[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-explosion--spacy-llm":3,"tool-explosion--spacy-llm":64},[4,17,27,35,44,52],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":16},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,3,"2026-04-05T11:01:52",[13,14,15],"开发框架","图像","Agent","ready",{"id":18,"name":19,"github_repo":20,"description_zh":21,"stars":22,"difficulty_score":23,"last_commit_at":24,"category_tags":25,"status":16},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",140436,2,"2026-04-05T23:32:43",[13,15,26],"语言模型",{"id":28,"name":29,"github_repo":30,"description_zh":31,"stars":32,"difficulty_score":23,"last_commit_at":33,"category_tags":34,"status":16},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",107662,"2026-04-03T11:11:01",[13,14,15],{"id":36,"name":37,"github_repo":38,"description_zh":39,"stars":40,"difficulty_score":10,"last_commit_at":41,"category_tags":42,"status":16},4292,"Deep-Live-Cam","hacksider\u002FDeep-Live-Cam","Deep-Live-Cam 是一款专注于实时换脸与视频生成的开源工具，用户仅需一张静态照片，即可通过“一键操作”实现摄像头画面的即时变脸或制作深度伪造视频。它有效解决了传统换脸技术流程繁琐、对硬件配置要求极高以及难以实时预览的痛点，让高质量的数字内容创作变得触手可及。\n\n这款工具不仅适合开发者和技术研究人员探索算法边界，更因其极简的操作逻辑（仅需三步：选脸、选摄像头、启动），广泛适用于普通用户、内容创作者、设计师及直播主播。无论是为了动画角色定制、服装展示模特替换，还是制作趣味短视频和直播互动，Deep-Live-Cam 都能提供流畅的支持。\n\n其核心技术亮点在于强大的实时处理能力，支持口型遮罩（Mouth Mask）以保留使用者原始的嘴部动作，确保表情自然精准；同时具备“人脸映射”功能，可同时对画面中的多个主体应用不同面孔。此外，项目内置了严格的内容安全过滤机制，自动拦截涉及裸露、暴力等不当素材，并倡导用户在获得授权及明确标注的前提下合规使用，体现了技术发展与伦理责任的平衡。",88924,"2026-04-06T03:28:53",[13,14,15,43],"视频",{"id":45,"name":46,"github_repo":47,"description_zh":48,"stars":49,"difficulty_score":23,"last_commit_at":50,"category_tags":51,"status":16},3704,"NextChat","ChatGPTNextWeb\u002FNextChat","NextChat 是一款轻量且极速的 AI 助手，旨在为用户提供流畅、跨平台的大模型交互体验。它完美解决了用户在多设备间切换时难以保持对话连续性，以及面对众多 AI 模型不知如何统一管理的痛点。无论是日常办公、学习辅助还是创意激发，NextChat 都能让用户随时随地通过网页、iOS、Android、Windows、MacOS 或 Linux 端无缝接入智能服务。\n\n这款工具非常适合普通用户、学生、职场人士以及需要私有化部署的企业团队使用。对于开发者而言，它也提供了便捷的自托管方案，支持一键部署到 Vercel 或 Zeabur 等平台。\n\nNextChat 的核心亮点在于其广泛的模型兼容性，原生支持 Claude、DeepSeek、GPT-4 及 Gemini Pro 等主流大模型，让用户在一个界面即可自由切换不同 AI 能力。此外，它还率先支持 MCP（Model Context Protocol）协议，增强了上下文处理能力。针对企业用户，NextChat 提供专业版解决方案，具备品牌定制、细粒度权限控制、内部知识库整合及安全审计等功能，满足公司对数据隐私和个性化管理的高标准要求。",87618,"2026-04-05T07:20:52",[13,26],{"id":53,"name":54,"github_repo":55,"description_zh":56,"stars":57,"difficulty_score":23,"last_commit_at":58,"category_tags":59,"status":16},2268,"ML-For-Beginners","microsoft\u002FML-For-Beginners","ML-For-Beginners 是由微软推出的一套系统化机器学习入门课程，旨在帮助零基础用户轻松掌握经典机器学习知识。这套课程将学习路径规划为 12 周，包含 26 节精炼课程和 52 道配套测验，内容涵盖从基础概念到实际应用的完整流程，有效解决了初学者面对庞大知识体系时无从下手、缺乏结构化指导的痛点。\n\n无论是希望转型的开发者、需要补充算法背景的研究人员，还是对人工智能充满好奇的普通爱好者，都能从中受益。课程不仅提供了清晰的理论讲解，还强调动手实践，让用户在循序渐进中建立扎实的技能基础。其独特的亮点在于强大的多语言支持，通过自动化机制提供了包括简体中文在内的 50 多种语言版本，极大地降低了全球不同背景用户的学习门槛。此外，项目采用开源协作模式，社区活跃且内容持续更新，确保学习者能获取前沿且准确的技术资讯。如果你正寻找一条清晰、友好且专业的机器学习入门之路，ML-For-Beginners 将是理想的起点。",84991,"2026-04-05T10:45:23",[14,60,43,61,15,62,26,13,63],"数据工具","插件","其他","音频",{"id":65,"github_repo":66,"name":67,"description_en":68,"description_zh":69,"ai_summary_zh":69,"readme_en":70,"readme_zh":71,"quickstart_zh":72,"use_case_zh":73,"hero_image_url":74,"owner_login":75,"owner_name":76,"owner_avatar_url":77,"owner_bio":78,"owner_company":79,"owner_location":79,"owner_email":80,"owner_twitter":79,"owner_website":81,"owner_url":82,"languages":83,"stars":96,"forks":97,"last_commit_at":98,"license":99,"difficulty_score":23,"env_os":100,"env_gpu":101,"env_ram":100,"env_deps":102,"category_tags":106,"github_topics":107,"view_count":23,"oss_zip_url":79,"oss_zip_packed_at":79,"status":16,"created_at":125,"updated_at":126,"faqs":127,"releases":156},4161,"explosion\u002Fspacy-llm","spacy-llm","🦙 Integrating LLMs into structured NLP pipelines","spacy-llm 是一款将大型语言模型（LLM）无缝集成到 spaCy 结构化自然语言处理流水线中的开源工具。它旨在解决传统 NLP 模型依赖大量标注数据进行训练的痛点，让用户无需任何训练样本，仅通过灵活的提示词（Prompting）即可快速实现命名实体识别、文本分类、情感分析、关系抽取及翻译等多种任务。\n\n该工具特别适合希望利用 LLM 强大理解能力构建稳健 NLP 应用的开发者与研究人员。其核心亮点在于模块化的系统设计：用户可轻松定义提示策略与解析逻辑，将 LLM 原本非结构化的输出自动转化为 spaCy 标准的结构化数据对象。spacy-llm 不仅原生支持 OpenAI、Anthropic、Google PaLM 等主流商业模型 API，还能对接 Hugging Face 上的 Falcon、Llama 2 等开源模型，甚至兼容 LangChain 生态。此外，它还内置了针对长文本的“分片 - 合并”机制，有效突破模型上下文长度限制。对于需要快速原型验证或处理复杂语言任务的团队而言，spacy-llm 提供了一条高效、低门槛的技术路径。","\u003Ca href=\"https:\u002F\u002Fexplosion.ai\">\u003Cimg src=\"https:\u002F\u002Fexplosion.ai\u002Fassets\u002Fimg\u002Flogo.svg\" width=\"125\" height=\"125\" align=\"right\" \u002F>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fexplosion.ai\">\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fexplosion_spacy-llm_readme_4fb1f905ad58.png\" width=\"125\" height=\"125\" align=\"left\" style=\"margin-right:30px\" \u002F>\u003C\u002Fa>\n\n\u003Ch1 align=\"center\">\n\u003Cspan style=\"font: bold 38pt'Courier New';\">spacy-llm\u003C\u002Fspan>\n\u003Cbr>Structured NLP with LLMs\n\u003C\u002Fh1>\n\u003Cbr>\u003Cbr>\n\n[![GitHub Workflow Status](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Factions\u002Fworkflow\u002Fstatus\u002Fexplosion\u002Fspacy-llm\u002Ftest.yml?branch=main)](https:\u002F\u002Fgithub.com\u002Fexplosion\u002Fspacy-llm\u002Factions\u002Fworkflows\u002Ftest.yml)\n[![pypi