NeuML/txtai-neuml-linkedin
Sentence Similarity
•
Updated
•
10
•
1
Post title
stringlengths 27
1.25k
| Post link
stringlengths 72
72
| Post type
stringclasses 1
value | Posted by
stringclasses 1
value | Created date
stringlengths 10
10
| Audience
stringclasses 1
value | Impressions
int64 136
7.2k
| Clicks
int64 1
337
| Click through rate (CTR)
float64 0
0.15
| Likes
int64 1
81
| Comments
int64 0
15
| Reposts
int64 0
7
| Engagement rate
float64 0.01
0.18
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|
🚀 Time for our first release of 2025 - txtai 8.2!
This release simplifies LLM chat messages, adds attribute filtering to Graph RAG and enables multi-cpu/gpu vector encoding.
txtai is the best framework around for Agents, RAG and Vector Search! Read more below.
GitHub: https://lnkd.in/dxWDeey
Release Notes: https://lnkd.in/e89BtUZS
PyPI: https://lnkd.in/eE_Jvft
Docker Hub: https://lnkd.in/e598zTHb | https://www.linkedin.com/feed/update/urn:li:activity:7283225139486326784 | Organic | David Mezzetti | 01/09/2025 | All followers | 2,044 | 112 | 0.054795 | 23 | 0 | 4 | 0.068004 |
125 ⭐'s to go for txtai 10K!
https://lnkd.in/dxWDeey | https://www.linkedin.com/feed/update/urn:li:activity:7282595210184863744 | Organic | David Mezzetti | 01/08/2025 | All followers | 320 | 10 | 0.03125 | 3 | 0 | 0 | 0.040625 |
🦠 Here is a dataset with PubMed article metadata related to Human metapneumovirus (HMPV). It's also a resource for those who want to get more educated on the matter with reliable data. Don't just believe what you see on social media.
Dataset: https://lnkd.in/eVrkKgHB
CSV: https://lnkd.in/egYqgmuF | https://www.linkedin.com/feed/update/urn:li:activity:7282582586344800257 | Organic | David Mezzetti | 01/08/2025 | All followers | 259 | 4 | 0.015444 | 3 | 0 | 1 | 0.030888 |
📄⚙️ Want to parse medical and scientific papers? Then check out paperetl.
With paperetl, you can select a subset of the PubMed archive using id's, MeSH codes and/or keywords. Build your own medical knowledge base. It also supports parsing full PDFs. All code no AI in this project.
https://lnkd.in/ecuuVG2
| https://www.linkedin.com/feed/update/urn:li:activity:7282485854973902851 | Organic | David Mezzetti | 01/07/2025 | All followers | 448 | 19 | 0.042411 | 4 | 1 | 1 | 0.055804 |
Want to build your own Speech to Speech RAG pipeline? Then check out this video tutorial!
https://lnkd.in/eEGGsXBQ | https://www.linkedin.com/feed/update/urn:li:activity:7282473058655715328 | Organic | David Mezzetti | 01/07/2025 | All followers | 231 | 8 | 0.034632 | 2 | 0 | 1 | 0.047619 |
🦠 We're hearing a lot about H5N1. Here is a dataset with PubMed article metadata related to H5N1 in hopes it can help those looking to prevent this from growing into something bigger. It's also a resource for those who want to get more educated on the matter with reliable data.
Dataset: https://lnkd.in/eK82xAWq
CSV: https://lnkd.in/epFGB6AY | https://www.linkedin.com/feed/update/urn:li:activity:7282054158587682817 | Organic | David Mezzetti | 01/06/2025 | All followers | 417 | 14 | 0.033573 | 5 | 0 | 1 | 0.047962 |
🧬⚕️🔬 What if we said you can have a competitive 8 million parameter model that can index as fast as BM25? Or a 100K parameter 200KB model that retains knowledge?
