iiinova — where your digital self keeps evolving

.txt @dottxtai.bsky.social · 12/03/2026

A lot of extraction systems have the same problem: EITHER cram thousands of possible values into the schema OR leave the field open & fix mistakes later The better option: keep the structure fixed and narrow the valid values at runtime See how it works: blog.dottxt.ai/external-constraints

010

.txt @dottxtai.bsky.social · 03/09/2025

The .txt team is growing. We're looking for a: - Founding DevRel Engineer, - Member of Technical Staff - LLM Inference, and - a Rust Developer Apply here: jobs.ashbyhq.com/dottxt

030

.txt @dottxtai.bsky.social · 20/08/2025

Be our developer zero. Join .txt as our Founding DevRel Engineer. Apply here: jobs.ashbyhq.com/dottxt

061

.txt @dottxtai.bsky.social · 13/07/2025

We changed the way we handle multimodal inputs in Outlines. Most libraries will give the prompt a special role, separate it from the other inputs. We don’t anymore.

061

.txt @dottxtai.bsky.social · 03/06/2025

Check out this cool project by @aastroza.bsky.social to automatically convert any TV stream into a real-time movie script! Amazingly cool project using Outlines to enforce movie script formatting, and generally an impressive project. Here's the entire JSON schema he uses. github.com/aastroza/tvtxt

github.com

GitHub - aastroza/tvtxt: [WIP] AI that "reads" live TV and writes it as a movie script in real-time.

[WIP] AI that "reads" live TV and writes it as a movie script in real-time. - aastroza/tvtxt

031

.txt @dottxtai.bsky.social · 30/05/2025

Another example from our internal context-free grammar hackathon. You can force a model to generate an acrostic, i.e. make each line start with a letter in a word -- HELLO in this case. Grammars are cool.

152

.txt @dottxtai.bsky.social · 29/05/2025

During our internal hackathon, one of our teams (named "Too Many Cooks") wrote a grammar-powered recipe generator. The language model can ONLY generate text consistent with this recipe format. Here's a recipe for the universe.

051

Reposted by .txt

Cameron @cameron.stream · 14/05/2025

🎵 Cameron with your eyes so bright 🎵 won't you join our Discord tonight

031

.txt @dottxtai.bsky.social · 14/05/2025

We've got a Discord channel if you want to talk about AI engineer, compound AI systems, low overhead agents (though who knows what that means), constrained decoding, or whatever you're building. Come on by! discord.gg/ErZ8XnCmkQ

discord.gg

Join the .txt Discord Server!

Check out the .txt community on Discord - hang out with 1565 other members and enjoy free voice and text chat.

020

Reposted by .txt

Cameron @cameron.stream · 13/05/2025

Loved this paper. Smart idea. Basically they "guess and check" whether a token would appease the compiler when an LLM is doing code generation.

092

Reposted by .txt

Alonso Silva @alonsosilva.bsky.social · 23/04/2025

My talk "Building Knowledge Graph-Based Agents with Structured Text Generation" at PyData Global 2024 is now available on YouTube: www.youtube.com/watch?v=94yu... #PyData #PyDataGlobal @dottxtai.bsky.social @pydata.bsky.social

youtube.com

Alonso Silva - Building Knowledge Graph-Based Agents with Structured Text Generation

YouTube video by PyData

041

Reposted by .txt

Cameron @cameron.stream · 28/04/2025

Do people like this type of thread? They're fun to write.

2111

.txt @dottxtai.bsky.social · 28/04/2025

We recently came across an interesting paper that helps LLMs be better at handling domain-specific languages like database queries or probabilistic programming languages, using an approach called "grammar prompting". Link + brief thread below.

1142

.txt @dottxtai.bsky.social · 14/04/2025

Check out the sieves library from Mantis NLP! sieves is a library for structured document processing. It works with Outlines! Go check out the library, it's pretty cool. Repo: github.com/mantisai/sie...

github.com

GitHub - MantisAI/sieves: NLP tasks with zero- and few-shot models.

NLP tasks with zero- and few-shot models. . Contribute to MantisAI/sieves development by creating an account on GitHub.

020

.txt @dottxtai.bsky.social · 10/04/2025

We're delighted to sponsor the AI User Conference! It's next week, April 15th-17th. Find us at online or at San Francisco's delightful Fort Mason. Tickets: www.aiuserconference.com @cameron_pfiffer will be presenting a design for Minerva, a frontier space colony resource management system.

011

.txt @dottxtai.bsky.social · 09/04/2025

Check out our advocacy team @willkurt.bsky.social and @cameron.pfiffer.org on the Weaviate Podcast!

032

Reposted by .txt

Cameron @cameron.stream · 02/04/2025

YOOOOOOOOOOOOOOOOO I got to teach a @deeplearningai.bsky.social course! I'm so happy. It was a really fun experience -- go take a look at the course and let us know how it went. bit.ly/4cdQlzG

bit.ly

Getting Structured LLM Output

Learn how to generate structured outputs to power production-ready LLM software applications.

2254

.txt @dottxtai.bsky.social · 02/04/2025

We are excited to announce our partnership with Andrew Ng and @deeplearningai.bsky.social! Our advocacy team @willkurt.bsky.social and @cameron.pfiffer.org created a course on building AI systems using structured output from LLMs. Oh, and it’s free. bit.ly/4cdQlzG

bit.ly

Getting Structured LLM Output

Learn how to generate structured outputs to power production-ready LLM software applications.

181

.txt @dottxtai.bsky.social · 01/04/2025

People love our Outlines library. We figured we'd release our next big product -- Inlines! Inlines are high-performance inline roller blades for the discerning AI engineer in your life. Reply in the comments and we'll add you to the pre-order list!

151

.txt @dottxtai.bsky.social · 31/03/2025

Wednesday, April 2nd, at 8am PST. Stay tuned.

040

.txt @dottxtai.bsky.social · 25/03/2025

Outlines supports "recursive schemas", which are JSON schemas that refer to themselves. For example, here is a simple schema to generate a recursive concept map. This defines a concept, which has a title, a description, and a list of related concepts with the same schema.

140

.txt @dottxtai.bsky.social · 17/03/2025

Outlines v1.0 has a few interface changes coming up. One is the addition of the Application class, which combines a prompt templating function and an output type. Then, you call that application with the language model you want to use, as well as the arguments to the prompt function.

040

.txt @dottxtai.bsky.social · 12/03/2025

Here's what happens if you disable the use of "R" and "r" in the thinking block of Qwen 7B R1. We asked if roses are red, and it invented the phrase "conclusion point" to use instead of "answer". It eventually got the right answe.

110

.txt @dottxtai.bsky.social · 11/03/2025

What features do you want to see in Outlines?

010

.txt @dottxtai.bsky.social · 06/03/2025

Is that our advocacy team @willkurt.bsky.social and @cameron.pfiffer.org hanging out with Andrew Ng? Wonder why.

030

.txt @dottxtai.bsky.social · 04/03/2025

Let us know how you're using this!

061

.txt @dottxtai.bsky.social · 03/03/2025

Highly recommend this post by @simonwillison.net about extracting structured data from text with LLMs. Simon puts it well: "the single most commercially valuable application of LLMs is turning unstructured content into structured data."

buff.ly

Structured data extraction from unstructured content using LLM schemas

LLM 0.23 is out today, and the signature feature is support for schemas—a new way of providing structured output from a model that matches a specification provided by the user. …

191

.txt @dottxtai.bsky.social · 28/02/2025

Théophile Cantelobre built a map of France's culinary universe by extracting structured data from restaurant reviews using Outlines. They created a knowledge graph connecting chefs and restaurants that traditional prompting couldn't achieve.

buff.ly

Mapping the French Culinary Universe | Théophile Cantelobre

Creating a knowledge graph of restaurants featured in lefooding.com reviews using `outlines` and `gpt4o-mini`.

063

.txt @dottxtai.bsky.social · 26/02/2025

We have an open discussion on GitHub about providing token-triggered logits for Outlines, which would allow separate logit processors for different parts of the model's response, e.g. - Easy reasoning blocks for R1 and others - Disallow/allow words in think blocks 👇

100

.txt @dottxtai.bsky.social · 26/02/2025

Your quarterly reminder to just use structured generation. Credit: @voooooogel on twitter

A webcomic.

Top right: this meme made by structured generation gang

First panel: a gymnast woman on a mat who simply walks across the mat and raises her arms as if she has completed a complex task. The text reads "just sample what you want".

Second panel: A series of images of characters doing complicated gymnastics, including bars, rings, pommel horse, and even a person on fire flying over a burning car. The text reads

- "Sample arbitrary text"
- "Oh no it's not the right format"
- "Add prompt hints"
- "Install langchain and add a 7 call chain of repaired self consistency pipeline you saw on arxiv"
- "ONLY OUTPUT JSON OR YOU WILL BE FIRED"

091

.txt @dottxtai.bsky.social · 24/02/2025

Outlines 0.2.1 is out! - Renaming `prompt` to `Template`. `prompt` is now deprecated. - First commit by .txt rockstar Dan Palmer-Bancel - Documentation fixes by EdAbati - Adding our bluesky account to the docs (come find us there) Full changelog: buff.ly/4bbwWyN

010

.txt @dottxtai.bsky.social · 18/02/2025

Check out our stellar engineer Yvan at the Rust In Paris conference! Discount link: buff.ly/3CMTE3E Speakers: buff.ly/4b7fQlF Schedule: buff.ly/4gKwc4P

022

.txt @dottxtai.bsky.social · 11/02/2025

We're looking for a Staff Rust Developer! We're looking for people with - Great communication skills - Experience withlarge Rust projects - Skills in performant parsing - Understanding of automata/state machines, text indexing + search, serialization Application here if you fit:

buff.ly

Staff Rust Developer

RESPONSIBILITIES - Initiate and lead the development of Rust projects - Write performance-focused parsing code backed by automated benchmarking - Research and implement efficient data structures for…

042

.txt @dottxtai.bsky.social · 11/02/2025

Looking for great learning resources about structured generation? Check out this GitHub repo from Sacha Ichbiah, one of our Discord users. Repo: buff.ly/4jOI3Bz Our Discord: buff.ly/4fULKn9

061

.txt @dottxtai.bsky.social · 06/02/2025

We're always looking for blog posts and writing demonstrating how to use Outlines and structured generation. If you've written anything you'd like us to read and share, please reach out!

111

.txt @dottxtai.bsky.social · 04/02/2025

DeepSeek's R1 model is a reasoning model, which means that it generates "thinking text" before responding. A challenge is allowing the model to "think" while structuring its final output. Here's some experimental code to get JSON from R1, using Outlines! Let the tokens flow.

buff.ly

Using Outlines to get structured output from R1

Using Outlines to get structured output from R1. GitHub Gist: instantly share code, notes, and snippets.

141

.txt @dottxtai.bsky.social · 03/02/2025

The .txt Discord office hours are coming up. Tuesday January 4th at 8am PST! It's a good place to - Get technical support - Share what you're working on - Chat about AI systems buff.ly/3Ck3zgV

000

.txt @dottxtai.bsky.social · 30/01/2025

We'll be hosting weekly office hours on our Discord server! Our developer relations engineer Cameron will be there to answer questions, talk about AI engineering, and generally chat about what you're building. Come see us on Tuesday mornings at 8am PST!

buff.ly

Join the .txt Discord Server!

Check out the .txt community on Discord - hang out with 1407 other members and enjoy free voice and text chat.

021

.txt @dottxtai.bsky.social · 30/01/2025

Want to talk about building AI systems? People have been sharing all the cool things they're working on, and they'd like to see whatever you're working on too. Come join our Discord!

buff.ly

Join the .txt Discord Server!

Check out the .txt community on Discord - hang out with 1375 other members and enjoy free voice and text chat.

001