Clelia Astra Bertelli

The anatomy of a document processing agent Exploring the building blocks of LobsterX, a document-processing agent inspired by OpenClaw

📦 Install: 𝘶𝘷 𝘵𝘰𝘰𝘭 𝘪𝘯𝘴𝘵𝘢𝘭𝘭 𝘭𝘰𝘣𝘴𝘵𝘦𝘳𝘹 --𝘱𝘳𝘦𝘳𝘦𝘭𝘦𝘢𝘴𝘦=𝘢𝘭𝘭𝘰𝘸
📚 Read more: github.com/AstraBert/w...
📝 Read the blog: clelia.dev/2026-02-09-...

3. The agent is self-hostable and can be used both as a Docker image and as a uv tool, and is available as a Telegram bot💬

2. The agent has only access to a virtual filesystem (AgentFS by @tur.so) to avoid damaging your real one, and cannot execute arbitrary bash commands, preventing it from performing dangerous or security-critical operations🔒

1. LobsterX uses @llamaindex.bsky.social cloud products to parse, extract structured data and classify files

The tech world went crazy for @openclaw, so I decided to build a similar crustacean agent called LobsterX🦞, with a focus on document-processing tasks.
I wrote a blog about it: clelia.dev/2026-02-09-..., but, as a tl;dr:

Everybody’s talking about OpenClaw, so @cle-does-things.bsky.social decided to build her own crustacean AI assistant for document workflows: LobsterX 🦞

GitHub - AstraBert/multipilot: Orchestration layer to run multiple Copilot tasks concurrently with Temporal Workflows Orchestration layer to run multiple Copilot tasks concurrently with Temporal Workflows - AstraBert/multipilot

You can visualize the progress (or the session, once it's finished), through ad-hoc HTML rendering of logs in the browser, all packaged within the same easy-to-use CLI tool!🎨
Give it a try 👉 github.com/AstraBert/m... (it's still in alpha, feedback is highly appreciated✨✨)

Built on top of the Copilot SDK in @golang.org, multipilot leverages @temporal.io workflows to spawn and manage several Copilot sessions on different working directories in a resilient and durable way, keeping track of everything through JSONL log files👩‍💻

Hey, I just released 𝗺𝘂𝗹𝘁𝗶𝗽𝗶𝗹𝗼𝘁, an orchestration tool to run multiple @github.com Copilot sessions concurrently!👋

Skills vs MCP tools for agents: when to use what LlamaIndex is a simple, flexible framework for building knowledge assistants using LLMs connected to your enterprise data.

The main takeaway? Skills are easier to set up but tougher to keep up-to-date, whereas MCPs are more dev-facing, but offer a more deterministic toolset with less maintenance overhead.
Check out the article 👉 www.llamaindex.ai/blog/skills...

Are you choosing between MCP servers and skills for your agent?
@tuana.dev and I wrote a blog post about the differences between the two and their PROs and CONs, informed by the lessons learnt while building our own coding agent (LlamaAgents Builder).

GitHub - AstraBert/PdfItDown: Convert Everything to PDF Convert Everything to PDF. Contribute to AstraBert/PdfItDown development by creating an account on GitHub.

📦 Install with: 𝘶𝘷 𝘵𝘰𝘰𝘭 𝘪𝘯𝘴𝘵𝘢𝘭𝘭 𝘱𝘥𝘧𝘪𝘵𝘥𝘰𝘸𝘯-𝘮𝘤𝘱-𝘴𝘦𝘳𝘷𝘦𝘳
📚 Take a look at the docs: pdfitdown.eu
⭐ Star the repo on GitHub: github.com/AstraBert/P...

PdfItDown, the python library I created to convert (almost) any file format to PDF, has now an MCP server!🚀
Built on top of FastMCP and running over stdio transport, PdfItDown MCP server allows your agents to seamlessly convert one or more files to PDF in one go!

Why don't I vibe code more? Yapping about my experience with vibe-coding

Here is the blog that came out of my self-reflection: clelia.dev/2026-01-31-...

While the world seems to be going crazy for OpenClaw/MoltBot (great work @steipete.me btw) and agents having their own Reddit, I decided to take a step back and reflect on why I don't vibe-code more (for those who don't know me: hi, I'm cle✨ and I hand-write approx. 70-80% of my code).

