{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-09-20T18:00:45.505Z","headline":"Hugging Face 发布开源工具 funes，为编码智能体提供可本地持有的记忆层","description":"Hugging Face 发布开源工具 funes，为 Claude Code、Codex、pi、Hermes 等编码智能体提供本地记忆层，把已有会话记录索引成 Lance 数据集，一条 funes add 命令即可让 Agent 自主召回原始出处（Agent、时间戳、会话、轮次）。","url":"https://www.aioga.com/news/cmtlg0ht406hlrow5clznd0ld/","mainEntityOfPage":"https://www.aioga.com/news/cmtlg0ht406hlrow5clznd0ld/","datePublished":"2026-09-03T00:00:00.000Z","dateModified":"2026-09-03T00:00:00.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://huggingface.co/blog/funes","https://aihot.virxact.com/items/cmtlg0ht406hlrow5clznd0ld"],"canonicalUrl":"https://www.aioga.com/news/cmtlg0ht406hlrow5clznd0ld/","directAnswer":{"@type":"Answer","text":"Hugging Face 发布开源工具 funes，为 Claude Code、Codex、pi 和 Hermes 等编码智能体提供本地记忆层。工具可将已有会话记录索引为 Lance 数据集，并通过一条命令接入智能体工作流。","url":"https://www.aioga.com/news/cmtlg0ht406hlrow5clznd0ld/","dateCreated":"2026-09-03T00:00:00.000Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"huggingface.co source article","url":"https://huggingface.co/blog/funes","datePublished":"2026-09-03T00:00:00.000Z","provider":{"@type":"Organization","name":"huggingface.co","url":"https://huggingface.co/blog/funes"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmtlg0ht406hlrow5clznd0ld","datePublished":"2026-09-03T00:00:00.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmtlg0ht406hlrow5clznd0ld"}}],"aggregationSource":"Hugging Face：Blog（RSS）","originalPublisher":{"name":"huggingface.co","url":"https://huggingface.co/blog/funes"},"geoDeepAnswer":null,"article":{"id":"cmtlg0ht406hlrow5clznd0ld","slug":"cmtlg0ht406hlrow5clznd0ld","url":"https://www.aioga.com/news/cmtlg0ht406hlrow5clznd0ld/","title":"Hugging Face 发布开源工具 funes，为编码智能体提供可本地持有的记忆层","title_en":"","summary":"Hugging Face 发布开源工具 funes，为 Claude Code、Codex、pi、Hermes 等编码智能体提供本地记忆层，把已有会话记录索引成 Lance 数据集，一条 funes add 命令即可让 Agent 自主召回原始出处（Agent、时间戳、会话、轮次）。","source":"Hugging Face：Blog（RSS）","sourceUrl":"https://huggingface.co/blog/funes","aiHotUrl":"https://aihot.virxact.com/items/cmtlg0ht406hlrow5clznd0ld","publishedAt":"2026-09-03T00:00:00.000Z","category":"行业动态","score":72,"selected":true,"articleBody":["Add memory to the agent you already use ：#add-memory-to-the-agent-you-already-use A memory is a dataset, not a service ：#a-memory-is-a-dataset-not-a-service Ask first, wire later ：#ask-first-wire-later Switching agents without losing the thread ：#switching-agents-without-losing-the-thread The cheapest way out of a long session ：#the-cheapest-way-out-of-a-long-session Stop starting from zero ：#stop-starting-from-zero Built on open source ：#built-on-open-source I work across several machines, and I switch coding agents depending on the task. Every one of them meets my projects as a stranger. The reasoning from “last Tuesday” disappears when the session ends. Each new agent, on each new host, starts from zero.","Earlier this year, Software Forgets: Agent Traces Are the Memory made the case that coding agents already produce the record we keep losing. As they search a codebase, try approaches, hit errors, read documentation, and change direction, they leave behind a dense account of not just what changed, but why .","While the diagnosis is correct, traces are only potential memory. The session logs of an agent are still just an archive. You cannot grep your way to “why did we move off the streaming parser?” across ten thousand turns. For an agent to use those traces while it works, they need indexing, retrieval, ranking, and exact provenance.","That is what funes：https://github.com/huggingface/funes provides. It is a durable memory layer for your agents (Claude Code, Codex, pi, and Hermes). It is built from the sessions already on your machine. It works locally and becomes part of your agent's normal workflow with one command. When you want it to, it can also travel to a Hugging Face dataset you own, private by default.","funes is a single binary. Its default inference backend has no ML runtime dependency, and embedding and reranking happen on your machine . Install it:","That one add command builds the first index, gives the agent recall and get tools, and installs the automation that indexes each completed turn. Indexing is incremental, with new runs adding new turns rather than embedding the whole history again. The older and deeper content can backfill in bounded steps.","From there, you just work. When a task touches a past decision, rationale, or finding, the agent can reach for recall itself. You do not need to remember the old session or paste its context into the new one.","：https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/funes/recall.gif","With funes added, recall happens inside the conversation. The agent reaches for its memory on its own and names the session behind its answer.","recall