We're excited to release static PubMedBERT Embeddings models! These are a set of static models distilled with the great Model2Vec library by The Minish Lab Thank you to Stéphan Tulkens and Thomas van Dongen for creating Model2Vec!
https://lnkd.in/esgtC_2X | https://www.linkedin.com/feed/update/urn:li:activity:7281057260766695424 | Organic | David Mezzetti | 01/03/2025 | All followers | 2,850 | 179 | 0.062807 | 44 | 5 | 3 | 0.081053 |
🤖📄 It's 2025 and you want to get into AI? Perhaps you've tried ChatGPT and want to see how something can be done with your own data locally?
Then check out our RAG application. It's an easy-to-run Docker container that has Vector RAG, Graph RAG, PDF text extraction, vector search and and more all built-in with a UI! It's also a great example on how to use txtai.
https://lnkd.in/evdB5HgN | https://www.linkedin.com/feed/update/urn:li:activity:7280955767967346688 | Organic | David Mezzetti | 01/03/2025 | All followers | 690 | 30 | 0.043478 | 9 | 0 | 1 | 0.057971 |
🔥 Happy to share this new article on using txtai with LLM APIs by Igor Ribeiro Lima!
https://lnkd.in/ercixQKS | https://www.linkedin.com/feed/update/urn:li:activity:7280647029133832193 | Organic | David Mezzetti | 01/02/2025 | All followers | 584 | 22 | 0.037671 | 3 | 0 | 1 | 0.044521 |
📰🔥 NeuML - 2024 Year in Review
Agents, GraphRAG, Speech to Speech RAG, Postgres, Streaming LLMs, Docling, Model2Vec and Consulting!
Lots of fun ahead in 2025. This article is worth a read!
https://lnkd.in/e2PWVEjS | https://www.linkedin.com/feed/update/urn:li:activity:7280323095284834304 | Organic | David Mezzetti | 01/01/2025 | All followers | 660 | 23 | 0.034848 | 9 | 1 | 2 | 0.05303 |
Happy New Year! Wishing you all the best in 2025 🍾🥂🎉 | https://www.linkedin.com/feed/update/urn:li:activity:7280031297962405888 | Organic | David Mezzetti | 01/01/2025 | All followers | 241 | 1 | 0.004149 | 2 | 0 | 1 | 0.016598 |
💥 Surprise! One last release in 2024.
🧬⚕️🔬 This time it's paperai and paperetl 2.3. paperai is a semantic search and workflow application for medical/scientific papers. paperetl is a companion library for parsing papers.
This update enables the latest txtai features and fixes some long standing bugs. More to come in 2025 and maybe one more thing in 2024!
paperai: https://lnkd.in/egFSSP4
paperetl: https://lnkd.in/ecuuVG2 | https://www.linkedin.com/feed/update/urn:li:activity:7278873244105904128 | Organic | David Mezzetti | 12/28/2024 | All followers | 5,229 | 337 | 0.064448 | 77 | 4 | 3 | 0.080513 |
Merry Christmas and Happy Holidays to all! 🎄🎅❄️ | https://www.linkedin.com/feed/update/urn:li:activity:7277350467746107392 | Organic | David Mezzetti | 12/24/2024 | All followers | 342 | 5 | 0.01462 | 3 | 0 | 1 | 0.026316 |
See why txtai had over 4000 (real) ⭐'s in 2024 | https://www.linkedin.com/feed/update/urn:li:activity:7276223906695319552 | Organic | David Mezzetti | 12/21/2024 | All followers | 576 | 15 | 0.026042 | 7 | 2 | 1 | 0.043403 |
The NeuML Holiday 🎄 🎅 ❄️ 🎆 newsletter is out!
https://lnkd.in/eEAhfu9N
| https://www.linkedin.com/feed/update/urn:li:activity:7275902067007709184 | Organic | David Mezzetti | 12/20/2024 | All followers | 364 | 2 | 0.005495 | 1 | 0 | 1 | 0.010989 |
🎄🎅🎶 With the year winding down and us heading into the holidays, it's amazing to think what is now and what will be possible soon.
Still 🤯 that a set of open source tools can clone my voice and recite the night before Christmas with an AI-generated music track to go along with it!
Read more: https://lnkd.in/dPmt-bZm | https://www.linkedin.com/feed/update/urn:li:activity:7275539285963882498 | Organic | David Mezzetti | 12/19/2024 | All followers | 353 | 11 | 0.031161 | 3 | 0 | 1 | 0.042493 |
2025 is the year of the AI vertical you say? | https://www.linkedin.com/feed/update/urn:li:activity:7275488346255085568 | Organic | David Mezzetti | 12/19/2024 | All followers | 264 | 5 | 0.018939 | 3 | 0 | 1 | 0.034091 |
Medium article covering the AnnotateAI project.
https://lnkd.in/ec6KGvsg | https://www.linkedin.com/feed/update/urn:li:activity:7275330519762972672 | Organic | David Mezzetti | 12/19/2024 | All followers | 312 | 8 | 0.025641 | 1 | 0 | 1 | 0.032051 |
🚚 📦 We're shipping today! This time it's AnnotationAI v0.2!
This updates the app to be able to run embeddings searches against txtai's ArXiv database when there is no active url.
GitHub: https://lnkd.in/ev4xH34m
Release Notes: https://lnkd.in/eWvD66tf
PyPI: https://lnkd.in/ehSq8uAA
Docker Hub: https://lnkd.in/e9RURbPD | https://www.linkedin.com/feed/update/urn:li:activity:7275213835186126849 | Organic | David Mezzetti | 12/18/2024 | All followers | 1,012 | 87 | 0.085968 | 13 | 1 | 1 | 0.100791 |
The txtai-arxiv embeddings database on the Hugging Face Hub has been updated with data through December 2024!
https://lnkd.in/eSCCs-Jz | https://www.linkedin.com/feed/update/urn:li:activity:7275211742740201472 | Organic | David Mezzetti | 12/18/2024 | All followers | 413 | 17 | 0.041162 | 6 | 0 | 1 | 0.058111 |
📚 Excited to release another new project, RAG Data.
While we have long provided publicly available embeddings databases for ArXiv and Wikipedia, the code wasn't easy to find. This project fixes that!
This project can be used as a starting point for those building their own large embeddings databases. It has multiprocessing and other performance enhancements built-in.
GitHub: https://lnkd.in/eTBJCiRr
PyPI: https://lnkd.in/eW2W6yVW
| https://www.linkedin.com/feed/update/urn:li:activity:7275211468378144768 | Organic | David Mezzetti | 12/18/2024 | All followers | 516 | 34 | 0.065891 | 8 | 0 | 1 | 0.083333 |
🚀 Happy to release v0.9.0 of the txtai rag app! This version adds support for extracting text from documents via Docling.
GitHub: https://lnkd.in/evdB5HgN
Release Notes: https://lnkd.in/eJsxDBX7
Docker Hub: https://lnkd.in/e-3Cx_68 | https://www.linkedin.com/feed/update/urn:li:activity:7274821495594315778 | Organic | David Mezzetti | 12/17/2024 | All followers | 396 | 24 | 0.060606 | 2 | 0 | 1 | 0.068182 |
Want to add highlights and text annotations to PDFs, then check out txtmarker!
https://lnkd.in/dqkstw5 | https://www.linkedin.com/feed/update/urn:li:activity:7274065142089240577 | Organic | David Mezzetti | 12/15/2024 | All followers | 515 | 13 | 0.025243 | 2 | 0 | 1 | 0.031068 |
There is now a Docker app available for AnnotateAI. The web app automatically annotates papers from URLs. It also renders a version of the annotated PDF right in the browser along with being available for download.