GitHub - AstraBert/scpr: Web scraper CLI and MCP built for human and coding agents Web scraper CLI and MCP built for human and coding agents - AstraBert/scpr

GitHub repo 👉 github.com/AstraBert/scpr
Download 👉 𝘯𝘱𝘮 𝘪𝘯𝘴𝘵𝘢𝘭𝘭 @𝘤𝘭𝘦-𝘥𝘰𝘦𝘴-𝘵𝘩𝘪𝘯𝘨𝘴/𝘴𝘤𝘱𝘳

Clelia Astra Bertelli — Open Source Engineer and passionate AI developer 🚀 Open Source Engineer and passionate AI developer

Take a look at the demo, where I use scpr to scrape content from one of my blog entries and all crossed-linked pages on my domain (clelia.dev, in total 14 pages) in 0.2 seconds☝️

I built a small web-scraping CLI app with @golang🌐
It's called 𝘀𝗰𝗽𝗿, and allows you to:
🔄 Scrape pages recursively using concurrent tasks
🔒 Visit only trusted domains
📁 Download scraped pages as markdown
🤖 Interact with your coding agents, since it comes with an 𝗠𝗖𝗣 𝘀𝗲𝗿𝘃𝗲𝗿 𝗮𝗻𝗱 𝘀𝗸𝗶𝗹𝗹𝘀

We updated our contributions guide with guidelines on how to use AI to propose changes to the @llamaindex.bsky.social main repository: github.com/run-llama/l...
tl;dr: you can use AI, but you need to understand what you're creating and be able to maintain it long-term

We're launching LlamaAgents Builder today—a new way to build document processing agents 🔥

On this Holocaust Remembrance Day, my heart goes out to all humans fighting systemic oppression, from Gaza to the United States, Ukraine, Yemen, Sudan, Syria, Iran, and countless other places.
May the suffering of those who came before us help us avoid the same horrors today🫂☮️

GitHub - AstraBert/rag-rs: A Rust-native implementation of the RAG stack A Rust-native implementation of the RAG stack. Contribute to AstraBert/rag-rs development by creating an account on GitHub.

Or find it on GitHub 👉 github.com/AstraBert/r...

Just released the first stable version of 𝗿𝗮𝗴-𝗿𝘀, my Rust native implementation of the RAG stack🦀
The package now has a solid CI pipeline, with thorough testing and linting🚀
You can install v0.2.1 via cargo and npm:
𝘤𝘢𝘳𝘨𝘰 𝘪𝘯𝘴𝘵𝘢𝘭𝘭 𝘳𝘢𝘨-𝘳𝘴
𝘯𝘱𝘮 𝘪𝘯𝘴𝘵𝘢𝘭𝘭 @𝘤𝘭𝘦-𝘥𝘰𝘦𝘴-𝘵𝘩𝘪𝘯𝘨𝘴/𝘳𝘢𝘨-𝘳𝘴

GitHub - AstraBert/codemd: Command line tool to extract code from markdown files Command line tool to extract code from markdown files - AstraBert/codemd

Make sure to upgrade:
𝘤𝘢𝘳𝘨𝘰 𝘪𝘯𝘴𝘵𝘢𝘭𝘭 𝘤𝘰𝘥𝘦𝘮𝘥@0.2.1
𝘯𝘱𝘮 𝘪𝘯𝘴𝘵𝘢𝘭𝘭 @𝘤𝘭𝘦-𝘥𝘰𝘦𝘴-𝘵𝘩𝘪𝘯𝘨𝘴/𝘤𝘰𝘥𝘦𝘮𝘥@𝘭𝘢𝘵𝘦𝘴𝘵
And leave a little ⭐ on GitHub: github.com/AstraBert/c...

Shipped a small bugfix to codemd, my Rust-based tool to extract and run code from markdown🦀
The regex I had in place wasn't correctly capturing code blocks with backticks (which are fairly common il languages like TS or Go), with v0.2.1 that should be fixed!👩‍💻

Just surpassed 3k contributions on @github.com in the past year!
btw, if you're looking for ways to support my OSS work 👉 github.com/sponsors/As...

Get started with the SDK:
𝘱𝘪𝘱 𝘪𝘯𝘴𝘵𝘢𝘭𝘭 𝘭𝘭𝘢𝘮𝘢-𝘤𝘭𝘰𝘶𝘥>=1
𝘯𝘱𝘮 𝘪 𝘭𝘭𝘢𝘮𝘢-𝘤𝘭𝘰𝘶𝘥@𝘭𝘢𝘵𝘦𝘴𝘵
Take a look at our n8n guide: run-llama.github.io/n8n-llamacloud

We also revamped the @n8n.io integration @tuana.dev and I built, with more stable nodes and extended support for LlamaClassify and LlamaSheets 🚀

At @llamaindex.bsky.social we just released our new SDK for LlamaCloud, which means:
- Cleaner, unified client interfaces
- Full synchronous and asynchronous support
- Smoother typing (-> better autocomplete!)
- Parity between Python and Typescript SDK

GitHub - AstraBert/rag-rs: A Rust-native implementation of the RAG stack A Rust-native implementation of the RAG stack. Contribute to AstraBert/rag-rs development by creating an account on GitHub.

Install the alpha release with 𝘤𝘢𝘳𝘨𝘰 𝘪𝘯𝘴𝘵𝘢𝘭𝘭 𝘳𝘢𝘨-𝘳𝘴@0.2.0-𝘢𝘭𝘱𝘩𝘢 and find the repo on GitHub 👉 github.com/AstraBert/r...