returns the original text, not a summary, and shows exactly where it came from (the agent, timestamp, session, and turn). Each result includes a get command that opens the full turn and its surrounding context.","Underneath, one deterministic pipeline parses every supported trace into the same turn-and-block shape, chunks it, embeds it with a pinned local model, and writes it to a local Lance dataset. A query combines vector and BM25 search, fuses their rankings, reranks the candidates with a cross-encoder, reweights them by recency, and attaches neighboring chunks.","That design gives funes three important properties:","The agent as a stranger problem is already solved on one machine. But memory gets more useful when the next agent is running somewhere else.","To make a memory follow your work, bind one when you add funes to an agent:","The bind publishes your current memory there. funes then keeps it current, indexing each turn locally and publishing at session boundaries. The agent recalls from it throughout. Run the same command on another machine and the memory follows you there.","Underneath, the local memory is a Lance dataset, and the shared memory is a Hugging Face dataset (private by default) you own.","Before anything reaches the Hub, credentials have already been redacted during indexing. Publishing then scans every chunk again and withholds anything that still looks like a secret. The scanner behind this is documented in SECURITY.md ：https://github.com/huggingface/funes/blob/main/SECURITY.md, including what it does and doesn't cover.","When an agent reads a remote memory, funes caches the dataset files locally, so warm queries return to local speed. The Hub supplies the ownership, access control, versioning, and distribution it already supplies for other datasets. Your memory does not become an account in a separate memory service, and you do not rent it back through an API.","recall is shaped for agents. When you want to put a question to a memory yourself, use ask . It reads your local memory by default:","Or point it at a shared memory. We published a memory：https://huggingface.co/datasets/huggingface/funes-memory of funes development, so you can ask why funes works the way it does without creating a memory of your own:","：https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/funes/ask.gif","funes ask is the read-only, one-question sibling of funes add . It recalls the passages, hands them to a coding agent, and returns a grounded answer that names its sources. It does not install an integration or change the agent's persistent setup.","A retrieval miss is not papered over. If the passages do not support an answer, the agent says so. You can rephrase the question or add funes to the agent so it can search the memory iteratively during normal work.","A shared memory is not tied to the agent or model that created it. Start a task in Claude Code, continue it in Codex next week, and the second agent can recall the first agent's reasoning. Use pi with a local model or one served through the Hugging Face router, then return to Claude.","：https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/funes/cross-agents.gif","Claude makes a decision; a hook indexes it; Codex recalls it in another session. The older hits in the demo are earlier recordings of the same experiment: an append-only memory remembered the rehearsals too.","This matters in a few different scopes:","Published memories carry a dataset card and the funes tag, making them recognizable and discoverable on the Hub：https://huggingface.co/datasets?other=funes. The Hub already hosts open weights and datasets. funes adds open working memory. It holds the decisions, failed approaches, and rationale behind a project, queryable by another agent and traceable to the sessions that produced them.","A long investigation bloats a session until each turn costs more to carry the context than to do the work. The usual answers are to let the agent compact and carry on, or to write a handoff and start fresh. Recall is a third, so we measured them against each other on the handoff-vs-recall benchmark : two tasks whose answer cannot be reconstructed without the session prior knowledge.","Compaction is what most agents do by default, and it was the only one of the three whose result divided: it arrived on one task and never arrived on the other. Where it failed, its summary had flattened the findings that mattered. Recall returns the passages themselves, so a finding does not have to survive summarization.","Recall was the cheapest of the three on both tasks, 8x cheaper than a written handoff on one and 4x on the other.","：https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/funes/handoff-recall-cost.jpg","The lighter segment of each bar is the one-time charge for preparing the channel, the handoff or the