Categories

Autonomous and multi-agent systems: agent runtimes, tool use, planning, and orchestration of long-running tasks (LangGraph, CrewAI, AutoGen).

Image and video models and pipelines: detection, segmentation, OCR, and generation (Ultralytics, diffusion models, OCR).

Ingestion, chunking, parsing, scraping, and retrieval pipelines that feed context into LLMs (Unstructured, Firecrawl, document loaders).

SDKs, gateways, CLIs, and glue that connect AI systems together — the plumbing of an AI stack (LiteLLM, MCP servers, SDKs).

Tracing, evaluation, monitoring, and observability for LLM and agent systems — measuring quality, cost, and latency (Langfuse, Phoenix, OpenLIT).

Model inference, serving, and local runtimes — deploying and running models efficiently (Ollama, vLLM, llama.cpp, TGI).

Frameworks and orchestration libraries for building LLM applications — prompting, chaining, RAG, and agent scaffolding (LangChain, LlamaIndex, Haystack).

Training, fine-tuning, and post-training toolkits — from full pretraining to LoRA and RLHF (Unsloth, TRL, Axolotl).

Speech-to-text, text-to-speech, and audio models and pipelines (Whisper, TTS engines, ASR).

Vector stores and embedding databases for similarity search and retrieval — the memory/retrieval backend of a RAG stack (Qdrant, Chroma, pgvector, Milvus).