Check it out!
https://lnkd.in/e9RURbPD | https://www.linkedin.com/feed/update/urn:li:activity:7274049801124462592 | Organic | David Mezzetti | 12/15/2024 | All followers | 817 | 60 | 0.073439 | 10 | 0 | 1 | 0.086903 |
💥📝 We're excited to release a new project.....AnnotateAI
Who reads papers around here? Lots of us do we're sure! There are plenty of projects to summarize, search and build generative systems with papers. What about helping us read them?
Well that's where AnnotateAI comes in! AnnotateAI automatically annotates papers using LLMs. This project focuses on providing human readers with context as they read.
Click through the link below to learn more!
https://lnkd.in/dCjtYPeR | https://www.linkedin.com/feed/update/urn:li:activity:7273400490766315530 | Organic | David Mezzetti | 12/13/2024 | All followers | 1,382 | 112 | 0.081042 | 23 | 3 | 2 | 0.101302 |
Postgres is all you need for vectors. Start small with SQLite and move up to Postgres for the win with txtai.
https://lnkd.in/eFg3zE4A
| https://www.linkedin.com/feed/update/urn:li:activity:7272704947547168768 | Organic | David Mezzetti | 12/11/2024 | All followers | 377 | 3 | 0.007958 | 5 | 0 | 0 | 0.02122 |
💥 txtai 8.1 is out!
txtai 8.1 adds Docling integration, Embeddings context managers and significant database component enhancements. The latest version of the txtai RAG application is also available on Docker Hub.
See below for more.
GitHub: https://lnkd.in/dxWDeey
Release Notes: https://lnkd.in/ecK4KnU5
PyPI: https://lnkd.in/eE_Jvft
Docker Hub: https://lnkd.in/e598zTHb | https://www.linkedin.com/feed/update/urn:li:activity:7272309319247740929 | Organic | David Mezzetti | 12/10/2024 | All followers | 1,898 | 123 | 0.064805 | 27 | 15 | 1 | 0.08746 |
txtai has a robust and growing integration with Postgres. It has the ability to persist each of it's components to Postgres as follows:
- Vectors via pgvector
- Sparse vectors via full-text search
- Documents are persisted as JSON columns and fully searchable
- Graph edges and nodes
More features are underway.
https://lnkd.in/eFeFNgYK
| https://www.linkedin.com/feed/update/urn:li:activity:7271642302928850944 | Organic | David Mezzetti | 12/08/2024 | All followers | 398 | 4 | 0.01005 | 1 | 0 | 1 | 0.015075 |
When building AI apps moving into 2025, you have a choice. A framework where with a couple config changes you can move quickly from prototyping locally to production. Or you can use frameworks that require you to write a bunch of gobbledygook code and imports for everything. And maybe you'll get it into production 😀
Want to learn more, take a look at txtai: https://lnkd.in/dxWDeey | https://www.linkedin.com/feed/update/urn:li:activity:7270474156771577856 | Organic | David Mezzetti | 12/05/2024 | All followers | 319 | 11 | 0.034483 | 2 | 0 | 1 | 0.043887 |
Coming in txtai 8.1 - support for Docling!
Docling is rapidly growing in popularity due to robust PDF text extraction and table parsing support.
Support is in the main txtai in GitHub now.
https://lnkd.in/gUYgQByS
| https://www.linkedin.com/feed/update/urn:li:activity:7270031724686843905 | Organic | David Mezzetti | 12/04/2024 | All followers | 517 | 27 | 0.052224 | 5 | 0 | 1 | 0.06383 |
With AWS re:Invent upon us, did you know that txtai can flexibly plug into AWS?
Here is a diagram for hosting a RAG service. | https://www.linkedin.com/feed/update/urn:li:activity:7269046575736270851 | Organic | David Mezzetti | 12/01/2024 | All followers | 312 | 6 | 0.019231 | 4 | 0 | 1 | 0.035256 |
The LLM pipeline with txtai is designed to be flexible. It supports running models via Transformers, llama.cpp and LLM API services (via LiteLLM) through a single interface. Provide prompts as strings or chat messages. GGUF, AWQ and more.
