AI Engineering Toolbox
A collection of libraries for all phases of AI Engineering.
Phase: Data
Data Preparation
| Category | Tool | Remarks |
|---|---|---|
| Synthetic Data | curator, datadreamer, distillabel, fabricator, promptwright | |
| auto-label | Generate annotated datasets | |
| Pretraining pipelines | datatrove, nemo-curator | |
| Data Exploration | lilac | |
| Deduplication | semhash | |
| Web Scraping | trafilatura, crawl4ai |
Phase: Evaluation
Evals
| Category | Tool | Remarks |
|---|---|---|
| Libraries | athina-evals, deepeval, geval, openevals, promptfoo, inspect-ai | |
| evaluate, torchmetrics, torcheval | Off-the-shelf metrics | |
| mir-eval | ||
| Agents | agentsevals | |
| Benchmarks | openai-evals, yet-another-applied-llm-benchmark, lm-evaluation-harness | |
| llmtest-needleinahaystack | Needle in a Hatstack | |
| agentbench, textarena | Agents | |
| Diversity | diversity | |
| Hallucination | lettucedetect, hallucination-leaderboard | |
| RAG | ragas, ragchecker | |
| auto-evaluator | Generate synthetic QA pairs from docs | |
| chunking-evaluation | Evaluate chunking strategies | |
| Text Generation | bleurt |
Phase: Modeling
Experiment Tracking
| Category | Tool | Remarks |
|---|---|---|
| Libraries | trulens, mlflow, wandb |
Prompt Engineering
| Category | Tool | Remarks |
|---|---|---|
| Automatic Prompt Engineering | dspy, textgrad, adalflow | |
| Function Calling | functionary | |
| Structured Output | guidance, instructor, jsonformer,lm-format-enforcer, outlines, xgrammar, lqml, fructose | |
| llm-scraper | webpage to json | |
| Memory | mem0, letta, memobase, memary, langmem, memoripy, | |
| Rate Limiting | backoff, tenacity, ratelimit | |
| limits | Rate-limit for own APIs | |
| Code Interpreter | gpt-code-interpreter, open-interpreter, codeinterpreter-api |
RAG
| Category | Tool | Remarks |
|---|---|---|
| Libraries | llama-index, verba, fastrag, haystack | |
| Chunking | wtsplit, semchunk, chonkie, langchain-text-splitters | |
| open-parse | Layout parsing visually | |
| sparseprimingrepresentations | SPR | |
| Reranking | rerankers, flaskrank | |
| Retrieval | pyserini | |
| bm25s | Sparse retrieval | |
| Embeddings | fastembed, sentence-transformers | |
| model2vec | Static Vectors | |
| Vector Index | annoy, diskann, faiss, chroma, qdrant, pinecone, weviate, milvus | |
| pgvector, sqlite-vec | SQL Extensions | |
| simsimd | Faster dot-product on CPUs | |
| Late Interaction | ragatouille, pylate | Train ColBERT models |
| Graph RAG | fast-graphrag, graphrag, nano-graphrag | |
| OCR | textract | |
| nougat | Academic Documents | |
| Document Understanding | donut | |
| table-transformer | Table Extraction | |
| marker | PDF to markdown | |
| RAG on internal docs | onyx, xyne |
Agents
| Category | Tool | Remarks |
|---|---|---|
| Libraries | autogen, crewai, langroid, openai-agents, pydantic-ai, marvin, metagpt, semantic-kernel | |
| langgraph | Graphs | |
| smolagents | Code-based agent | |
| Web Use | browser-use | |
| deer-flow | Deep Research | |
| magnitude | Automated Testing | |
| Computer Use | ui-tars-desktop |
Finetuning
| Category | Tool | Remarks |
|---|---|---|
| LLM Finetuning | axolotl, unsloth, torchtune, peft, litgpt, llama-factory | |
| onebitllms | 1.58-bit LLMs | |
| Model Merging | mergekit, mergoo | |
| Multi-modal VLMs | maestro, nanovlm, mlx-vlm | |
| smol-vision | Recipes | |
| llava | Visual Instruction Tuning | |
| RL | art, trl, verl, openrlhf, atropos, retrain | |
| verifiers | Verifiers | |
| Distributed Training | metagron-lm, deepspeed, yafsdp, nanotron, fairscale, colossalai | |
| hivemind, psyche | Decentralized Training | |
| Self-instruct | airoboros | |
| Tokenization | supertokenizer | Train multi-word BPE |
Multimodality
Audio
| Category | Tool | Remarks |
|---|---|---|
| Voice Activity Detection | silero-vad | |
| General models | seamless-communication | |
| Speech to Text | faster-whisper, whisperx | |
| whisper-streaming | Realtime |
Vision
| Category | Tool | Remarks |
|---|---|---|
| Facial recognition | deepface | |
| VLM | CogVLM |
Phase: Deployment
Inference
| Category | Tool | Remarks |
|---|---|---|
| Inference Servers | ray, vllm, powerinfer, text-generation-inference, sglang, tensorrt-llm, ctransalte2, mlc-llm, deepspeed-mii, openllm, exllamav2, fastgen | |
| Batch Processing | skypilot | |
| Multi-token prediction | medusa | |
| Multi-LoRA inference | s-lora, punica, lorax | |
| LLM Gateway | litellm, portkey | |
| LLM Routing | routellm, automix, openrouter, awesome-ai-model-routing | |
| Model Cascade | frugalgpt | |
| Semantic Cache | gptcache | |
| Quantization | llm-compressor, bitsandbytes, optimum | |
| Embeddings | text-embeddings-inference, infinity | |
| Kernels | liger-kernel | |
| attorch | Triton kernels | |
| Local Servers | lmstudio, ollama, text-generation-webui, koboldcpp, llama.cpp | |
| Frontend UI | gradio, streamlit | |
| copilot-kit | Chat UI components |
Monitoring
| Category | Tool | Remarks |
|---|---|---|
| Observability | openllmetry, phoenix, logfire | |
| Guardrails | giskard, langkit, garak, deepchecks, nemo-guardrails | |
| rebuff | Prompt Injection Detection | |
| uqlm | Uncertainty Quantification | |
| Drift Detection | ft-drift | Detect drift in OpenAI messages |
| AI Detection | binoculars |