Version](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Fspacy-llm.svg?style=flat-square&logo=pypi&logoColor=white)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fspacy-llm\u002F)\n[![Code style: black](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fcode%20style-black-000000.svg?style=flat-square)](https:\u002F\u002Fgithub.com\u002Fambv\u002Fblack)\n\nThis package integrates Large Language Models (LLMs) into [spaCy](https:\u002F\u002Fspacy.io), featuring a modular system for **fast prototyping** and **prompting**, and turning unstructured responses into **robust outputs** for various NLP tasks, **no training data** required.\n\n## Feature Highlight\n\n- Serializable `llm` **component** to integrate prompts into your spaCy pipeline\n- **Modular functions** to define the [**task**](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#tasks) (prompting and parsing) and [**model**](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#models)\n- Interfaces with the APIs of\n  - **[OpenAI](https:\u002F\u002Fplatform.openai.com\u002Fdocs\u002Fapi-reference\u002F)**\n  - **[Cohere](https:\u002F\u002Fdocs.cohere.com\u002Freference\u002Fgenerate)**\n  - **[Anthropic](https:\u002F\u002Fdocs.anthropic.com\u002Fclaude\u002Freference\u002F)**\n  - **[Google PaLM](https:\u002F\u002Fai.google\u002Fdiscover\u002Fpalm2\u002F)**\n  - **[Microsoft Azure AI](https:\u002F\u002Fazure.microsoft.com\u002Fen-us\u002Fsolutions\u002Fai)**\n- Supports open-source LLMs hosted on Hugging Face 🤗:\n  - **[Falcon](https:\u002F\u002Fhuggingface.co\u002Ftiiuae)**\n  - **[Dolly](https:\u002F\u002Fhuggingface.co\u002Fdatabricks)**\n  - **[Llama 2](https:\u002F\u002Fhuggingface.co\u002Fmeta-llama)**\n  - **[OpenLLaMA](https:\u002F\u002Fhuggingface.co\u002Fopenlm-research)**\n  - **[StableLM](https:\u002F\u002Fhuggingface.co\u002Fstabilityai)**\n  - **[Mistral](https:\u002F\u002Fhuggingface.co\u002Fmistralai)**\n- Integration with [LangChain](https:\u002F\u002Fgithub.com\u002Fhwchase17\u002Flangchain) 🦜️🔗 - all `langchain` models and features can be used in `spacy-llm`\n- Tasks available out of the box:\n  - [Named Entity Recognition](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#ner)\n  - [Text classification](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#textcat)\n  - [Lemmatization](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#lemma)\n  - [Relationship extraction](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#rel)\n  - [Sentiment analysis](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#sentiment)\n  - [Span categorization](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#spancat)\n  - [Summarization](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#summarization)\n  - [Entity linking](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#nel)\n  - [Translation](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#translation)\n  - [Raw prompt execution for maximum flexibility](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#raw)\n  - Soon:\n    - Semantic role labeling\n- Easy implementation of **your own functions** via [spaCy's registry](https:\u002F\u002Fspacy.io\u002Fapi\u002Ftop-level#registry) for custom prompting, parsing and model integrations. For an example, see [here](https:\u002F\u002Fspacy.io\u002Fusage\u002Flarge-language-models#example-4).\n- [Map-reduce approach](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#task-sharding) for splitting prompts too long for LLM's context window and fusing the results back together\n\n## 🧠 Motivation\n\nLarge Language Models (LLMs) feature powerful natural language understanding capabilities. With only a few (and sometimes no) examples, an LLM can be prompted to perform custom NLP tasks such as text categorization, named entity recognition, coreference resolution, information extraction and more.\n\n[spaCy](https:\u002F\u002Fspacy.io) is a well-established library for building systems that need to work with language in various ways. spaCy's built-in components are generally powered by supervised learning or rule-based approaches.\n\nSupervised learning is much worse than LLM prompting for prototyping, but for many tasks it's much better for production. A transformer model that runs comfortably on a single GPU is extremely powerful, and it's likely to be a better choice for any task for which you have a well-defined output. You train the model with anything from a few hundred to a few thousand labelled examples, and it will learn to do exactly that. Efficiency, reliability and control are all better with supervised learning, and accuracy will generally be higher than LLM prompting as well.\n\n`spacy-llm` lets you have **the best of both worlds**. You can quickly initialize a pipeline with components powered by LLM prompts, and freely mix in components powered by other approaches. As your project progresses, you can look at replacing some or all of the LLM-powered components as you require.