compaction, paid before the first question is asked and counted once. A cross marks a channel that never arrives, and so has no cost per success.","“To think is to forget differences, generalize, make abstractions.” — Jorge Luis Borges, Funes the Memorious","Your agents already wrote the record. funes lives at github.com/huggingface/funes ：https://github.com/huggingface/funes, one command away from turning that record into a memory the next agent can read, on whichever machine you happen to be on.","funes invents little of this. It leans on open-source embedding models good enough to run locally, on Lance：https://github.com/lancedb/lance's append-only datasets with cheap incremental writes, and on the Hub's caching and content-dedup for datasets. The work is in fitting them into a memory an agent can actually use.","funes is open source too. Open an issue：https://github.com/huggingface/funes/issues for anything from an install snag to a recall that missed, or an agent you'd like supported."],"articleImages":[{"sourceUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/647995564be04c76ce4547b3/-tlmMQgr-CTm5TCIjGApB.jpeg","alt":"David Corvoysier's avatar","afterParagraph":0,"url":"/media/articles/cmtlg0ht406hlrow5clznd0ld/5735a7e939052217.webp"},{"sourceUrl":"https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/funes/recall.gif","alt":"A coding agent reaches for funes on its own, recalls an earlier decision, and grounds its answer in the retrieved session","afterParagraph":6,"url":"/media/articles/cmtlg0ht406hlrow5clznd0ld/2642a6bab9d48210.gif"},{"sourceUrl":"https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/funes/ask.gif","alt":"Asking the published funes memory why it is append-only; funes retrieves the relevant sessions and a coding agent answers from them","afterParagraph":19,"url":"/media/articles/cmtlg0ht406hlrow5clznd0ld/06bcba3c2a28c6e7.gif"},{"sourceUrl":"https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/funes/cross-agents.gif","alt":"Claude Code chooses an embedding model, then Codex recalls that decision in a separate session","afterParagraph":23,"url":"/media/articles/cmtlg0ht406hlrow5clznd0ld/75b62ff8eab258a6.gif"}],"mediaStatus":"ok","articleBodyZh":["为你已经使用的代理添加记忆：#add-memory-to-the-agent-you-already-use 记忆是数据集，而不是服务：#a-memory-is-a-dataset-not-a-service 先询问，后连接：#ask-first-wire-later 在不丢失上下文的情况下切换代理：#switching-agents-without-losing-the-thread 结束长会话的最便宜方法：#the-cheapest-way-out-of-a-long-session 停止从零开始：#stop-starting-from-zero 基于开源构建：#built-on-open-source 我在多台机器上工作，根据任务切换编码代理。每一个代理在接手我的项目时都是陌生的。“上周二”的推理在会话结束时消失。每一个新的代理，在每台新的主机上，都是从零开始。","今年早些时候，《软件会遗忘：代理轨迹就是记忆》提出了这样一个观点：编码代理已经生成了我们一直丢失的记录。当它们搜索代码库、尝试方法、遇到错误、阅读文档并改变方向时，它们留下的不仅是发生了什么变化，还有为什么发生的详细记录。","虽然这个诊断是正确的，但轨迹只是潜在的记忆。代理的会话日志仍然只是一个档案。你无法通过grep在一万次操作中找到“为什么我们放弃流式解析器？” 这样的原因。为了让代理在工作时使用这些轨迹，它们需要索引、检索、排序和精确的来源记录。","这就是 funes（https://github.com/huggingface/funes） 提供的功能。它为你的代理（Claude Code、Codex、pi 和 Hermes）提供了一个持久的记忆层。它是基于你机器上已有的会话构建的。本地运行，并且只需一个命令就能成为代理的正常工作流程的一部分。当你需要时，它也可以上传到你拥有的 Hugging Face 数据集，默认是私有的。","funes 是一个单一的二进制文件。它默认的推理后端没有机器学习运行时依赖，并且嵌入和重排序操作在你的本地机器上执行。安装它：","那一条添加命令会构建第一个索引，提供代理的回忆和获取工具，并安装自动化程序以索引每一个完成的操作。索引是增量的，新运行只会添加新的操作，而不会重新嵌入整个历史记录。较旧和较深的内容可以在有限的步骤中补充回填。","从那里开始，你只需工作即可。当一个任务涉及过去的决策、理由或发现时，代理可以直接调用记忆本身。你不需要记住旧的会话或将其内容粘贴到新的会话中。","：https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/funes/recall.gif","加入 funes 后，记忆会在对话中发生。代理会自己调用记忆，并标明答案背后的会话。","recall 返回原始文本，而不是摘要，并准确显示其来源（代理、时间戳、会话和步骤）。每个结果都包含一个 get 命令，可打开整个步骤及其周围的上下文。","在底层，一个确定性的流水线将每个支持的追踪解析为相同的步骤和块形状，对其分块，用固定的本地模型进行嵌入，并写入本地 Lance 数据集。查询结合向量搜索和 BM25 搜索，融合它们的排名，用跨编码器重新排序候选项，根据最近性重新加权，并附加相邻块。","这种设计赋予 funes 三个重要特性：","在一台机器上，代理作为一个陌生问题已经解决。但当下一个代理在其他地方运行时，记忆会变得更有用。","为了让记忆跟随你的工作，在向代理添加 funes 时绑定一个记忆：","绑定会将你当前的记忆发布到那儿。funes 随后保持其最新状态，在本地索引每个步骤，并在会话边界处发布。代理可以在整个过程中调用它。在另一台机器上运行相同的命令，记忆也会跟随你到那里。","在底层，本地记忆是一个 Lance 数据集，而共享记忆是你拥有的 Hugging Face 数据集（默认私有）。","在任何数据达到 Hub 之前，凭证已在索引过程中被处理。发布时会再次扫描每个块，并屏蔽任何仍然看起来像是秘密的信息。其背后的扫描器在 SECURITY.md 中有文档说明：https://github.com/huggingface/funes/blob/main/SECURITY.md，包括它覆盖和不覆盖的内容。","当代理读取远程内存时，funes 会将数据集文件缓存到本地，因此热查询会返回到本地速度。Hub 提供所有权、访问控制、版本管理和分发，就像它对其他数据集提供的一样。您的内存不会成为独立内存服务的账户，也不会通过 API 租回它。","回忆是为代理设计的。当您自己想向内存提问时，请使用 ask。它默认读取您的本地内存：","或者指向共享内存。我们发布了一个 memory：https://huggingface.co/datasets/huggingface/funes-memory 记录 funes 的开发情况，因此您可以在不创建自己内存的情况下询问 funes 为什么会以这种方式工作：","：https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/funes/ask.gif","funes ask 是 funes add 的只读、单次提问兄弟功能。它回忆片段，将其交给编码代理，并返回带有来源名称的有根据的答案。它不会安装集成或更改代理的持久设置。","检索失败不会被掩盖。如果这些片段无法支持答案，代理会说明。您可以重新表述问题，或将 funes 添加到代理中，以便它在正常工作期间迭代搜索内存。","共享内存不与创建它的代理或模型绑定。在 Claude Code 中开始任务，下周继续在 Codex 中进行，第二个代理可以回忆第一个代理的推理。使用 pi 进行本地模型或通过 Hugging Face 路由提供的模型，然后返回 Claude。","：https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/funes/cross-agents.gif","Claude 做出决策；钩子对其进行索引；Codex 在另一个会话中回忆它。