正在加载视频...
视频加载失败
Introducing Blume 2.0 The open-source docs framework for humans and agents. → 18 analytics integrations → Contentful, Payload + Strapi → Adapters for search, AI + deployment → Composable API references → Hundreds of fixes and improvements Upgrade now with 𝚗𝚙𝚡 𝚋𝚕𝚞𝚖𝚎@𝚕𝚊𝚝𝚎𝚜𝚝 𝚞𝚙𝚐𝚛𝚊𝚍𝚎 --𝚌𝚘𝚍𝚎𝚡 What's new ↓
56,678 次观看 • 23 天前 •via X (Twitter)
38 条评论

Keep your content where you write it. New Contentful, Payload and Strapi adapters join Markdown, remote MDX, GitHub Releases, Notion, Sanity and Obsidian. Mix sources in one site, with rich text and CMS draft support.

Analytics now has 18 built-in provider adapters, including PostHog, Plausible, Google Analytics, Mixpanel, Amplitude and Clarity. Combine providers in one list. Their native options pass through, so you can keep the settings you already use.

Search and Ask AI now use adapters too. Choose local search with Orama, FlexSearch or Pagefind, or hosted search like Algolia. For Ask AI, use AI Gateway, OpenRouter, LLM Gateway, Inkeep or your own OpenAI-compatible endpoint.

API references are composable. Mix OpenAPI, AsyncAPI, GraphQL and embedded Scalar references in one list. Add multiple references of the same kind, each with its own route. Your public API, internal API and event schemas can live in the same docs site.

Deployment follows the same pattern. Pick a Vercel, Netlify, Cloudflare or Node adapter, with the underlying Astro adapter's options available. Leave deployment unset and Blume still builds a static site you can host anywhere.

Each adapter declares the dependencies and secrets it needs. 𝚋𝚕𝚞𝚖𝚎 𝚍𝚘𝚌𝚝𝚘𝚛 can read back your stack, and the generated project, secrets checks and ejected Astro app all use those same declarations. Less configuration to keep in sync.

Your docs just got easier for agents to read. Accordions, file trees, code groups and tooltips become readable Markdown in .md pages, llms-full.txt and MCP responses. Rich components for readers. The same content for agents.

Moving from Mintlify, Fumadocs, Docusaurus, Starlight or Nextra? $ 𝚗𝚙𝚡 𝚋𝚕𝚞𝚖𝚎 𝚖𝚒𝚐𝚛𝚊𝚝𝚎 --𝚌𝚘𝚍𝚎𝚡 Blume detects your framework and opens Codex with a migration playbook covering config, content and build checks.

Upgrading from Blume 1? $ 𝚗𝚙𝚡 𝚋𝚕𝚞𝚖𝚎@𝚕𝚊𝚝𝚎𝚜𝚝 𝚞𝚙𝚐𝚛𝚊𝚍𝚎 --𝚌𝚘𝚍𝚎𝚡 Installs Blume 2, finds remaining changes by file and line, then opens Codex to apply them and run doctor and build checks.

Start a new docs site: 𝚗𝚙𝚡 𝚋𝚕𝚞𝚖𝚎 𝚒𝚗𝚒𝚝 It now installs dependencies with your package manager too. Upgrade with Codex: 𝚗𝚙𝚡 𝚋𝚕𝚞𝚖𝚎@𝚕𝚊𝚝𝚎𝚜𝚝 𝚞𝚙𝚐𝚛𝚊𝚍𝚎 --𝚌𝚘𝚍𝚎𝚡 Docs:

As always, free and open source.

I hope you're enjoying Blume! Follow me @haydenbleasel for more updates. Also, Like/Repost the quote below if you can:

lfg 🥹 i probably knew its coming soon one of the first people when was migrating our docs to Blume today and saw the 2.0

thank you so much this! @haydenbleasel

LFG! Fire launch video too

This looks gooood!

this is dope and neat 🙌

改动不少,老项目升级会不会有坑?

不,应该完全无缝。如果您遇到任何问题,请告诉我,我会进行调查。

我建议使用 npx blume@latest upgrade --codex

FIRE

Ohhhh baby

docs that agents can actually navigate feels overdue

I love Blume so much man

Blume is beauty.

🔥💣

Great !! I love Blume so much !

I really need to try it for my next OSS project 👀

yo nice!! i'm about to migrate all my projects with blume

wait blume start 2.0 fr

congrats on the release!!

Docs written for humans and agents makes sense. How do the CMS adapters handle drafts? If an agent writes a page into Contentful as a draft, can Blume preview it before someone publishes? That handoff is where we keep getting stuck.

Yep Contentful drafts work through preview. Run blume dev --preview (or blume build --preview) and it pulls unpublished entries via the Preview API with CONTENTFUL_PREVIEW_TOKEN. Production builds without --preview stay published-only, so that handoff is intentional. Same pattern for Payload, Strapi, and Notion.

> 18 analytics integrations no databuddy im offended

Put a PR up and I will merge it tonight

already broke my heart

the mixed API references caught my eye. can search and Ask AI work across OpenAPI, AsyncAPI and GraphQL in the same site too?

docs for agents is the right framing. what breaks is stale examples nothing ever marks as stale

