Coding agents might feel like magic, but they aren’t 🪄 That’s exactly why I’ve been eager lately to understand how they work under the hood, using the most natural approach I could think of: observability.

Coding agents shouldn’t be black boxes ❌ If you’re calling models through APIs and can configure provider settings (base URL, API key, etc.), I highly recommend putting a local instrumented API proxy in front of them. I built one for the Anthropic Messages API:

Do you actually own your document parsing infrastructure? 👀 At LlamaIndex, we wanted to make that easier, so we built 𝗹𝗶𝘁𝗲𝗽𝗮𝗿𝘀𝗲-𝘀𝗲𝗿𝘃𝗲𝗿, a lightweight HTTP backend built on top of LiteParse that can parse and generate page screenshots from PDFs, images, and Office documents🦙

I have one big problem with agentic engineering: I want agents to operate autonomously, but I also want granular, reversible control over every change they make. I could solve this by committing every intermediate step to Git, but that would completely pollute my repo history.

For 2026, I decided I'm going to build all of my side projects with Rust when possible🦀 So here's the first one: 𝗳𝗶𝗹𝗲-𝘀𝘁𝗼𝗿𝗮𝗴𝗲, a self-hostable file storage application built on a modern open source stack, with a full Rust backend and a @react.dev + @vite.dev frontend🚀 With file-storage, you get:

Scaling document pipelines is harder than it looks, and not because of tooling, but mostly because of orchestration⚙️ That's why I teamed up with @render.com and focused on building a system where each component does one job well, and scales independently:

The LlamaParse MCP got a new face, and it is now easier than ever to run document processing workflows from your agents🚀

I'll be speaking at the Applied AI Conf organized by Tech Europe!🇪🇺 On the 28th of May, I'll be in Berlin to give my talk about how you can build reliable document agents by leveraging harness engineering with the building blocks provided by @llamaindex.bsy.social 🦙

I'll be speaking at the upcoming Voxel51 event in Stuttgart this Tuesday!🚀 I'll talk about the anatomy of AI agents, with a focus on document agents and building good harnesses for them⚙️ Swing by if you're interested, and check the official page for more details: voxel51.com/events/stut...

I've been building sunbears, a typescript library for CSV data loading written in Rust🦀

ParseBench is here!📊 We’ve just released ParseBench, an open benchmark + dataset for evaluating document parsing at scale.

I'm building 𝘀𝘂𝗻𝗯𝗲𝗮𝗿𝘀, a CSV data loader library for TS written in Rust🦀 In Node, it can read a file with 1.000.000 rows in 0.3s, write the same amount of rows in 0.15s, respectively 4x and 2x faster than the `csv` package⚡

Visually rich documents are especially challenging for agents. Tables, charts, and images often break traditional document pipelines, making complex reasoning difficult📄 So we teamed up with LanceDB to build a structure-aware PDF QA pipeline🚀 Here’s how it works:

How can you improve your agentic search pipeline? I just wrote a blog post in collab with LanceDB to answer exactly that. TLDR:

At @llamaindex.bsky.social, we're committed to building the most capable document agents.

I just published a TypeScript library for loading CSV data, with an API inspired by Pandas and @pola.rs, but fully written in Rust 🦀

Our OSS engineer @cle-does-things.bsky.social recently built 𝗹𝗶𝘁𝗲𝘀𝗲𝗮𝗿𝗰𝗵, a fully local document ingestion and retrieval CLI/TUI application powered by LiteParse ⚡ litesearch demonstrates how developers can assemble a high-performance, local-first pipeline using tools from across the ecosystem:

Hey there 👋 , I built 𝗹𝗶𝘁𝗲𝘀𝗲𝗮𝗿𝗰𝗵, a fully local document ingestion and retrieval CLI and TUI app, powered by LiteParse⚡