演示中的早期命中是同一实验的早期记录：一个仅追加的内存也记住了排练过程。","这在几个不同的范围内很重要：","已发布的记忆带有数据集卡和 funes 标签，使其在 Hub 上易于识别和发现：https：//huggingface.co/datasets？other=funes。Hub 已经托管开放权重和数据集。Funes 增加了开放工作记忆。它保存项目背后的决策、失败方法和理由，可由其他代理查询，并可追溯到产生这些内容的会话。","长时间的调查会让会话膨胀，直到每回合承载上下文的成本超过完成工作的成本。通常的答案是让代理压缩并继续，或者写一个交接文件重新开始。回忆是第三个，所以我们在交接与回忆基准测试中进行了比较：这两个任务的答案在没有会话事先知识的情况下无法重建。","压缩是大多数代理默认会做的，而在三者中，它是唯一结果分裂的：它在一个任务上得出，另一个任务却从未得到。在失败的地方，其总结压平了重要的发现。回忆返回了段落本身，因此发现不必经得起摘要。","召回在这三种任务中最便宜，其中一项比书面交接便宜8倍，另一项是4倍。","：https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/funes/handoff-recall-cost.jpg","每条条中较轻的部分是一次性准备通道的费用，即切换或压缩，在第一个问题提出前支付，并计一次。叉号标记了一个从未到达的通道，因此没有每次成功的成本。","“思考就是忘记差异，泛化，抽象化。”——豪尔赫·路易斯·博尔赫斯，《记忆中的富内斯》","你的特工已经写好了记录。富内斯住在 github.com/huggingface/funes：https://github.com/huggingface/funes，只需一个命令，就能把那个记录变成下一个特工能读取的内存，无论你所在的哪台机器上。","Funes 几乎没有发明这些内容。它依赖于足够优秀的开源嵌入模型，可以在本地运行，使用 Lance：https：//github.com/lancedb/lance 的仅附加数据集，支持廉价的增量写入，以及 Hub 的缓存和内容去量。工作重点是将它们融入代理实际可用的内存中。","funes 也是开源的。对于从安装问题到未能成功召回的记忆，或是你希望支持的代理，任何事项都可以在 https://github.com/huggingface/funes/issues 上提交 issue。"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"Hugging Face 发布开源工具 funes，为 Claude Code、Codex、pi 和 Hermes 等编码智能体提供本地记忆层。工具可将已有会话记录索引为 Lance 数据集，并通过一条命令接入智能体工作流。","background":"来源称，编码智能体的会话日志记录了搜索代码、尝试方案、错误和调整方向，但原始日志仍是存档，难以直接检索决策原因。funes 提供索引、检索、排序和精确出处能力，并支持本地处理。","viewpoint":"Aioga 判断：funes 的核心价值在于把编码智能体已有的会话轨迹转化为可召回、可追溯的记忆数据，而不是另建一套独立服务。其本地运行和开源属性也使部署方式更贴近现有工作流。","implications":"可能影响：编码智能体之间切换时，历史决策和上下文有机会被继续使用，但这不代表召回结果始终完整或准确。使用方需要关注索引范围、出处核验、机器间同步方式及会话数据的隐私边界。","nextStep":"后续观察：应关注 funes 对不同智能体会话格式的支持范围、召回与排序效果、跨机器使用方式，以及本地数据上传至 Hugging Face 数据集时的实际控制选项。","evidenceRefs":["title","summary","articleBody","source"],"status":"published","aiGenerated":true,"autoApproved":true,"generatedBy":"aioga-editorial:gpt-5.6-sol","reviewedBy":"aioga-editorial-review:gpt-5.6-sol","generatedAt":"2026-09-03T12:05:14.458Z","sourceHash":"5cbb010b3dd42449","review":{"approved":true,"groundedness":94,"clarity":90,"duplicationRisk":18,"blockingIssues":[],"notes":["观点与可能影响已通过“Aioga 判断”“可能影响”“后续观察”等措辞明确标注，没有冒充来源事实。","可选优化：可将“一条命令”具体写为“funes add”，并说明其会建立首个索引、提供 recall 和 get 工具以及安装增量索引自动化。","关于召回完整性、索引范围、隐私边界和效果评估的表述属于合理风险提示，但并非来源已经验证的产品结论，当前措辞已保留不确定性。"]},"validation":{"passed":true,"mode":"ai-auto","revisions":0,"checks":["schema","length","source-attribution","editorial-labels","inference-boundary","low-source-overlap","no-html","independent-ai-review"]}},"tags":["行业动态","Hugging Face：Blog（RSS）"],"translations":{"zh-CN":{"title":"Hugging Face 发布开源工具 funes，为编码智能体提供可本地持有的记忆层","summary":"Hugging Face 发布开源工具 funes，为 Claude Code、Codex、pi、Hermes 等编码智能体提供本地记忆层，把已有会话记录索引成 Lance 数据集，一条 funes add 命令即可让 Agent 自主召回原始出处（Agent、时间戳、会话、轮次）。","category":"行业动态","source":"huggingface.co","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Hugging Face 发布开源工具 funes，为编码智能体提供可本地持有的记忆层 - Aioga AI资讯","description":"Hugging Face 发布开源工具 funes，为 Claude Code、Codex、pi、Hermes 等编码智能体提供本地记忆层，把已有会话记录索引成 Lance 数据集，一条 funes add 命令即可让 Agent 自主召回原始出处（Agent、时间戳、会话、轮次）。","url":"https://www.aioga.com/news/cmtlg0ht406hlrow5clznd0ld/","articleBody":["为你已经使用的代理添加记忆：#add-memory-to-the-agent-you-already-use 记忆是数据集，而不是服务：#a-memory-is-a-dataset-not-a-service 先询问，后连接：#ask-first-wire-later 在不丢失上下文的情况下切换代理：#switching-agents-without-losing-the-thread 结束长会话的最便宜方法：#the-cheapest-way-out-of-a-long-session 停止从零开始：#stop-starting-from-zero 基于开源构建：#built-on-open-source 我在多台机器上工作，根据任务切换编码代理。每一个代理在接手我的项目时都是陌生的。“上周二”的推理在会话结束时消失。每一个新的代理，在每台新的主机上，都是从零开始。","今年早些时候，《软件会遗忘：代理轨迹就是记忆》提出了这样一个观点：编码代理已经生成了我们一直丢失的记录。当它们搜索代码库、尝试方法、遇到错误、阅读文档并改变方向时，它们留下的不仅是发生了什么变化，还有为什么发生的详细记录。","虽然这个诊断是正确的，但轨迹只是潜在的记忆。代理的会话日志仍然只是一个档案。你无法通过grep在一万次操作中找到“为什么我们放弃流式解析器？” 这样的原因。为了让代理在工作时使用这些轨迹，它们需要索引、检索、排序和精确的来源记录。","这就是 funes（https://github.com/huggingface/funes） 提供的功能。它为你的代理（Claude Code、Codex、pi 和 Hermes）提供了一个持久的记忆层。它是基于你机器上已有的会话构建的。本地运行，并且只需一个命令就能成为代理的正常工作流程的一部分。当你需要时，它也可以上传到你拥有的 Hugging Face 数据集，默认是私有的。","funes 是一个单一的二进制文件。它默认的推理后端没有机器学习运行时依赖，并且嵌入和重排序操作在你的本地机器上执行。安装它：","那一条添加命令会构建第一个索引，提供代理的回忆和获取工具，并安装自动化程序以索引每一个完成的操作。索引是增量的，新运行只会添加新的操作，而不会重新嵌入整个历史记录。较旧和较深的内容可以在有限的步骤中补充回填。","从那里开始，你只需工作即可。当一个任务涉及过去的决策、理由或发现时，代理可以直接调用记忆本身。你不需要记住旧的会话或将其内容粘贴到新的会话中。","：https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/funes/recall.gif","加入 funes 后，记忆会在对话中发生。代理会自己调用记忆，并标明答案背后的会话。","recall 返回原始文本，而不是摘要，并准确显示其来源（代理、时间戳、会话和步骤）。每个结果都包含一个 get 命令，可打开整个步骤及其周围的上下文。","在底层，一个确定性的流水线将每个支持的追踪解析为相同的步骤和块形状，对其分块，用固定的本地模型进行嵌入，并写入本地 Lance 数据集。查询结合向量搜索和 BM25 搜索，融合它们的排名，用跨编码器重新排序候选项，根据最近性重新加权，并附加相邻块。","这种设计赋予 funes 三个重要特性：","在一台机器上，代理作为一个陌生问题已经解决。但当下一个代理在其他地方运行时，记忆会变得更有用。","为了让记忆跟随你的工作，在向代理添加 funes 时绑定一个记忆：","绑定会将你当前的记忆发布到那儿。funes 随后保持其最新状态，在本地索引每个步骤，并在会话边界处发布。代理可以在整个过程中调用它。在另一台机器上运行相同的命令，记忆也会跟随你到那里。","在底层，本地记忆是一个 Lance 数据集，而共享记忆是你拥有的 Hugging Face 数据集（默认私有）。","在任何数据达到 Hub 之前，凭证已在索引过程中被处理。发布时会再次扫描每个块，并屏蔽任何仍然看起来像是秘密的信息。其背后的扫描器在 SECURITY.md 中有文档说明：https://github.com/huggingface/funes/blob/main/SECURITY.md，包括它覆盖和不覆盖的内容。","当代理读取远程内存时，funes 会将数据集文件缓存到本地，因此热查询会返回到本地速度。Hub 提供所有权、访问控制、版本管理和分发，就像它对其他数据集提供的一样。您的内存不会成为独立内存服务的账户，也不会通过 API 租回它。","回忆是为代理设计的。当您自己想向内存提问时，请使用 ask。它默认读取您的本地内存：","或者指向共享内存。我们发布了一个 memory：https://huggingface.co/datasets/huggingface/funes-memory 记录 funes 的开发情况，因此您可以在不创建自己内存的情况下询问 funes 为什么会以这种方式工作：","：https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/funes/ask.gif","funes ask 是 funes add 的只读、单次提问兄弟功能。它回忆片段，将其交给编码代理，并返回带有来源名称的有根据的答案。它不会安装集成或更改代理的持久设置。","检索失败不会被掩盖。如果这些片段无法支持答案，代理会说明。您可以重新表述问题，或将 funes 添加到代理中，以便它在正常工作期间迭代搜索内存。","共享内存不与创建它的代理或模型绑定。在 Claude Code 中开始任务，下周继续在 Codex 中进行，第二个代理可以回忆第一个代理的推理。使用 pi 进行本地模型或通过 Hugging Face 路由提供的模型，然后返回 Claude。","：https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/funes/cross-agents.gif","Claude 做出决策；钩子对其进行索引；Codex 在另一个会话中回忆它。演示中的早期命中是同一实验的早期记录：一个仅追加的内存也记住了排练过程。","这在几个不同的范围内很重要：","已发布的记忆带有数据集卡和 funes 标签，使其在 Hub 上易于识别和发现：https：//huggingface.co/datasets？other=funes。Hub 已经托管开放权重和数据集。Funes 增加了开放工作记忆。它保存项目背后的决策、失败方法和理由，可由其他代理查询，并可追溯到产生这些内容的会话。","长时间的调查会让会话膨胀，直到每回合承载上下文的成本超过完成工作的成本。通常的答案是让代理压缩并继续，或者写一个交接文件重新开始。回忆是第三个，所以我们在交接与回忆基准测试中进行了比较：这两个任务的答案在没有会话事先知识的情况下无法重建。","压缩是大多数代理默认会做的，而在三者中，它是唯一结果分裂的：它在一个任务上得出，另一个任务却从未得到。在失败的地方，其总结压平了重要的发现。回忆返回了段落本身，因此发现不必经得起摘要。","召回在这三种任务中最便宜，其中一项比书面交接便宜8倍，另一项是4倍。","：https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/funes/handoff-recall-cost.jpg","每条条中较轻的部分是一次性准备通道的费用，即切换或压缩，在第一个问题提出前支付，并计一次。叉号标记了一个从未到达的通道，因此没有每次成功的成本。","“思考就是忘记差异，泛化，抽象化。”——豪尔赫·路易斯·博尔赫斯，《记忆中的富内斯》","你的特工已经写好了记录。富内斯住在 github.com/huggingface/funes：https://github.com/huggingface/funes，只需一个命令，就能把那个记录变成下一个特工能读取的内存，无论你所在的哪台机器上。","Funes 几乎没有发明这些内容。它依赖于足够优秀的开源嵌入模型，可以在本地运行，使用 Lance：https：//github.com/lancedb/lance 的仅附加数据集，支持廉价的增量写入，以及 Hub 的缓存和内容去量。工作重点是将它们融入代理实际可用的内存中。","funes 也是开源的。对于从安装问题到未能成功召回的记忆，或是你希望支持的代理，任何事项都可以在 https://github.com/huggingface/funes/issues 上提交 issue。"]},"en":{"title":"Hugging Face Releases Open-Source Tool Funes, Providing Local Memory Layers for Coding Agents","summary":"Hugging Face has released the open-source tool Funes, offering local memory layers for coding agents like Claude Code, Codex, pi, and Hermes. It indexes existing conversation records into a Lance dataset, and a single `funes add` command allows the agent to autonomously recall the original source (agent, timestamp, conversation, round).","category":"Industry","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Hugging Face Releases Open-Source Tool Funes, Providing Local Memory Layers for Coding Agents - Aioga AI News","description":"Hugging Face has released the open-source tool Funes, offering local memory layers for coding agents like Claude Code, Codex, pi, and Hermes. It indexes existing conversation recor...","url":"https://www.aioga.com/en/news/cmtlg0ht406hlrow5clznd0ld/","contentTranslated":true,"sourceHash":"a7fce6d15e609b67","translatedAt":"2026-09-03T12:02:28.162Z"},"ja":{"title":"Hugging Faceがオープンソースツールfunesを公開、エージェントのローカルメモリ層を提供","summary":"Hugging Faceはオープンソースツールfunesを公開し、Claude Code、Codex、pi、Hermesなどのコーディングエージェントにローカルメモリ層を提供する。既存の会話記録をLanceデータセットとしてインデックス化し、単一のfunes addコマンドでエージェントが元の出典（エージェント、タイムスタンプ、会話、ラウンド）を自主的に呼び出せる。","category":"業界動向","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Hugging Faceがオープンソースツールfunesを公開、エージェントのローカルメモリ層を提供 - Aioga AIニュース","description":"Hugging Faceはオープンソースツールfunesを公開し、Claude Code、Codex、pi、Hermesなどのコーディングエージェントにローカルメモリ層を提供する。既存の会話記録をLanceデータセットとしてインデックス化し、単一のfunes addコマンドでエージェントが元の出典（エージェント、タイムスタンプ、会話、ラウンド）を自主的に呼び...","url":"https://www.aioga.com/ja/news/cmtlg0ht406hlrow5clznd0ld/","contentTranslated":true,"sourceHash":"a7fce6d15e609b67","translatedAt":"2026-09-03T12:02:29.337Z"},"ko":{"title":"Hugging Face가 오픈소스 도구 funes를 발표, 코딩 에이전트에 로컬 기억 계층 제공","summary":"Hugging Face가 오픈소스 도구 funes를 발표하여 Claude Code, Codex, pi, Hermes 등 코딩 에이전트에 로컬 메모리 계층을 제공하며, 기존 대화 기록을 Lance 데이터셋으로 인덱싱할 수 있습니다. 단일 funes add 명령으로 에이전트가 원본 출처(에이전트, 타임스탬프, 대화, 라운드)를 자율적으로 회상할 수 있습니다.","category":"업계 동향","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Hugging Face가 오픈소스 도구 funes를 발표, 코딩 에이전트에 로컬 기억 계층 제공 - Aioga AI 뉴스","description":"Hugging Face가 오픈소스 도구 funes를 발표하여 Claude Code, Codex, pi, Hermes 등 코딩 에이전트에 로컬 메모리 계층을 제공하며, 기존 대화 기록을 Lance 데이터셋으로 인덱싱할 수 있습니다. 