions in Semantica graphs.

Visualize the Semantica knowledge graph — topology, centrality, communities, paths, embeddings, decision insights, and temporal evolution.

Local text-to-speech via sherpa-onnx (offline, no cloud)

Reference-free preference alignment, simpler than DPO.

RL post-training for LLMs with Megatron and SGLang.

Audio spectrograms/features (mel, chroma, MFCC) via CLI.

Text-to-image generation, inpainting, and img2img.

Audit PostHog survey SDK features and version requirements

TRL: SFT, DPO, GRPO, RLOO reward modeling for LLM RLHF.

Unsloth: 2-5x faster LoRA/QLoRA fine-tuning, less VRAM.

Transcribe and translate speech in 99 languages.