\n\nOf course, there can be components in your system for which the power of an LLM is fully justified. If you want a system that can synthesize information from multiple documents in subtle ways and generate a nuanced summary for you, bigger is better. However, even if your production system needs an LLM for some of the task, that doesn't mean you need an LLM for all of it. Maybe you want to use a cheap text classification model to help you find the texts to summarize, or maybe you want to add a rule-based system to sanity check the output of the summary. These before-and-after tasks are much easier with a mature and well-thought-out library, which is exactly what spaCy provides.\n\n## ⏳ Install\n\n`spacy-llm` will be installed automatically in future spaCy versions. For now, you can run the following in the same virtual environment where you already have `spacy` [installed](https:\u002F\u002Fspacy.io\u002Fusage).\n\n```bash\npython -m pip install spacy-llm\n```\n\n> ⚠️ This package is still experimental and it is possible that changes made to the interface will be breaking in minor version updates.\n\n## 🐍 Quickstart\n\nLet's run some text classification using a GPT model from OpenAI. \n\nCreate a new API key from openai.com or fetch an existing one, and ensure the\nkeys are set as environmental variables. For more background information, see\nthe documentation around setting [API keys](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#api-keys).\n\n### In Python code\n\nTo do some quick experiments, from 0.5.0 onwards you can run:\n\n```python\nimport spacy\n\nnlp = spacy.blank(\"en\")\nllm = nlp.add_pipe(\"llm_textcat\")\nllm.add_label(\"INSULT\")\nllm.add_label(\"COMPLIMENT\")\ndoc = nlp(\"You look gorgeous!\")\nprint(doc.cats)\n# {\"COMPLIMENT\": 1.0, \"INSULT\": 0.0}\n```\n\nBy using the `llm_textcat` factory, the latest version of the built-in textcat task is used, \nas well as the default GPT-3-5 model from OpenAI.\n\n### Using a config file\n\nTo control the various parameters of the `llm` pipeline, we can use \n[spaCy's config system](https:\u002F\u002Fspacy.io\u002Fapi\u002Fdata-formats#config).\nTo start, create a config file `config.cfg` containing at least the following (or see the\nfull example\n[here](https:\u002F\u002Fgithub.com\u002Fexplosion\u002Fspacy-llm\u002Ftree\u002Fmain\u002Fusage_examples\u002Ftextcat_openai)):\n\n```ini\n[nlp]\nlang = \"en\"\npipeline = [\"llm\"]\n\n[components]\n\n[components.llm]\nfactory = \"llm\"\n\n[components.llm.task]\n@llm_tasks = \"spacy.TextCat.v3\"\nlabels = [\"COMPLIMENT\", \"INSULT\"]\n\n[components.llm.model]\n@llm_models = \"spacy.GPT-4.v2\"\n```\n\nNow run:\n\n```python\nfrom spacy_llm.util import assemble\n\nnlp = assemble(\"config.cfg\")\ndoc = nlp(\"You look gorgeous!\")\nprint(doc.cats)\n# {\"COMPLIMENT\": 1.0, \"INSULT\": 0.0}\n```\n\nThat's it! There's a lot of other features - prompt templating, more tasks, logging etc. For more information on how to\nuse those, check out https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models.\n\n\n## 🚀 Ongoing work\n\nIn the near future, we will\n\n- Add more example tasks\n- Support a broader range of models\n- Provide more example use-cases and tutorials\n\nPRs are always welcome!\n\n## 📝️ Reporting issues\n\nIf you have questions regarding the usage of `spacy-llm`, or want to give us feedback after giving it a spin, please use\nthe [discussion board](https:\u002F\u002Fgithub.com\u002Fexplosion\u002Fspacy-llm\u002Fdiscussions).\nBug reports can be filed on the [spaCy issue tracker](https:\u002F\u002Fgithub.com\u002Fexplosion\u002Fspacy-llm\u002Fissues). Thank you!\n\n## Migration guides\n\nPlease refer to our [migration guide](migration_guide.md).\n","\u003Ca href=\"https:\u002F\u002Fexplosion.ai\">\u003Cimg src=\"https:\u002F\u002Fexplosion.ai\u002Fassets\u002Fimg\u002Flogo.svg\" width=\"125\" height=\"125\" align=\"right\" \u002F>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fexplosion.ai\">\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fexplosion_spacy-llm_readme_4fb1f905ad58.png\" width=\"125\" height=\"125\" align=\"left\" style=\"margin-right:30px\" \u002F>\u003C\u002Fa>\n\n\u003Ch1 align=\"center\">\n\u003Cspan style=\"font: bold 38pt'Courier New';\">spacy-llm\u003C\u002Fspan>\n\u003Cbr>使用大语言模型进行结构化自然语言处理\n\u003C\u002Fh1>\n\u003Cbr>\u003Cbr>\n\n[![GitHub 工作流状态](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Factions\u002Fworkflow\u002Fstatus\u002Fexplosion\u002Fspacy-llm\u002Ftest.yml?branch=main)](https:\u002F\u002Fgithub.com\u002Fexplosion\u002Fspacy-llm\u002Factions\u002Fworkflows\u002Ftest.yml)\n[![PyPI 版本](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Fspacy-llm.svg?style=flat-square&logo=pypi&logoColor=white)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fspacy-llm\u002F)\n[![代码风格：black](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fcode%20style-black-000000.svg?style=flat-square)](https:\u002F\u002Fgithub.com\u002Fambv\u002Fblack)\n\n本包将大型语言模型（LLMs）集成到 [spaCy](https:\u002F\u002Fspacy.io) 中，提供模块化的系统，用于**快速原型设计**和**提示工程**，并将非结构化的响应转化为适用于各种 NLP 任务的**稳健输出**，且**无需训练数据**。\n\n## 功能亮点\n\n- 可序列化的 `llm` **组件**，可将提示整合到你的 spaCy 管道中\n- **模块化函数**，用于定义 [**任务**](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#tasks)（提示与解析）和 [**模型**](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#models)\n- 支持以下 API：\n  - **[OpenAI](https:\u002F\u002Fplatform.openai.com\u002Fdocs\u002Fapi-reference\u002F)**\n  - **[Cohere](https:\u002F\u002Fdocs.cohere.com\u002Freference\u002Fgenerate)**\n  - **[Anthropic](https:\u002F\u002Fdocs.anthropic.com\u002Fclaude\u002Freference\u002F)**\n  - **[Google PaLM](https:\u002F\u002Fai.google\u002Fdiscover\u002Fpalm2\u002F)**\n  - **[Microsoft Azure AI](https:\u002F\u002Fazure.microsoft.com\u002Fen-us\u002Fsolutions\u002Fai)**\n- 支持托管在 Hugging Face 🤗 上的开源 LLM：\n  - **[Falcon](https:\u002F\u002Fhuggingface.co\u002Ftiiuae)**\n  - **[Dolly](https:\u002F\u002Fhuggingface.co\u002Fdatabricks)**\n  - **[Llama 2](https:\u002F\u002Fhuggingface.co\u002Fmeta-llama)**\n  - **[OpenLLaMA](https:\u002F\u002Fhuggingface.co\u002Fopenlm-research)**\n  - **[StableLM](https:\u002F\u002Fhuggingface.co\u002Fstabilityai)**\n  - **[Mistral](https:\u002F\u002Fhuggingface.co\u002Fmistralai)**\n- 与 [LangChain](https:\u002F\u002Fgithub.com\u002Fhwchase17\u002Flangchain) 🦜️🔗 集成——所有 `langchain` 模型和功能均可在 `spacy-llm` 中使用\n- 开箱即用的任务：\n  - [命名实体识别](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#ner)\n  - [文本分类](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#textcat)\n  - [词形还原](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#lemma)\n  - [关系抽取](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#rel)\n  - [情感分析](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#sentiment)\n  - [跨度分类](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#spancat)\n  - [摘要生成](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#summarization)\n  - [实体链接](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#nel)\n  - [翻译](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#translation)\n  - [原始提示执行，以获得最大灵活性](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#raw)\n  - 即将推出：\n    - 语义角色标注\n- 通过 [spaCy 的注册机制](https:\u002F\u002Fspacy.io\u002Fapi\u002Ftop-level#registry)，可轻松实现**自定义函数**，用于自定义提示、解析及模型集成。示例请参见 [这里](https:\u002F\u002Fspacy.io\u002Fusage\u002Flarge-language-models#example-4)。\n- [Map-reduce 方法](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#task-sharding)，用于拆分超出 LLM 上下文窗口的提示，并将结果重新合并\n\n## 🧠 动机\n\n大型语言模型（LLMs）具备强大的自然语言理解能力。只需少量（有时甚至无需）示例，即可通过提示让 LLM 执行自定义的 NLP 任务，例如文本分类、命名实体识别、共指消解、信息抽取等。\n\n[spaCy](https:\u002F\u002Fspacy.io) 是一个成熟的库，可用于构建需要以多种方式处理语言的系统。spaCy 的内置组件通常基于监督学习或规则驱动的方法。\n\n对于原型设计而言，监督学习远不如 LLM 提示有效；但对于生产环境中的许多任务来说，监督学习则更为优越。能够在单个 GPU 上流畅运行的 Transformer 模型功能强大，如果你的任务有明确的输出要求，它往往是更好的选择。只需几百到几千个标注样本，你就可以训练出一个能够精确完成该任务的模型。监督学习在效率、可靠性和可控性方面都更胜一筹，准确率通常也高于 LLM 提示。\n\n`spacy-llm` 让你能够**兼得两者的优点**。你可以快速初始化一个由 LLM 提示驱动的组件组成的管道，并自由地混合其他方法驱动的组件。随着项目的推进，你可以根据需求逐步替换部分或全部 LLM 驱动的组件。\n\n当然，你的系统中也可能存在完全适合使用 LLM 的组件。如果你希望系统能够以微妙的方式综合多份文档的信息，并为你生成一份细致入微的摘要，那么使用更强大的模型无疑是更好的选择。然而，即使你的生产系统在某些任务上需要使用 LLM，也不意味着所有任务都需要 LLM。也许你可以使用一个廉价的文本分类模型来帮助你找到需要摘要的文本，或者添加一个基于规则的系统来校验摘要的合理性。这些前后处理任务，使用成熟且经过深思熟虑的库会更加容易，而 spaCy 正是这样的工具。\n\n## ⏳ 安装\n\n未来版本的 spaCy 将自动安装 `spacy-llm`。目前，你可以在已安装 `spacy` 的同一虚拟环境中运行以下命令：\n\n```bash\npython -m pip install spacy-llm\n```\n\n> ⚠️ 本包仍处于实验阶段，接口的更改可能会导致小版本更新时出现不兼容的情况。\n\n## 🐍 快速入门\n\n让我们使用 OpenAI 的 GPT 模型来进行文本分类。\n\n从 openai.com 创建一个新的 API 密钥，或获取现有的密钥，并确保将其设置为环境变量。更多背景信息，请参阅关于设置 [API 密钥](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#api-keys) 的文档。\n\n### 在 Python 代码中\n\n为了快速实验，从 0.5.0 版本开始，你可以运行如下代码：\n\n```python\nimport spacy\n\nnlp = spacy.blank(\"en\")\nllm = nlp.add_pipe(\"llm_textcat\")\nllm.add_label(\"INSULT\")\nllm.add_label(\"COMPLIMENT\")\ndoc = nlp(\"You look gorgeous!\")\nprint(doc.cats)\n# {\"COMPLIMENT\": 1.0, \"INSULT\": 0.0}\n```\n\n通过使用 `llm_textcat` 工厂，将采用最新版本的内置 textcat 任务，以及 OpenAI 默认的 GPT-3.5 模型。\n\n### 使用配置文件\n\n为了控制 `llm` 管道的各种参数，我们可以使用 \n[spaCy 的配置系统](https:\u002F\u002Fspacy.io\u002Fapi\u002Fdata-formats#config)。\n首先，创建一个名为 `config.cfg` 的配置文件，至少包含以下内容（或参阅完整的示例\n[这里](https:\u002F\u002Fgithub.com\u002Fexplosion\u002Fspacy-llm\u002Ftree\u002Fmain\u002Fusage_examples\u002Ftextcat_openai))：\n\n```ini\n[nlp]\nlang = \"en\"\npipeline = [\"llm\"]\n\n[components]\n\n[components.llm]\nfactory = \"llm\"\n\n[components.llm.task]\n@llm_tasks = \"spacy.TextCat.v3\"\nlabels = [\"COMPLIMENT\", \"INSULT\"]\n\n[components.llm.model]\n@llm_models = \"spacy.GPT-4.v2\"\n```\n\n然后运行：\n\n```python\nfrom spacy_llm.util import assemble\n\nnlp = assemble(\"config.cfg\")\ndoc = nlp(\"你看起来真漂亮！\")\nprint(doc.cats)\n# {\"COMPLIMENT\": 1.0, \"INSULT\": 0.0}\n```\n\n就是这样！还有许多其他功能——提示模板化、更多任务、日志记录等。有关如何使用这些功能的更多信息，请访问 https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models。\n\n\n## 🚀 持续工作\n\n在不久的将来，我们将\n\n- 添加更多示例任务\n- 支持更广泛的模型\n- 提供更多使用案例和教程\n\n欢迎随时提交 Pull Request！\n\n## 📝️ 报告问题\n\n如果您对 `spacy-llm` 的使用有任何疑问，或者在试用后希望向我们提供反馈，请使用\n[讨论区](https:\u002F\u002Fgithub.com\u002Fexplosion\u002Fspacy-llm\u002Fdiscussions)。\nBug 报告可以在 [spaCy 问题跟踪器](https:\u002F\u002Fgithub.com\u002Fexplosion\u002Fspacy-llm\u002Fissues) 上提交。感谢您的支持！\n\n## 迁移指南\n\n请参阅我们的 [迁移指南](migration_guide.md)。","# spacy-llm 快速上手指南\n\n`spacy-llm` 是一个将大型语言模型（LLM）集成到 spaCy 中的工具包。它允许你通过提示词（Prompting）快速构建结构化 NLP 管道，无需训练数据即可执行命名实体识别、文本分类、情感分析等任务。\n\n## 环境准备\n\n在开始之前，请确保你的开发环境满足以下要求：\n\n*   **Python 版本**：建议 Python 3.8 或更高版本。\n*   **前置依赖**：必须先安装 `spacy` 核心库。\n*   **API Key**：如果你使用云端模型（如 OpenAI, Cohere, Anthropic 等），需要提前注册账号并获取 API Key，并将其设置为环境变量（例如 `OPENAI_API_KEY`）。\n    *   *注：也支持本地部署的开源模型（如 Llama 2, Mistral 等），需自行配置相应的推理后端。*\n\n## 安装步骤\n\n请在已安装 `spacy` 的虚拟环境中执行以下命令安装 `spacy-llm`：\n\n```bash\npython -m pip install spacy-llm\n```\n\n> ⚠️ **注意**：该包目前处于实验阶段，小版本更新可能会包含破坏性接口变更。\n\n## 基本使用\n\n以下是两种最常用的快速启动方式：直接在 Python 代码中调用或使用配置文件。\n\n### 方式一：Python 代码快速实验\n\n从 0.5.0 版本开始，你可以直接使用内置工厂函数快速构建管道。以下示例使用 OpenAI 的 GPT 模型进行文本分类（默认为 GPT-3.5）：\n\n```python\nimport spacy\n\n# 创建一个空的英文管道\nnlp = spacy.blank(\"en\")\n\n# 添加 llm_textcat 组件，自动加载默认的文本分类任务和模型\nllm = nlp.add_pipe(\"llm_textcat\")\n\n# 定义分类标签\nllm.add_label(\"INSULT\")\nllm.add_label(\"COMPLIMENT\")\n\n# 处理文本\ndoc = nlp(\"You look gorgeous!\")\n\n# 输出结果\nprint(doc.cats)\n# 预期输出: {\"COMPLIMENT\": 1.0, \"INSULT\": 0.0}\n```\n\n### 方式二：使用配置文件 (config.cfg)\n\n对于更复杂的生产环境，建议使用 spaCy 的配置文件系统来精确控制模型参数和任务定义。\n\n1. 创建名为 `config.cfg` 的文件，内容如下：\n\n```ini\n[nlp]\nlang = \"en\"\npipeline = [\"llm\"]\n\n[components]\n\n[components.llm]\nfactory = \"llm\"\n\n[components.llm.task]\n@llm_tasks = \"spacy.TextCat.v3\"\nlabels = [\"COMPLIMENT\", \"INSULT\"]\n\n[components.llm.model]\n@llm_models = \"spacy.GPT-4.v2\"\n```\n\n2. 在 Python 中加载配置并运行：\n\n```python\nfrom spacy_llm.util import assemble\n\n# 从配置文件组装管道\nnlp = assemble(\"config.cfg\")\n\n# 处理文本\ndoc = nlp(\"You look gorgeous!\")\n\n# 输出结果\nprint(doc.cats)\n# 预期输出: {\"COMPLIMENT\": 1.0, \"INSULT\": 0.0}\n```\n\n通过以上步骤，你即可利用 LLM 的强大能力快速原型化 NLP 任务。更多高级功能（如自定义 Prompt 模板、支持 LangChain、Map-Reduce 长文本处理等）请参考官方文档。","某电商公司的数据团队需要每天从数万条非结构化的用户评论中，精准提取“产品缺陷”与“情感倾向”，以生成质量监控报表。\n\n### 没有 spacy-llm 时\n- **冷启动困难**：传统机器学习模型依赖大量标注数据，针对新出现的网络流行语或特定品类缺陷，重新收集和标注数据耗时数周。\n- **流程割裂**：调用大模型 API 的脚本与现有的 spaCy 预处理流水线（如分词、清洗）完全分离，代码维护成本高且容易出错。\n- **输出不稳定**：直接请求大模型返回的 JSON 格式经常因幻觉或标点错误而解析失败，导致整个批处理任务中断。\n- **调试黑盒**：难以将提示词（Prompt）工程模块化，每次调整策略都需要修改底层调用逻辑，无法快速验证效果。\n\n### 使用 spacy-llm 后\n- **零样本即时上线**：利用 spacy-llm 内置的 `textcat` 和 `ner` 任务模板，无需任何训练数据，仅通过自然语言描述即可让模型识别全新的缺陷类型。\n- **无缝管道集成**：将大模型作为标准组件直接嵌入现有的 spaCy 流水线，前端的文本清洗与后端的大模型推理在同一框架下流畅运行。\n- **结构化强约束**：工具自动处理提示词构建与结果解析，强制将大模型的非结构化回答转换为鲁棒的 Python 对象，彻底消除解析报错。\n- **敏捷提示迭代**：通过模块化配置轻松切换底层模型（如从 OpenAI 切至 Llama 2）或调整提示策略，像调节超参数一样快速优化业务效果。\n\nspacy-llm 的核心价值在于它将大模型的灵活性与工业级 NLP 流水线的稳定性完美结合，让开发者无需训练数据即可在几分钟内构建出生产级的智能文本分析系统。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fexplosion_spacy-llm_4fb1f905.png","explosion","Explosion","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fexplosion_6bc9ac84.png","Software company specializing in developer tools and tailored solutions for AI and Natural Language Processing",null,"contact@explosion.ai","https:\u002F\u002Fexplosion.ai","https:\u002F\u002Fgithub.com\u002Fexplosion",[84,88,92],{"name":85,"color":86,"percentage":87},"Python","#3572A5",96.6,{"name":89,"color":90,"percentage":91},"Jinja","#a52a22",3.3,{"name":93,"color":94,"percentage":95},"Shell","#89e051",0,1376,106,"2026-04-01T01:05:05","MIT","未说明","非必需。主要依赖外部 API（如 OpenAI, Azure 等）；若使用本地开源模型（如 Llama 2, Mistral 等托管于 Hugging Face），则需根据具体模型大小配置相应的 GPU 和显存，README 中未指定具体型号或版本。",{"notes":103,"python":100,"dependencies":104},"该工具主要作为 spaCy 的插件，通过 API 调用大型语言模型（LLM），因此通常不需要本地高性能计算资源。若选择连接本地部署的开源模型（通过 Hugging Face），则需自行配置对应的运行环境（如 PyTorch, Transformers 等）及硬件资源。使用前需设置相应服务提供商（如 OpenAI, Cohere, Anthropic 等）的 API 密钥为环境变量。该包目前处于实验阶段，接口可能在次版本更新中发生破坏性变更。",[105,67],"spacy",[13,26],[108,109,110,105,111,112,113,114,115,116,117,118,119,120,121,122,123,124],"large-language-models","llm","openai","dolly","gpt-3","gpt-4","machine-learning","named-entity-recognition","natural-language-processing","nlp","text-classification","prompt-engineering","anthropic","claude","cohere","falcon","llama","2026-03-27T02:49:30.150509","2026-04-06T12:04:04.191638",[128,133,137,142,147,152],{"id":129,"question_zh":130,"answer_zh":131,"source_url":132},18952,"为什么运行 README 示例时会报错 'cannot import name SimpleFrozenDict from confection'？","这通常是由于依赖包版本不兼容导致的。请确保安装了正确版本的 `confection` 和 `spacy-llm`。建议升级 `spacy-llm` 到最新版本（例如 0.6.2 或更高），因为新版本修复了相关的导入问题和依赖兼容性。可以使用命令 `pip install --upgrade spacy-llm` 进行更新。如果问题依旧，请检查 `confection` 的版本是否符合 `spacy-llm` 的要求。","https:\u002F\u002Fgithub.com\u002Fexplosion\u002Fspacy-llm\u002Fissues\u002F289",{"id":134,"question_zh":135,"answer_zh":136,"source_url":132},18953,"在 Intel Mac 上运行 Hugging Face 模型（如 Dolly, StableLM）时遇到 dtype 不支持或 GPU 警告怎么办？","Intel Mac 不支持 MPS 后端或某些特定的数据类型（如 BFloat16），且没有 CUDA GPU。解决方案包括：1. 使用 GGML 格式的模型（需自行集成，不在 spaCy 核心库中）；2. 升级到 Apple Silicon (M1\u002FM2) 芯片的 Mac 以获得更好的支持；3. 强制模型在 CPU 上运行，但速度会较慢。错误信息中提到的 'device_map:auto' 会导致模型加载到硬盘或 CPU，这是在没有 CUDA 时的预期行为。",{"id":138,"question_zh":139,"answer_zh":140,"source_url":141},18954,"使用 Anthropic Claude 模型时报错 'anthropic-version header is required' 如何解决？","这是因为 Anthropic API 更新了头部字段要求。该问题已在 `spacy-llm` 0.6.2 版本中修复，新版本将配置键从 `anthropic_version` 更改为 `anthropic-version` 以符合 API 要求。请将 `spacy-llm` 升级到 0.6.2 或更高版本：`pip install --upgrade spacy-llm`。升级后无需手动修改配置文件，库会自动处理正确的头部格式。","https:\u002F\u002Fgithub.com\u002Fexplosion\u002Fspacy-llm\u002Fissues\u002F323",{"id":143,"question_zh":144,"answer_zh":145,"source_url":146},18955,"如何在 spacy-llm 中使用 Azure OpenAI 服务或 GPT-3.5\u002FGPT-4 模型？","早期版本的 `spacy-llm` LangChain 集成仅支持旧的 `langchain.llms` 接口，不兼容 GPT-3.5\u002F4 所需的 `langchain.chat_models` 接口，导致报错 'Must provide an engine or deployment_id' 或 'completion operation does not work'。此问题已在 `spacy-llm` 0.6.3 版本中修复。请升级到 0.6.3 或更高版本以支持 Azure OpenAI 和最新的 GPT 模型聊天补全接口。","https:\u002F\u002Fgithub.com\u002Fexplosion\u002Fspacy-llm\u002Fissues\u002F311",{"id":148,"question_zh":149,"answer_zh":150,"source_url":151},18956,"运行 NER 任务时出现 'TimeoutError: The read operation timed out' 错误怎么办？","该错误表示请求 OpenAI API 时发生超时。解决方法包括：1. 增加配置文件中的 `max_request_time` 参数值（默认可能较短）；2. 检查网络连接是否稳定，特别是访问 OpenAI 服务的连通性；3. 如果输入文本过长，尝试减少 batch size 或截断输入；4. 在配置中增加重试次数 (`max_tries`) 和间隔时间 (`interval`) 以应对临时网络波动。","https:\u002F\u002Fgithub.com\u002Fexplosion\u002Fspacy-llm\u002Fissues\u002F355",{"id":153,"question_zh":154,"answer_zh":155,"source_url":132},18957,"spacy-llm 支持哪些类型的本地模型和云端模型？","spacy-llm 支持多种模型来源：1. 云端 API：OpenAI (GPT-3.5\u002F4), Anthropic (Claude), Azure OpenAI；2. Hugging Face 本地模型：如 Dolly, StableLM, OpenLLaMA, Falcon 等（需注意硬件兼容性，Intel Mac 可能受限）；3. LangChain 集成的其他模型。对于本地部署，建议使用具备 CUDA 支持的 GPU 以获得最佳性能，否则模型可能会回退到 CPU 运行导致速度变慢。",[157,162,167,172,177,182,187,192,197,202,207,212,217,222,227,232,237,242,247,252],{"id":158,"version":159,"summary_zh":160,"released_at":161},115385,"release-v0.7.4","# 生成的注释（2026-03-24 23:31）\n\nPython 3.14 支持、Pydantic v2 迁移及依赖项更新\n\n## 功能\n\n- 增加对 Python 3.13 和 3.14 的支持（已弃用 Python 3.9；最低版本现为 3.10）\n- 从 Pydantic v1 迁移到 Pydantic v2 原生 API\n- 更新 LangChain 集成，要求 langchain >= 1.0\n\n## 修复\n\n- 通过要求 confection >= 1.3.3，修复与 spaCy 3.8.13 的兼容性问题\n- 在 Python 3.14 上过滤来自 langchain-core 的虚假 Pydantic v1 弃用警告\n\n## 其他\n\n- 更新 LangChain 的导入方式为延迟加载，以兼容 Python 3.14\n- 更新开发依赖：langchain 1.x、openai 1.x，移除过时的固定版本锁定","2026-03-24T22:33:25",{"id":163,"version":164,"summary_zh":165,"released_at":166},115386,"release-v0.7.3","Jinja 模板库允许任意代码执行，除非在沙箱环境中运行。如果使用不受信任的配置文件来加载流水线，这可能导致任意代码执行。","2025-01-13T12:11:11",{"id":168,"version":169,"summary_zh":170,"released_at":171},115387,"v0.7.2","## ✨ 新功能和改进\n\n- 支持 Python 3.12 (#466)。\n\n## 🔴 错误修复\n\n- 确保仅在 Torch 可用时才导入或使用它 (#467)。\n\n## ⚠️ 向后不兼容的变更\n\n与 v0.7.1 相比，没有向后不兼容的变更。\n\n## 👥 贡献者\n\n@honnibal、@ines、@magdaaniol、@svlandeg\n","2024-05-17T09:36:17",{"id":173,"version":174,"summary_zh":175,"released_at":176},115388,"v0.7.1","## 🔴 错误修复\n\n- 移除 OpenAI REST 实现中对固定端点的误判检查 (#429)\n- 修复 EL 任务中，当文本中存在未高亮显示的实体时的 bug (#375)\n- 更新 `langchain` 集成，以支持 `langchain` >= 0.1, \u003C 0.2 (#433)\n\n## ⚠️ 向后不兼容变更\n\n与 v0.7.0 相比，无向后不兼容变更。\n\n## 📖 文档和示例\n\n- 将自述文件中的任务提及链接到文档 (#421)\n\n## 👥 贡献者\n\n@honnibal、@ines、@magdaaniol、@rmitsch、@svlandeg\n","2024-01-29T13:46:07",{"id":178,"version":179,"summary_zh":180,"released_at":181},115403,"v0.3.0","## :sparkles: New features and improvements\r\n\r\n- NEW: Optional storing of prompts and responses in `Doc` objects  (#127)\r\n- NEW: Optional logging of prompts and responses (#80)\r\n- NEW: Streamlit demo (#102)\r\n- NEW: Support for Cohere in backend `spacy.REST.v1` (#165)\r\n- NEW: Support for Anthropic in backend `spacy.REST.v1` (#157)\r\n- NEW: Support for OpenLLaMA via HuggingFace with the backend `spacy.OpenLLaMa_HF.v1` (#151)\r\n- NEW: Support for StableLM via HuggingFace with the backend `spacy.StableLM_HF.v1` (#141)\r\n- NEW: Lemmatization task `spacy.Lemma.v1` (#164)\r\n\r\n##  🔴 Bug fixes\r\n- Fix bug with sending empty prompts if all `Doc` objects are cached (#166)\r\n- Fix issue with `LangChain` model creation due to updated argument name (#162)\r\n \r\n## 👥 Contributors\r\n@adrianeboyd, @bdura, @honnibal, @ines, @kabirkhan, @ljvmiranda921, @rmitsch, @svlandeg, @victorialslocum, @vin-ivar ","2023-06-14T09:42:00",{"id":183,"version":184,"summary_zh":185,"released_at":186},115404,"v0.2.1","## :sparkles: New features and improvements\r\n\r\n- NEW: `llm` component supports scoring, like other spaCy components (#135)\r\n- Labels for `spacy.NER.v2`, `spacy.REL.v1`, `spacy.SpanCat.v2`, `spacy.TextCat.v2` can be specified as list (#137)\r\n\r\n##  🔴 Bug fixes\r\n- Fix type comparison in type checks failing on some platforms (#158)\r\n- Fix example 3 in readme failing (#137)\r\n\r\n## 👥 Contributors\r\n@adrianeboyd, @bdura, @honnibal, @ines, @kabirkhan, @KennethEnevoldsen, @rmitsch, @svlandeg, @vin-ivar","2023-06-05T14:21:07",{"id":188,"version":189,"summary_zh":190,"released_at":191},115400,"v0.4.0","## :sparkles: New features and improvements\r\n\r\n- NEW: Refactored to transition from backend- to model-centric architecture. Note: this is breaking, you'll need to adjust your configs  (#176)\r\n- NEW: Support for Falcon via HuggingFace (#179)\r\n- NEW: Extract prompt examples from component initialization (#163)\r\n- NEW: Summary task `spacy.Summary.v1` (#181)\r\n- NEW: Sentiment analysis task `spacy.Sentiment.v1` (#200)\r\n- More thorough check for label inconsistencies in span-related tasks NER, REL, SpanCat, TextCat (#183)\r\n- Update `langchain` pin (#196)\r\n- Make fewshot file reader more robust w.r.t. file formats (#184)\r\n\r\n## ⚠️ Backwards incompatibilities\r\n- Built-in support for MiniChain was dropped (#176)\r\n- The switch from a backend- to a model-centric architecture (#176) requires light adjustments in your config. Check out the [migration guide](https:\u002F\u002Fgithub.com\u002Fexplosion\u002Fspacy-llm\u002Fblob\u002Ffb393388e4b85ef1ac9b033dcda489b83307ecca\u002Fmigration_guide.md#03x-to-04x) to see how to update your config\r\n \r\n## 👥 Contributors\r\n@bdura, @honnibal, @ines, @kabirkhan, @koaning, @rmitsch, @shadeMe, @svlandeg, @vin-ivar ","2023-07-06T12:16:49",{"id":193,"version":194,"summary_zh":195,"released_at":196},115401,"v0.3.2","##  🔴 Bug fixes\r\n- Use `doc._context` to ensure that  `nlp.pipe(..., as_tuples=True)` works (#188)\r\n- Fix issue with caching that prevented last doc in cache batch being cached with their LLM IO data (i. e. raw prompt and LLM response) (#191)\r\n \r\n## 👥 Contributors\r\n@honnibal, @ines, @kabirkhan, @rmitsch","2023-06-26T12:24:50",{"id":198,"version":199,"summary_zh":200,"released_at":201},115389,"v0.7.0","## ✨ 新功能和改进\n\n- **新增**：通过 map-reduce 和文档分片支持任意长度的文档。更多信息请参见 [这里](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#task-sharding) (#332)\n- **新增**：新任务——\n  - `EntityLinkingTask`：使用大语言模型将文本中的实体提及链接到自定义知识库中的实体。更多信息请参见 [这里](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#nel) (#241)\n  - `TranslationTask`：将文档翻译成任意语言或从任意语言翻译过来。更多信息请参见 [这里](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#translation) (#396)\n  - `RawTask`：仅向模型提供文档内容，不使用任何提示模板，以获得最大的灵活性。更多信息请参见 [这里](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#raw) (#395)\n- 允许为所有模型配置自定义端点，以支持微调模型的使用 (#390)\n\n## 🔴 错误修复\n\n- 修复了之前 LangChain 对象集成不完整的问题 (#417)\n\n## ⚠️ 向后不兼容性\n\n与 v0.6.x 版本相比，无向后不兼容性。\n\n## 📖 文档和示例\n\n- `EntityLinkingTask`、`TranslationTask` 和 `RawTask` 的描述（https:\u002F\u002Fgithub.com\u002Fexplosion\u002FspaCy\u002Fpull\u002F12988、https:\u002F\u002Fgithub.com\u002Fexplosion\u002FspaCy\u002Fpull\u002F13183、https:\u002F\u002Fgithub.com\u002Fexplosion\u002FspaCy\u002Fpull\u002F13180）\n- 文档分片\u002Fmap-reduce 的介绍及说明，用于支持无限长文档（https:\u002F\u002Fgithub.com\u002Fexplosion\u002FspaCy\u002Fpull\u002F13214）\n\n## 👥 贡献者\n\n@honnibal、@ines、@rmitsch、@svlandeg","2024-01-19T16:27:19",{"id":203,"version":204,"summary_zh":205,"released_at":206},115390,"v0.6.4","## 🔴 错误修复\n\n- 修复 `langchain` 模型初始化，适用于那些不将 `model_name` 作为模型 ID 参数的模型 (#374)\n- 忽略由 `torch` 引发的 Pydantic 已弃用警告 (#375)\n\n## ⚠️ 向后不兼容性\n\n与 v0.6.0 相比，无向后不兼容性。\n\n## 👥 贡献者\n\n@honnibal、@ines、@shadeMe、@rmitsch\n","2023-11-17T17:19:16",{"id":208,"version":209,"summary_zh":210,"released_at":211},115391,"v0.6.3","## ✨ 新功能和改进\n\n- 为情感分析任务添加评分支持 (#331)\n- 为 `LabeledTasks` 添加标签定义 (#340)\n- 支持 `langchain` 0.0.335 (#365)\n- 通过允许为现有模型系列使用任意模型名称，支持新的 OpenAI 模型 (#356)\n\n## 🔴 错误修复\n\n- 修复 REL 提示生成时输出标签的内部 ID 而不是字符串值的问题 (#367)\n- 修复 HF 模型的设备处理问题，并允许在配置中传递 `torch_dtype` 参数 (#359)\n- 修复 Azure OpenAI 模型的基础 URL 错误 (#337)\n\n## ⚠️ 向后不兼容性\n\n与 v0.6.0 相比，无向后不兼容性。\n\n## 👥 贡献者\n\n@habibhaidari1, @honnibal, @ines, @shadeMe, @rmitsch, @svlandeg, @viveksilimkhan1","2023-11-13T18:26:36",{"id":213,"version":214,"summary_zh":215,"released_at":216},115392,"v0.6.2","## 🔴 修复的 bug\n\n- Anthropic 请求现在包含 `anthropic-version` 头（而不是 `anthropic_version`）(#328)\n\n## ⚠️ 向后不兼容的变更\n\n与 v0.6.0 相比，没有向后不兼容的变更。\n\n## 👥 贡献者\n\n@honnibal、@ines、@rmitsch\n","2023-10-16T09:17:41",{"id":218,"version":219,"summary_zh":220,"released_at":221},115393,"v0.6.1","## 🔴 修复的 bug\n\n- Llama 2 初始化默认不再使用 `use_auth_token=True` (#325)\n- Anthropic 请求中包含 `anthropic_version` 头部信息 (#325)\n\n以上两个问题在我们这边均无法复现，因此我们无法完全保证这些问题现已修复。相关讨论请参见：https:\u002F\u002Fgithub.com\u002Fexplosion\u002Fspacy-llm\u002Fissues\u002F323 和 https:\u002F\u002Fgithub.com\u002Fexplosion\u002Fspacy-llm\u002Fdiscussions\u002F324，以获取后续进展。\n\n## ⚠️ 向后不兼容的变更\n\n与 v0.6.0 相比，无向后不兼容的变更。\n\n## 👥 贡献者\n\n@honnibal、@ines、@rmitsch","2023-10-13T17:55:59",{"id":223,"version":224,"summary_zh":225,"released_at":226},115394,"v0.6.0","## ✨ 新功能和改进\n\n- **新增**: 原生 REST 支持以下模型：\n  - PaLM (#305)\n  - Azure OpenAI (#316)\n  - Mistral 7B (#313)\n  - OpenAI `gpt-3.5-turbo-instruct` — 支持批量处理！