🚀 New release: 𝗯𝗲𝘁𝘁𝗲𝗿-𝗿𝗲𝘀𝘂𝗹𝘁-𝗽𝘆 𝘃𝟭.𝟭.𝟬 Inspired by the better-result library in the JS ecosystem, I built better-result-py to bring a Rust-style Result type to Python. With v1.1.0, I introduced the 𝙾𝚔 and 𝙴𝚛𝚛 classes to make control flow clearer and more explicit:

The Google DeepMind team really cooked with Gemini 3.1 in the Live API: it's fast and the output quality is great🔥 That's why at @llamaindex.bsky.social we decided to test it out with our bread and butter: document processing📄

I created 𝘀𝗸𝗶𝗹𝗹𝘇𝘆, a simple CLI written in Rust for your agent skills📝 → 𝘴𝘬𝘪𝘭𝘭𝘻𝘺 𝘪𝘯𝘪𝘵 will allow you to create the frontmatter for a skill, also generating the required folders and skill file → 𝘴𝘬𝘪𝘭𝘭𝘻𝘺 𝘤𝘩𝘦𝘤𝘬 will validate your existing skills against the agentskills.io specification

𝗷𝗮𝗸𝗲, the Make-like task executor for Unix OS I wrote in Rust🦀, has now reached 𝘃𝟬.𝟳.𝟬! Here's what's new:

notion-cli, the app I built in Golang to interact with @notion.com pages from your terminal and with your agents, got to v0.3.0🚀 This version adds a search command, which allows your (or your agent's) flow to be smoother: search → read → modify.

So excited to see live the article Vishal and I wrote on LlamaParse x Google Gemini!

Improve document parsing accuracy by 15% for financial PDFs. Use LlamaParse and Gemini 3.1 Pro to extract high-quality data from unstructured brokerage statements and complex tables. 📈 Precise reasoning 📂 Structured PDF data ⚡️ Event-driven scaling Dive into the code on GitHub → goo.gle/4dCPjjd

A few people reached out asking if they could contribute to 𝗯𝗲𝘁𝘁𝗲𝗿-𝗿𝗲𝘀𝘂𝗹𝘁-𝗽𝘆, the library I created with a minimal, fully-typed and dependency-free implementation of the Rust Result type🦀

Inspired by @dillonis.online's better-result library for the JS ecosystem, I published 𝗯𝗲𝘁𝘁𝗲𝗿-𝗿𝗲𝘀𝘂𝗹𝘁-𝗽𝘆, a minimal library with the Rust Result type implementation for Python. It's fully typed, dependency-free, and supports both sync and async functions.

Today at @llamaindex.bsky.social we are launching 𝗟𝗶𝘁𝗲𝗣𝗮𝗿𝘀𝗲, a lightweight and open source CLI and TS library for document parsing🚀 We all know that document parsing pipelines have become complicated fast: different tools for different formats, many libraries, failures in structural detection.

I've been seeing some users that joined GitHub very recently (4-7 days ago) having dozens of opened PRs, most of which, in my experience, are duplicates and/or clearly AI slop: could someone from @github.com explain to me how/why aren't these users getting automatically blocked by the platform?

SqUIte, the GUI app I wrote for SQLite in Rust🦀, just reached v0.2.0 :) Now it supports all SQL statements, with a caveat: if you want to modify data, you have to acknowledge it by ticking the dedicated checkbox, so that you never accidentally run non-readonly queries🔐

I just released v0.5.0 of jake, my Rust-based alternative for Make🦀 Here's what's new:

Gemini 2 Embedding dropped today and at @llamaindex.bsky.social we built a demo to showcase how you can integrate it with our ecosystem, from LlamaParse to LlamaAgents. It's called audio-kb, a local knowledge base for your audio notes:

Lol, agents opening a PR and pinging me for a review not event a minute after🙈 At least they said "please" and "thanks"🤷‍♀️ Seriously, tho, this should stop🫠

At @llama_index we built Surreal Slides, an agentic tool to manage and search your slides and presentations directly from the terminal. Here’s how it works:

I built 𝗻𝗼𝘁𝗶𝗼𝗻-𝗰𝗹𝗶 over the weekend: a lightweight CLI tool (and agent skill) to 𝗿𝗲𝗮𝗱, 𝘄𝗿𝗶𝘁𝗲, 𝗮𝗻𝗱 𝗺𝗼𝗱𝗶𝗳𝘆 𝗡𝗼𝘁𝗶𝗼𝗻 𝗽𝗮𝗴𝗲𝘀 𝘄𝗶𝘁𝗵𝗼𝘂𝘁 𝗹𝗲𝗮𝘃𝗶𝗻𝗴 𝘁𝗵𝗲 𝘁𝗲𝗿𝗺𝗶𝗻𝗮𝗹. Built in Go, it’s fast, simple, and works for both humans and AI agents.

Happy International Women's Day to all my sisters out there!♀️🏳️‍🌈🏳️‍⚧️ I see and I celebrate all of you, and that's what we should do every day, empowering this and the next generations of women through support, equal opportunities and visibility💜✨

In these times of war and uncertainty, I stumbled across an old song by an Italian rapper that I used to enjoy some years ago.

jake, the Make alternative I built with Rust, just reached v0.4.0 🦀 Here’s what’s new: - Execute composite commands using operators like && and | 👩‍💻 - Run tasks from any subdirectory where jakefile.toml is defined ⚙️ There’s also a brand-new documentation website: jake-build.com 👀

🎥 Watch the full video here: 📘 Or get started right away with the docs (UI + code examples): developers.llamaindex.ai/python/clou...

Rust and Go are often compared across many dimensions. Since I’ve been working with both languages recently, I decided to put them to the test by building the same project in each: an in-memory, single-node KV store📁 I focused on comparing developer experience and performance in my own use case👩‍💻

New Rust project: 𝗦𝗾𝗨𝗜𝘁𝗲, a UI for SQLite🦀 Having heard some pretty interesting things about Rust's desktop ecosystem, I decided to give 𝗲𝗴𝘂𝗶 a try, building a lightweight desktop app that allows users to visualize results from SQL queries as tables✨

New blog post alert!🦙✨ At @llamaindex.bsky.social we shipped LlamaAgents Builder to make it the fastest way to build a document agent only with natural language.

I've been seeing an alarming pattern in OSS GitHub communities lately: people just scraping issues, pinging maintainers with urgency to get the issues assigned to them and producing PRs (mostly through coding agents) within minutes of maintainers agreeing to assign the issue.

Hey Rustacean friends!🦀 This Friday I'll be talking about how we migrated to Qdrant Edge for the storage layer of Semtools, @llamaindex.bsky.social's Rust-powered CLI toolkit for local document intelligence🦙

We’ve been cooking at @llamaindex.bsky.social🍳 Almost a month ago, we launched LlamaAgents Builder, a natural language interface for building agentic document workflows📁 Today, we’re excited to announce file upload support: add your PDFs, and the agent will use them to build your use case🚀

Are you looking for a Make alternative?👀 Meet 𝗷𝗮𝗸𝗲, a task executor I wrote in Rust🦀 - Tasks are defined with TOML syntax - No .PHONY, no strict spacing rules - Tasks can depend on one another, and there are circular dependency checks - You can pass additional options to task when invoking them

🚀 Big drop from Google DeepMind: Gemini 3.1 Pro is here, and we built a hands-on demo powered by LlamaCloud to put it to work and turn your receipt photos into real financial insights!

Friendly reminder that pinging a maintainer on Discord more than once to ask a review for your PR or writing to their personal email (!) for the same purpose is not (and will never be) an acceptable behavior🤗

I built a Make-like CLI tool in Rust🦀 I noticed that my Makefiles mostly included simple commands, for whom Make syntax felt a tad too overengineered⚙️ That's why I created 𝗷𝗮𝗸𝗲 (that stands for "just make"):

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: