iiinova — where your digital self keeps evolving

Tuana @tuana.dev · 17/09/2026

Example coming soon and you can read the paper here: arxiv.org/pdf/2605.10616

Tuana @tuana.dev · 17/09/2026

It's motivated by MulTaBench from Alan Arazi & co: 40 datasets where combining the table and the image/text provably beats either alone. Turns out that's a real, common problem, and tabular models had no native answer to it. `pip install tabpfn-extensions[image]`

100

Tuana @tuana.dev · 17/09/2026

A new TabPFN extension is cooking (well, cooked actually) to handle images in tables: TabPFNWithImages

100

Tuana @tuana.dev · 15/09/2026

There's also TabPFN-3.5-Fast, Plus and Thinking Mode too. 🎉 We're also hosting a hackathon until the 6th of October!! Join the hackathon here: platrorm.priorlabs.ai/hackathon-3...

000

Tuana @tuana.dev · 15/09/2026

It adds +250 Elo points over the strongest previous baseline in BeyondArena for text-rich, high-cardinality and high-dimensional data. +52 Elo over the previous benchmark leader TabArena. With Thinking mode adding another +44 Elo to that!

100

Tuana @tuana.dev · 15/09/2026

We released TabPFN-3.5 today, and it's the next SOTA tabular foundation model now ranking first on TabArena and BeyondArena for 1M rows and up to 20k features.

100

Tuana @tuana.dev · 14/09/2026

I need to send you some more!!

010

Tuana @tuana.dev · 06/08/2026

I did a very loose experiment by asking Claude Code about how to build an agent that makes predictions over structured data. It pulled the actual numbers and picked best accuracy/latency compromise (TabPFN-3), then wired it in as a tool the LLM calls rather than a prediction the LLM makes itself.

101

Tuana @tuana.dev · 06/08/2026

Another step in getting tabular foundation models into the working knowledge of agents: @LennartPurucker built a TabArena MCP server, so an agent can query the leaderboards directly.

200

Tuana @tuana.dev · 03/08/2026

Tabular foundation models work as a prediction tool in agent workflows. The agent passes it structured data and a prediction task. Our TFM at @prior_labs - TabPFN uses historical structured data as context and returns results, in a single forward-pass.

000

Tuana @tuana.dev · 03/08/2026

LLMs handle reasoning and orchestration. But when it reasons that it needs to predict an outcome based on structured data, it should also reason about routing to an appropriate tool: a tabular foundation model.

100

Tuana @tuana.dev · 03/08/2026

I'm as LLM reliant as the next person.. But we forget there are other models too, specialized models that are cheaper and more effective to use depending on the task.. E.g.: Agents need tools for structured data predictions.

110

Tuana @tuana.dev · 20/03/2026

`npx skills add run-llama/llamaparse-agent-skills --skill liteparse`

100

Tuana @tuana.dev · 19/03/2026

When you hit more complex territory like scanned docs, dense tables, or multi-column layouts, that's where LlamaParse picks up. Same philosophy, more horsepower for the hard stuff.

100

Tuana @tuana.dev · 19/03/2026

The output is designed to be fed straight into agents so they can read parsed text and reason over screenshots without any extra wrangling.

100

Tuana @tuana.dev · 19/03/2026

It's built for developers who want parsing that stays on their own infrastructure and gets out of their way. Clean PDFs, DOCX, HTML: run it, get your text, move on.

100

Tuana @tuana.dev · 19/03/2026

We just open-sourced LiteParse 🎉 A lightweight, local document parser in the shape of an easy-to-use CLI. No API calls, no external service, no cloud dependency. Just fast text extraction from common file formats, right from your terminal.

100

Tuana @tuana.dev · 19/02/2026

Thanks! Fixed 🫶

000

Tuana @tuana.dev · 19/02/2026

Give it a try and let me know what you build. We're running a contest for the most difficult document workflow you can throw at it. Full video here: youtu.be/0Zhf5z2Onjs...

youtube.com

Vibe-Code a Document Agent with LlamaAgents

In this video, Senior DevRel Enginner Tuana Celik walks through how to use LlamaAgent Builder to create an end-to-end agent workflow. The example here is an ...

000

Tuana @tuana.dev · 19/02/2026

the whole point is you're describing the problem, not building the pipeline. the agent builder decides the architecture BUT with the caveat that coding agents aren't perfect and the code is yours to edit and perfect!

100

Tuana @tuana.dev · 19/02/2026

The agent builder figured out it needed Split and Extract, configured both, built the workflow, and deployed it: API + UI, code in my GitHub

100

Tuana @tuana.dev · 19/02/2026

I revisited my old demo: I took a resume book from NYU (resumes mixed with cover pages and curriculum pages) and just told the agent builder: split this into individual resumes, ignore the rest, extract graduation year, work experience, etc.

100

Tuana @tuana.dev · 19/02/2026

I filmed a walkthrough of LlamaAgent Builder, our new tool for building document agents by just describing what you want @llama_index

110

Tuana @tuana.dev · 29/01/2026

Announcement: www.llamaindex.ai/blog/llamaa... Documentation: developers.llamaindex.ai/python/llam...

developers.llamaindex.ai

Agent Builder

000

Tuana @tuana.dev · 29/01/2026

Note: our first release is a beta release, tuned and optimized for agents that do complex document extractions. We'll be adding more use cases as we test and confirm quality. If you try it out, please let us know! Drop us feedback in the UI or DM me 💛

110

Tuana @tuana.dev · 29/01/2026

But: ultimately the workflow is code, which the builder will create a repository for wherever you want, and deploy to LlamaCloud after. You get the ease of no-code to begin with, but full flexibility to customize (or fully change) the code if you want.

100

Tuana @tuana.dev · 29/01/2026

🧠 The Builder will generate the agent workflow code, ask you clarifying questions etc 🎨 While this is happening, you'll see a visualisation of the resulting workflow

100

Tuana @tuana.dev · 29/01/2026

We have a new tool to help you build and deploy document agents in LlamaCloud. The LlamaAgents Builder is kiiinda no-code, but not: 🦋 We have a new chat interface: just describe the document processing task you want in natural language

100

Tuana @tuana.dev · 22/01/2026

· processing_options for fine-grained control when you need it. Plus new llama-cloud SDKs (Python + TypeScript) with way better developer experience. If you're on v1, you're fine, we're maintaining support. But for new projects, v2 and new SDKs is the way.

100

Tuana @tuana.dev · 22/01/2026

We rebuilt LlamaParse's API from the ground up, and also released new SDKs for LlamaCloud in its entirety. API v2 for LlamaParse simplfies parsing config into structured objects · input_options for file-specific settings. · output_options for controlling what you get back.

110

Tuana @tuana.dev · 29/12/2025

Latest example in our docs at @llama_index: LlamaSheets (one of our latest products within LlamaCloud) has an example listed on how you can use LlamaSheets alongside coding agents: developers.llamaindex.ai/python/clou...

developers.llamaindex.ai

Using LlamaSheets with Coding Agents

Step-by-step guide to analyzing spreadsheet data extracted by LlamaSheets using Coding Agents

000

Tuana @tuana.dev · 29/12/2025

For that to work, both the developer and the agent needs to understand the context that they're working in.

100

Tuana @tuana.dev · 29/12/2025

that makes the most sense. Something that bothers me: In an ideal world, the purpose isn't to replace the developer, but to get to a place where we have coding agents that can sit alongside us in the development process.

100

Tuana @tuana.dev · 29/12/2025

A completely new way of thinking about documentation, oss projects, developer tools we provide has actually been figuring out how to structure them so that without compromising the main audience (the developer) - we can make sure coding agents also get the most relevant context, formatted in the way

100

Tuana @tuana.dev · 29/12/2025

I went offline for a couple of days to be with family and it seems like all we talked about on this platform has been coding agents.

110

Tuana @tuana.dev · 02/12/2025

And of course, bring along any questions too!

000

Tuana @tuana.dev · 02/12/2025

That's a lot, so this week, we want to take the time to hear from you! On Thursday, me, @LoganMarkewich and @itsclelia will be in our Discords voice channel for an hour dedicated to chatting about these two new tools. Drop by for our office hours, we'd love to hear from you.

100

Tuana @tuana.dev · 02/12/2025

LlamaSheets: Another addition to LlamaCloud that parses, extracts information and deep context (also hidden in metadata) from tables and spreadsheets, as well as identifying sub-groups from complex sheets

110

Tuana @tuana.dev · 02/12/2025

The team at LlamaIndex have been cooking! 🧑‍🍳 🍳 Over the last few weeks, we released: LlamaAgents: This is agent workflows that come with complete, deployable templates (more coming on this this week!)

120

Tuana @tuana.dev · 25/11/2025

Humans parse this instantly. Agents however, struggle. So LlamaSheets handles the problem by extracting 40+ features per cell, clustering regions, preserving hierarchy, and finally outputting typed parquet files. Your agent gets clean data instead of formatted chaos.

100

Tuana @tuana.dev · 25/11/2025

We shipped LlamaSheets today (beta, free). It approaches spreadsheets as a visual structure: Bold headers, merged cells, color-coded categories.

100

Tuana @tuana.dev · 19/11/2025

- Initialize a document agent project starting with one of our templates (In the blog, we used the SEC Insights Agent as an example) - Serve agents locally - Deploy them to LlamaCloud - Use all the LlamaCloud tooling like Extract and Classify as inherent components to your agent workflows

100

Tuana @tuana.dev · 19/11/2025

Yesterday we announced the open preview for LlamaAgents So, me and my colleague Adrian wrote this intro blog to help you get started. Learn about all the document agent templates available to you via llamactl and:

111

Tuana @tuana.dev · 18/11/2025

For example, here's me extracting the train time table from CalTrain schedule which is 2 large tables in a PDF. I also used the schema generation option for this one. Just prompt -> check schema,.

000