단일 funes add 명령으로 에이전트가 원본 출처(에이전트, 타임스탬프, 대화, 라운드)...","url":"https://www.aioga.com/ko/news/cmtlg0ht406hlrow5clznd0ld/","contentTranslated":true,"sourceHash":"a7fce6d15e609b67","translatedAt":"2026-09-03T12:02:36.191Z"},"es":{"title":"Hugging Face lanza la herramienta de código abierto funes, proporcionando una capa de memoria local para agentes de codificación","summary":"Hugging Face lanza la herramienta de código abierto funes, ofreciendo una capa de memoria local para agentes de codificación como Claude Code, Codex, pi, Hermes, indexando los registros de conversaciones existentes en conjuntos de datos Lance. Un solo comando `funes add` permite que el agente recupere de manera autónoma la fuente original (Agente, marca de tiempo, conversación, ronda).","category":"Industria","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Hugging Face lanza la herramienta de código abierto funes, proporcionando una capa de memoria local para agentes de codificación - Aioga Noticias de IA","description":"Hugging Face lanza la herramienta de código abierto funes, ofreciendo una capa de memoria local para agentes de codificación como Claude Code, Codex, pi, Hermes, indexando los regi...","url":"https://www.aioga.com/es/news/cmtlg0ht406hlrow5clznd0ld/","contentTranslated":true,"sourceHash":"a7fce6d15e609b67","translatedAt":"2026-09-03T12:02:36.546Z"},"fr":{"title":"Hugging Face publie l'outil open source funes, fournissant une couche de mémoire locale pour les agents de codage","summary":"Hugging Face publie l'outil open source funes, qui fournit une couche de mémoire locale pour les agents de codage tels que Claude Code, Codex, pi, Hermes. Il indexe les enregistrements de conversation existants en tant que jeu de données Lance, et une seule commande funes add permet à l'agent de rappeler de manière autonome la source originale (Agent, horodatage, conversation, tour).","category":"Industrie","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Hugging Face publie l'outil open source funes, fournissant une couche de mémoire locale pour les agents de codage - Aioga Actualités IA","description":"Hugging Face publie l'outil open source funes, qui fournit une couche de mémoire locale pour les agents de codage tels que Claude Code, Codex, pi, Hermes. Il indexe les enregistrem...","url":"https://www.aioga.com/fr/news/cmtlg0ht406hlrow5clznd0ld/","contentTranslated":true,"sourceHash":"a7fce6d15e609b67","translatedAt":"2026-09-03T12:02:42.977Z"},"de":{"title":"Hugging Face veröffentlicht das Open-Source-Tool funes, das Codierungsagenten eine lokal gehaltene Speicherebene bietet","summary":"Hugging Face veröffentlicht das Open-Source-Tool funes, das Claude Code, Codex, pi, Hermes und anderen Codierungsagenten eine lokale Speicherebene bietet. Bestehende Gesprächsaufzeichnungen werden in einen Lance-Datensatz indexiert, und ein einfacher Befehl \"funes add\" ermöglicht es dem Agenten, die ursprüngliche Quelle (Agent, Zeitstempel, Sitzung, Runde) selbstständig abzurufen.","category":"行业动态","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Hugging Face veröffentlicht das Open-Source-Tool funes, das Codierungsagenten eine lokal gehaltene Speicherebene bietet - Aioga KI-News","description":"Hugging Face veröffentlicht das Open-Source-Tool funes, das Claude Code, Codex, pi, Hermes und anderen Codierungsagenten eine lokale Speicherebene bietet. Bestehende Gesprächsaufze...","url":"https://www.aioga.com/de/news/cmtlg0ht406hlrow5clznd0ld/","contentTranslated":true,"sourceHash":"a7fce6d15e609b67","translatedAt":"2026-09-03T12:02:42.988Z"},"pt-BR":{"title":"Hugging Face lança ferramenta open source funes, fornecendo uma camada de memória local para agentes de programação","summary":"Hugging Face lança ferramenta open source funes, fornecendo uma camada de memória local para agentes de programação como Claude Code, Codex, pi, Hermes, indexando registros de conversas existentes em datasets Lance. Um único comando funes add permite que o Agente recupere de forma autônoma a fonte original (Agente, timestamp, conversa, rodada).","category":"行业动态","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Hugging Face lança ferramenta open source funes, fornecendo uma camada de memória local para agentes de programação - Aioga Notícias de IA","description":"Hugging Face lança ferramenta open source funes, fornecendo uma camada de memória local para agentes de programação como Claude Code, Codex, pi, Hermes, indexando registros de conv...","url":"https://www.aioga.com/pt-BR/news/cmtlg0ht406hlrow5clznd0ld/","contentTranslated":true,"sourceHash":"a7fce6d15e609b67","translatedAt":"2026-09-03T12:02:50.966Z"},"ru":{"title":"Hugging Face выпустила открытый инструмент funes, предоставляющий локально сохраняемый слой памяти для программистических агентов","summary":"Hugging Face выпустила инструмент с открытым исходным кодом funes, предоставляющий локальный слой памяти для программных агентов, таких как Claude Code, Codex, pi и Hermes, индексируя существующие сессионные записи в набор данных Lance. Всего одной командой добавления фунеса агент может автономно отзвать исходный исходный код (агент, временная метка, сессия, ротация).","category":"行业动态","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Hugging Face выпустила открытый инструмент funes, предоставляющий локально сохраняемый слой памяти для программистических агентов - Aioga Новости ИИ","description":"Hugging Face выпустила инструмент с открытым исходным кодом funes, предоставляющий локальный слой памяти для программных агентов, таких как Claude Code, Codex, pi и Hermes, индекси...","url":"https://www.aioga.com/ru/news/cmtlg0ht406hlrow5clznd0ld/","contentTranslated":true,"sourceHash":"a7fce6d15e609b67","translatedAt":"2026-09-03T12:02:51.918Z"},"ar":{"title":"أصدرت Hugging Face أداة مفتوحة المصدر funes لتوفير طبقة ذاكرة يمكن الاحتفاظ بها محليًا للوكلاء البرمجيين","summary":"أصدرت Hugging Face أداة مفتوحة المصدر funes لتوفير طبقة ذاكرة محلية لوكلاء الترميز مثل Claude Code و Codex و pi و Hermes، حيث تقوم بفهرسة سجلات المحادثات الحالية كبيانات Lance، وأمر واحد فقط funes add يسمح للوكيل باسترجاع المصدر الأصلي تلقائيًا (الوكيل، الطابع الزمني، المحادثة، الجولة).","category":"行业动态","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"أصدرت Hugging Face أداة مفتوحة المصدر funes لتوفير طبقة ذاكرة يمكن الاحتفاظ بها محليًا للوكلاء البرمجيين - Aioga أخبار الذكاء الاصطناعي","description":"أصدرت Hugging Face أداة مفتوحة المصدر funes لتوفير طبقة ذاكرة محلية لوكلاء الترميز مثل Claude Code و Codex و pi و Hermes، حيث تقوم بفهرسة سجلات المحادثات الحالية كبيانات Lance، وأم...","url":"https://www.aioga.com/ar/news/cmtlg0ht406hlrow5clznd0ld/","contentTranslated":true,"sourceHash":"a7fce6d15e609b67","translatedAt":"2026-09-03T12:02:58.316Z"},"hi":{"title":"हगिंग फेस ने ओपन-सोर्स टूल फनेस जारी किया, जो कोडिंग एजेंटों के लिए स्थानीय रूप से बनाए रखने योग्य मेमोरी परत प्रदान करता है","summary":"हगिंग फेस ने ओपन-सोर्स टूल फनेस जारी किया, जो क्लाउड कोड, कोडेक्स, पाई और हेमीज़ जैसे कोडिंग एजेंटों के लिए एक स्थानीय मेमोरी परत प्रदान करता है, जो मौजूदा सत्र रिकॉर्ड को लांस डेटासेट में अनुक्रमित करता है। केवल एक फनेस ऐड कमांड के साथ, एजेंट स्वायत्त रूप से मूल स्रोत (एजेंट, टाइमस्टैम्प, सत्र, रोटेशन) को याद कर सकता है।","category":"行业动态","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"हगिंग फेस ने ओपन-सोर्स टूल फनेस जारी किया, जो कोडिंग एजेंटों के लिए स्थानीय रूप से बनाए रखने योग्य मेमोरी परत प्रदान करता है - Aioga AI समाचार","description":"हगिंग फेस ने ओपन-सोर्स टूल फनेस जारी किया, जो क्लाउड कोड, कोडेक्स, पाई और हेमीज़ जैसे कोडिंग एजेंटों के लिए एक स्थानीय मेमोरी परत प्रदान करता है, जो मौजूदा सत्र रिकॉर्ड को लांस डेट...","url":"https://www.aioga.com/hi/news/cmtlg0ht406hlrow5clznd0ld/","contentTranslated":true,"sourceHash":"a7fce6d15e609b67","translatedAt":"2026-09-03T12:03:00.769Z"},"it":{"title":"Hugging Face rilascia lo strumento open source funes, fornendo un layer di memoria locale per agenti di codifica","summary":"Hugging Face rilascia lo strumento open source funes, fornendo un layer di memoria locale per agenti di codifica come Claude Code, Codex, pi, Hermes, indicizzando le conversazioni esistenti in dataset Lance; un singolo comando funes add consente all'Agent di richiamare autonomamente le fonti originali (Agent, timestamp, conversazione, turno).","category":"行业动态","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Hugging Face rilascia lo strumento open source funes, fornendo un layer di memoria locale per agenti di codifica - Aioga Notizie IA","description":"Hugging Face rilascia lo strumento open source funes, fornendo un layer di memoria locale per agenti di codifica come Claude Code, Codex, pi, Hermes, indicizzando le conversazioni...","url":"https://www.aioga.com/it/news/cmtlg0ht406hlrow5clznd0ld/","contentTranslated":true,"sourceHash":"a7fce6d15e609b67","translatedAt":"2026-09-03T12:03:07.646Z"},"nl":{"title":"Hugging Face brengt open-source tool funes uit, die een lokaal geheugenlaag voor codeeragentschappen biedt","summary":"Hugging Face heeft de open-source tool funes uitgebracht, die voor Claude Code, Codex, pi, Hermes en andere codeeragentschappen een lokaal geheugenlaag biedt, waarbij bestaande gesprekslogboeken worden geïndexeerd als Lance-dataset. Een enkele funes add-commando kan de Agent in staat stellen de oorspronkelijke bron (Agent, tijdstempel, gesprek, ronde) zelfstandig terug te halen.","category":"行业动态","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Hugging Face brengt open-source tool funes uit, die een lokaal geheugenlaag voor codeeragentschappen biedt - Aioga AI-nieuws","description":"Hugging Face heeft de open-source tool funes uitgebracht, die voor Claude Code, Codex, pi, Hermes en andere codeeragentschappen een lokaal geheugenlaag biedt, waarbij bestaande ges...","url":"https://www.aioga.com/nl/news/cmtlg0ht406hlrow5clznd0ld/","contentTranslated":true,"sourceHash":"a7fce6d15e609b67","translatedAt":"2026-09-03T12:03:07.612Z"},"tr":{"title":"Hugging Face, kodlama ajanlarına yerel olarak tutulabilecek bir bellek katmanı sağlayan açık kaynak aracı funes'i yayınladı","summary":"Hugging Face, Claude Code, Codex, pi, Hermes gibi kodlama ajanlarına yerel bellek katmanı sağlayan açık kaynak aracı funes'i yayınladı; mevcut sohbet kayıtlarını Lance veri setine dizinleyebiliyor ve bir funes add komutu ile Ajan'ın orijinal kaynağı (Ajan, zaman damgası, sohbet, tur) kendi kendine geri çağırmasını sağlıyor.","category":"行业动态","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Hugging Face, kodlama ajanlarına yerel olarak tutulabilecek bir bellek katmanı sağlayan açık kaynak aracı funes'i yayınladı - Aioga AI Haberleri","description":"Hugging Face, Claude Code, Codex, pi, Hermes gibi kodlama ajanlarına yerel bellek katmanı sağlayan açık kaynak aracı funes'i yayınladı; mevcut sohbet kayıtlarını Lance veri setine...","url":"https://www.aioga.com/tr/news/cmtlg0ht406hlrow5clznd0ld/","contentTranslated":true,"sourceHash":"a7fce6d15e609b67","translatedAt":"2026-09-03T12:03:14.263Z"},"vi":{"title":"Hugging Face phát hành công cụ mã nguồn mở funes, cung cấp lớp bộ nhớ có thể lưu trữ cục bộ cho các tác nhân lập trình","summary":"Hugging Face phát hành công cụ mã nguồn mở funes, cung cấp lớp bộ nhớ cục bộ cho các tác nhân lập trình như Claude Code, Codex, pi, Hermes, chuyển các bản ghi hội thoại hiện có thành bộ dữ liệu Lance, chỉ với một lệnh funes add là Agent có thể tự truy hồi nguồn gốc gốc (Agent, dấu thời gian, hội thoại, vòng).","category":"行业动态","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Hugging Face phát hành công cụ mã nguồn mở funes, cung cấp lớp bộ nhớ có thể lưu trữ cục bộ cho các tác nhân lập trình - Tin tức AI Aioga","description":"Hugging Face phát hành công cụ mã nguồn mở funes, cung cấp lớp bộ nhớ cục bộ cho các tác nhân lập trình như Claude Code, Codex, pi, Hermes, chuyển các bản ghi hội thoại hiện có thà...","url":"https://www.aioga.com/vi/news/cmtlg0ht406hlrow5clznd0ld/","contentTranslated":true,"sourceHash":"a7fce6d15e609b67","translatedAt":"2026-09-03T12:03:14.159Z"},"id":{"title":"Hugging Face merilis alat open source funes, menyediakan lapisan memori lokal untuk agen pengkodean","summary":"Hugging Face merilis alat open source funes, menyediakan lapisan memori lokal untuk agen pengkodean seperti Claude Code, Codex, pi, Hermes, dengan mengindeks catatan percakapan yang ada menjadi dataset Lance. Satu perintah funes add cukup untuk membuat Agent dapat memanggil kembali sumber asli secara mandiri (Agent, stempel waktu, percakapan, putaran).","category":"行业动态","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Hugging Face merilis alat open source funes, menyediakan lapisan memori lokal untuk agen pengkodean - Berita AI Aioga","description":"Hugging Face merilis alat open source funes, menyediakan lapisan memori lokal untuk agen pengkodean seperti Claude Code, Codex, pi, Hermes, dengan mengindeks catatan percakapan yan...","url":"https://www.aioga.com/id/news/cmtlg0ht406hlrow5clznd0ld/","contentTranslated":true,"sourceHash":"a7fce6d15e609b67","translatedAt":"2026-09-03T12:03:21.301Z"},"th":{"title":"Hugging Face เปิดเครื่องมือโอเพนซอร์ส funes สำหรับให้เอเจนต์โค้ดมีชั้นความจำที่สามารถเก็บไว้ในเครื่องได้","summary":"Hugging Face เปิดเครื่องมือโอเพนซอร์ส funes สำหรับ Claude Code, Codex, pi, Hermes และเอเจนต์โค้ดอื่น ๆ เพื่อให้มีชั้นความจำในเครื่อง สามารถจัดทำดัชนีบันทึกการสนทนาที่มีอยู่เป็นชุดข้อมูล Lance เพียงใช้คำสั่ง funes add ครั้งเดียวก็สามารถให้ Agent เรียกคืนต้นทางเดิมได้เอง (Agent, เวลา, การสนทนา, รอบ)","category":"行业动态","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Hugging Face เปิดเครื่องมือโอเพนซอร์ส funes สำหรับให้เอเจนต์โค้ดมีชั้นความจำที่สามารถเก็บไว้ในเครื่องได้ - ข่าว AI Aioga","description":"Hugging Face เปิดเครื่องมือโอเพนซอร์ส funes สำหรับ Claude Code, Codex, pi, Hermes และเอเจนต์โค้ดอื่น ๆ เพื่อให้มีชั้นความจำในเครื่อง สามารถจัดทำดัชนีบันทึกการสนทนาที่มีอยู่เป็นชุดข...","url":"https://www.aioga.com/th/news/cmtlg0ht406hlrow5clznd0ld/","contentTranslated":true,"sourceHash":"a7fce6d15e609b67","translatedAt":"2026-09-03T12:03:20.898Z"},"pl":{"title":"Hugging Face wydało otwartoźródłowe narzędzie funes, zapewniające lokalnie zachowywalną warstwę pamięci dla agentów kodujących","summary":"Hugging Face wydało otwartoźródłowe narzędzie funes, zapewniające lokalną warstwę pamięci dla agentów kodujących takich jak Claude Code, Codex, pi i Hermes, indeksując istniejące rekordy sesji do zbioru danych Lance. Za pomocą jednego polecenia funes add agent może autonomicznie odwołać oryginalne źródło (Agent, znacznik czasu, sesja, rotacja).","category":"行业动态","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Hugging Face wydało otwartoźródłowe narzędzie funes, zapewniające lokalnie zachowywalną warstwę pamięci dla agentów kodujących - Aioga Wiadomości AI","description":"Hugging Face wydało otwartoźródłowe narzędzie funes, zapewniające lokalną warstwę pamięci dla agentów kodujących takich jak Claude Code, Codex, pi i Hermes, indeksując istniejące r...","url":"https://www.aioga.com/pl/news/cmtlg0ht406hlrow5clznd0ld/","contentTranslated":true,"sourceHash":"a7fce6d15e609b67","translatedAt":"2026-09-03T12:03:30.547Z"}},"evidenceTier":"verified-news","reviewStatus":"editorial-selected","indexable":true,"editorialCover":""}}