(#300)\n- 支持 `langchain==0.0.302` (#308)\n\n## 🔴 错误修复\n\n- Hugging Face 模型仅返回提示响应，而非提示 + 提示响应 (#315)\n\n## ⚠️ 向后不兼容性\n\n与 v0.5.x 版本相比，无向后不兼容性。\n\n## 📖 文档和示例\n\n- 新增使用 `spacy.NER.v3` 结合 Dolly 的用法示例 (#302)\n\n## 👥 贡献者\n\n@honnibal, @ines,  @rmitsch","2023-10-05T17:54:42",{"id":228,"version":229,"summary_zh":230,"released_at":231},115395,"v0.5.1","## 🔴 Bug fixes\r\n\r\n- Fix  Fix `confection` dependency pin (#290)","2023-09-11T09:01:02",{"id":233,"version":234,"summary_zh":235,"released_at":236},115396,"v0.5.0","## ✨ New features and improvements\r\n\r\n- **NEW**: More accurate [Chain-of-Thought (CoT) NER](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#ner-v3) prompting with `spacy.NER.v3`(#180)\r\n- **NEW**: Task-specific [component factories](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#config) for `llm_ner` , `llm_spancat`, `llm_rel`,`llm_textcat` , `llm_sentiment`,  `llm_summarization`(#243,  #283)\r\n- **NEW**: Implementation of `add_label` functionality to more easily work with an `llm` component directly in Python code and not (necessarily) through the config system (#277)\r\n- New `v2` model versions for the OpenAI models that set [reasonable defaults](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models#models-rest) for `temperature` and `max_tokens` (#236)\r\n- Functionality to ignore occassional blocking errors from Cohere (#233) \r\n- Support for Pydantic v1 and v2 (#261, #275)\r\n- Internal refactoring, including renaming of v1 Jinja templates (#242)\r\n- Empty the cache of `torch.cuda` in between calls (#242)\r\n- Various improvements to the test suite and CI\r\n\r\n## 🔴 Bug fixes\r\n\r\n- Fix Anthropic chat endpoints (#230)\r\n\r\n## ⚠️ Backwards incompatibilities\r\n\r\n- Though significant refactoring of internal modules has happened, this release should not introduce any backwards incompatibilities for user-facing functionality.\r\n- Check our [migration guide](https:\u002F\u002Fgithub.com\u002Fexplosion\u002Fspacy-llm\u002Fblob\u002Fmain\u002Fmigration_guide.md#04x-to-05x) if you want to update the SpanCat or NER task from `v1` or `v2` to `v3`.\r\n\r\n## 📖 Documentation and examples\r\n\r\n- Updated [usage](https:\u002F\u002Fspacy.io\u002Fusage\u002Flarge-language-models) documentation\r\n- Updated [API](https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models) documentation\r\n- New Chain-of-Though [example](https:\u002F\u002Fgithub.com\u002Fexplosion\u002Fspacy-llm\u002Ftree\u002Fmain\u002Fusage_examples\u002Fner_v3_openai) with GPT 3.5\r\n- New `textcat` [example](https:\u002F\u002Fgithub.com\u002Fexplosion\u002Fspacy-llm\u002Ftree\u002Fmain\u002Fusage_examples\u002Ftextcat_dolly) with Dolly\r\n\r\n## 👥 Contributors\r\n\r\n@adrianeboyd, @honnibal, @ines, @kabirkhan, @ljvmiranda921,  @rmitsch, @svlandeg, @victorialslocum, @vinbo8\r\n","2023-09-08T09:29:28",{"id":238,"version":239,"summary_zh":240,"released_at":241},115397,"v0.4.3","## :sparkles: New features and improvements\r\n- NEW: Support for LLama 2 via Huggingface (#225)\r\n- NEW: Support for Anthropic's Claude 2  (#231)\r\n\r\n##  🔴 Bug fixes\r\n- Corrects how runtime arguments are passed to Falcon (#222)\r\n- Fixes incorrect capitalization in example in readme (#227)\r\n\r\n## 📖 Documentation and examples\r\nWe've moved most of our documentation to the main spaCy docs:\r\n- Usage pages: https:\u002F\u002Fspacy.io\u002Fusage\u002Flarge-language-models\r\n- API pages: https:\u002F\u002Fspacy.io\u002Fapi\u002Flarge-language-models\r\n \r\n## 👥 Contributors\r\n@honnibal, @ines, @koaning, @rmitsch, @svlandeg, @victorialslocum ","2023-07-25T14:02:55",{"id":243,"version":244,"summary_zh":245,"released_at":246},115398,"v0.4.2","##  🔴 Bug fixes\r\n- Guard REL against entity indices higher than the total number of entities available (#219)\r\n- Warn instead of raising an error when API credentials are wrong for REST-based models (#218)\r\n \r\n## 👥 Contributors\r\n@honnibal, @ines, @rmitsch, @svlandeg ","2023-07-14T09:47:52",{"id":248,"version":249,"summary_zh":250,"released_at":251},115399,"v0.4.1","## :sparkles: New features and improvements\r\n- Verify authentication details at init (instead of at run) time for Anthropic and Cohere models (#206)\r\n\r\n##  🔴 Bug fixes\r\n- Update OpenLLaMA model names after updates on HuggingFace (#209)\r\n- Fix incorrectly spelled model names for OpenAI models in migration guide (#210)\r\n \r\n## 👥 Contributors\r\n@honnibal, @ines, @rmitsch, @svlandeg ","2023-07-11T11:36:48",{"id":253,"version":254,"summary_zh":255,"released_at":256},115402,"v0.3.1","## :sparkles: New features and improvements\r\n\r\n- Make type validation optional with the new `validate_types` flag (#178)\r\n\r\n##  🔴 Bug fixes\r\n- Fixed `nlp.pipe_labels()` not working for the `llm` component (#175)\r\n \r\n## 👥 Contributors\r\n@honnibal, @ines, @kabirkhan, @rmitsch","2023-06-23T13:22:54"]