工欲善其事,必先利其器!精选 Agent 开发必备工具
-
Cursor ⭐⭐⭐⭐⭐ 最推荐
- 链接:https://cursor.sh/
- 用途:AI 辅助编程,写 Agent 代码必备
- 特点:Claude 集成,代码补全,AI 问答
-
Claude Code
- 链接:https://claude.ai/
- 用途:复杂代码逻辑设计,架构咨询
- 特点:长上下文,代码理解能力强
-
腾讯元宝 ⭐⭐⭐⭐⭐ 最推荐
- 用途:读论文、总结论文、理解算法原理
- 特点:中文友好,免费
-
豆包
- 用途:快速理解论文核心思想
- 特点:响应快,免费
-
论文翻译工具
- 链接:https://hjfy.top/
- 用途:整篇论文翻译,保留格式
-
DeepWiki ⭐⭐⭐⭐⭐
- 链接:https://opendeep.wiki/
- 用途:理解复杂开源项目(如 LangChain 源码)
- 特点:AI 辅助代码理解
-
ZRead.ai
- 链接:https://zread.ai/
- 用途:快速理解 GitHub 项目
- GPT-4 / GPT-o3-mini
- 用途:头脑风暴,技术方案设计
- 配合高质量 Prompt 使用
-
Excalidraw ⭐⭐⭐⭐⭐
- 链接:https://excalidraw.com/
- 用途:绘制系统架构图、Agent 工作流图
- 特点:免费,简洁,适合技术图
-
Draw.io
- 链接:https://draw.io/
- 用途:复杂系统架构图
-
Grok (X.AI)
- 用途:论文润色、降重
- 特点:生成质量高
-
Claude 4.5
- 用途:Paper Review,技术文档审查
- 特点:逻辑严谨
- GitHub:搜索
awesome-{topic}或{topic}-dataset - Hugging Face Datasets:https://huggingface.co/datasets
- ModelScope:https://modelscope.cn/datasets
- OpenLabData:开放科研数据
- OpenKG:中文知识图谱数据
- Hugging Face Papers(趋势榜):https://huggingface.co/papers/trending
- ArXiv:https://arxiv.org/
- OpenReview:https://openreview.net/
- GitHub搜索技巧:
Awesome + 关键词- 示例:
Awesome RAG,Awesome Agent,Awesome LLM
- 示例:
- 国外AI趋势:https://www.theaivalley.com/
- Bing搜索技巧:关键词 +
pdf找研究报告
- Transformers:https://github.com/huggingface/transformers
- vLLM:https://github.com/vllm-project/vllm
- TGI (Text Generation Inference):https://github.com/huggingface/text-generation-inference
- LMDeploy:https://github.com/InternLM/lmdeploy
- SGLang:https://github.com/sgl-project/sglang
- DeepSpeed:https://github.com/microsoft/DeepSpeed
- FairScale:https://github.com/facebookresearch/fairscale
- Megatron-LM:https://github.com/NVIDIA/Megatron-LM
- ColossalAI:https://github.com/hpcaitech/ColossalAI
- Llama-Factory ⭐⭐⭐⭐⭐:https://github.com/hiyouga/LLaMA-Factory
- Swift:https://github.com/modelscope/swift
- Axolotl:https://github.com/OpenAccess-AI-Collective/axolotl
- Unsloth:https://github.com/unslothai/unsloth
- PEFT:https://github.com/huggingface/peft
RAG 框架:
- LangChain ⭐:https://github.com/langchain-ai/langchain
- LlamaIndex ⭐:https://github.com/run-llama/llama_index
- Haystack:https://github.com/deepset-ai/haystack
- RAGFlow:https://github.com/infiniflow/ragflow
- Dify:https://github.com/langgenius/dify
向量数据库:
- Milvus ⭐:https://github.com/milvus-io/milvus
- Weaviate:https://github.com/weaviate/weaviate
- Pinecone:https://www.pinecone.io/
- Chroma:https://github.com/chroma-core/chroma
- Qdrant:https://github.com/qdrant/qdrant
- FAISS:https://github.com/facebookresearch/faiss
Embedding 模型:
- BGE ⭐:https://github.com/FlagOpen/FlagEmbedding
- E5:https://github.com/microsoft/unilm/tree/master/e5
- Sentence-Transformers:https://github.com/UKPLab/sentence-transformers
- Instructor:https://github.com/xlang-ai/instructor-embedding
文档解析:
- MinerU ⭐:https://github.com/opendatalab/MinerU
- Unstructured:https://github.com/Unstructured-IO/unstructured
- LlamaParse:https://github.com/run-llama/llama_parse
- Docling:https://github.com/DS4SD/docling
- PyPDF2:https://github.com/py-pdf/pypdf2
Agent 框架:
- AutoGPT:https://github.com/Significant-Gravitas/AutoGPT
- LangGraph ⭐:https://github.com/langchain-ai/langgraph
- CrewAI:https://github.com/joaomdmoura/crewAI
- MetaGPT:https://github.com/geekan/MetaGPT
- AutoGen ⭐:https://github.com/microsoft/autogen
- Swarm:https://github.com/openai/swarm
工具调用:
- ToolBench:https://github.com/OpenBMB/ToolBench
- Gorilla:https://github.com/ShishirPatil/gorilla
- ToolLLM:https://github.com/OpenBMB/ToolLLM
记忆模块:
- Mem0 ⭐:https://github.com/mem0ai/mem0
- MemGPT:https://github.com/cpacker/MemGPT
- LangChain Memory:内置在 LangChain 中
- Zep:https://github.com/getzep/zep
GUI Agent:
- AppAgent:https://github.com/mnotgod96/AppAgent
- SeeAct:https://github.com/OSU-NLP-Group/SeeAct
- WebShop:https://github.com/princeton-nlp/WebShop
- Mind2Web:https://github.com/OSU-NLP-Group/Mind2Web
视觉理解:
- CLIP:https://github.com/openai/CLIP
- BLIP:https://github.com/salesforce/BLIP
- LLaVA:https://github.com/haotian-liu/LLaVA
- Qwen-VL:https://github.com/QwenLM/Qwen-VL
- InternVL:https://github.com/OpenGVLab/InternVL
OCR 技术:
- PaddleOCR ⭐:https://github.com/PaddlePaddle/PaddleOCR
- EasyOCR:https://github.com/JaidedAI/EasyOCR
- TrOCR:https://github.com/microsoft/unilm/tree/master/trocr
- GOT-OCR:https://github.com/Ucas-HaoranWei/GOT-OCR2.0
- Rapid_OCR:https://github.com/RapidAI/RapidOCR
模型服务:
- FastAPI ⭐:https://github.com/tiangolo/fastapi
- Gradio:https://github.com/gradio-app/gradio
- Streamlit:https://github.com/streamlit/streamlit
- BentoML:https://github.com/bentoml/BentoML
监控工具:
- LangSmith ⭐:https://smith.langchain.com/
- Weights & Biases:https://wandb.ai/
- MLflow:https://github.com/mlflow/mlflow
- TensorBoard:https://github.com/tensorflow/tensorboard
-
阿东的大模型实验室:https://ccn7vpu5l5y8.feishu.cn/wiki/Nh8NwmGpsi0cAykX6OWcHmS3nyc
- Agent 学习资料汇总
- 实战项目案例
- 面试题库
-
WayToAGI 小白入门:https://waytoagi.feishu.cn/wiki/QPe5w5g7UisbEkkow8XcDmOpn8e
- 零基础AI入门
- 系统化学习路径
-
吴恩达 LLM 课程(中文版):https://github.com/datawhalechina/llm-cookbook
- 大模型基础入门
- Datawhale 出品
- 宝藏AI产品知识库:https://zw73xyquvv.feishu.cn/wiki/V7JxwkK1ti8hjwkqxwAcf0cEngR
- AI开源产品解决方案库:https://d.aigclink.ai/
- 腾讯技术工程
- 阿里云开发者
- 阿里技术
- 大淘宝技术
Step 1: 读论文理解原理
工具:腾讯元宝 / 豆包
目标:理解 ReAct、Reflexion 等架构
Step 2: 阅读开源项目源码
工具:DeepWiki / ZRead.ai
目标:理解 LangChain、AutoGen 的实现
Step 3: 动手实现项目
工具:Cursor + Claude Code
目标:完成 3 个简历级项目
Step 4: 绘制架构图
工具:Excalidraw / Draw.io
目标:能清晰讲解系统设计
Step 5: 准备面试
工具:AgentGuide 面试题库
目标:掌握高频面试题
数据准备:MinerU(文档解析)
向量化:BGE-Embedding
存储:Milvus
框架:LangChain
监控:LangSmith
部署:FastAPI
框架:AutoGen / CrewAI
编排:LangGraph
记忆:Mem0
工具:各类API
监控:LangSmith
部署:FastAPI + Redis
读论文:腾讯元宝
读源码:DeepWiki
写代码:Cursor
画图:Excalidraw
做笔记:Feishu / Notion
- 找数据集:
{topic} dataset - 找综述:
Awesome {topic} - 找实现:
{论文名} implementation - 找教程:
{topic} tutorial Chinese
- 找热门论文:https://huggingface.co/papers/trending
- 找数据集:https://huggingface.co/datasets
- 找模型:https://huggingface.co/models
- ✅ 做笔记:每学一个知识点,立即记录
- ✅ 写博客:学完就写,加深理解
- ✅ 写代码:理论学完立即实践
- ✅ 问AI:遇到问题先问元宝/ChatGPT
- 搜索能力:快速找到高质量资源
- 动手能力:快速实现和验证想法
- 思考能力:不盲从,有自己的判断
核心理念:选择大于努力!选对工具,效率翻倍!