Read more here: https://lnkd.in/ebMr77S2 | https://www.linkedin.com/feed/update/urn:li:activity:7269010664717799424 | Organic | David Mezzetti | 12/01/2024 | All followers | 369 | 7 | 0.01897 | 1 | 0 | 1 | 0.02439 |
Happy Thanksgiving! 🦃 🍂 🍻 🏈 | https://www.linkedin.com/feed/update/urn:li:activity:7267676372100169728 | Organic | David Mezzetti | 11/27/2024 | All followers | 225 | 1 | 0.004444 | 2 | 0 | 1 | 0.017778 |
🧬⚕️🔬 For those working with medical and scientific literature, the OpenScholar project from Ai2 is quite promising.
We like to work with AWQ quantization. So, we've added a AWQ-quantized version to the Hugging Face Hub!
Read more below.
AWQ Model: https://lnkd.in/eDtPvZgx
AI2 Blog: https://lnkd.in/dBZv5EPf
AI2 Paper: https://lnkd.in/ezQUsgTJ
| https://www.linkedin.com/feed/update/urn:li:activity:7267573987772096512 | Organic | David Mezzetti | 11/27/2024 | All followers | 464 | 21 | 0.045259 | 5 | 3 | 1 | 0.064655 |
Cool to see txtai holding it's own against the big dogs 🐕🦺
https://lnkd.in/d2-HKx_d | https://www.linkedin.com/feed/update/urn:li:activity:7267524131661701120 | Organic | David Mezzetti | 11/27/2024 | All followers | 312 | 19 | 0.060897 | 7 | 2 | 1 | 0.092949 |
🚀 The txtai journey over the last two years in one article. Plus a bunch of tips on how you can do the same.
https://lnkd.in/eX4bn5V8 | https://www.linkedin.com/feed/update/urn:li:activity:7267212351513284608 | Organic | David Mezzetti | 11/26/2024 | All followers | 319 | 11 | 0.034483 | 4 | 0 | 1 | 0.050157 |
🤖✨ Build autonomous agents with txtai!
https://lnkd.in/eQzJQz6B | https://www.linkedin.com/feed/update/urn:li:activity:7266924019315093506 | Organic | David Mezzetti | 11/25/2024 | All followers | 245 | 6 | 0.02449 | 2 | 0 | 1 | 0.036735 |
There are lots of AI project charts out there with tiny logos that fit nicely into tiny boxes of functionality. Then it's only you to connect all the tiny logos together into a working system.
What if there is an all-in-one solution?
https://lnkd.in/dxWDeey | https://www.linkedin.com/feed/update/urn:li:activity:7266476728049446912 | Organic | David Mezzetti | 11/24/2024 | All followers | 588 | 27 | 0.045918 | 6 | 0 | 2 | 0.059524 |
Did you know txtai agents can load embeddings databases right from the Hugging Face Hub?
Want to add your own knowledge base contribution? Then what are you waiting for!
https://lnkd.in/eyTV9uwR | https://www.linkedin.com/feed/update/urn:li:activity:7266047086054227968 | Organic | David Mezzetti | 11/23/2024 | All followers | 352 | 2 | 0.005682 | 3 | 0 | 1 | 0.017045 |
The foundation of txtai is it's embeddings database. This is where knowledge is stored and it powers other components. From here we can build autonomous agents, retrieval augmented generation (RAG) processes and multi-model workflows.
Default configuration is provided out of the box for all these components, so you can get up and running fast!
https://lnkd.in/dxWDeey | https://www.linkedin.com/feed/update/urn:li:activity:7266038865071587328 | Organic | David Mezzetti | 11/23/2024 | All followers | 298 | 5 | 0.016779 | 4 | 0 | 1 | 0.033557 |
Agents are great but what if you have a known and repeatable process? Then don't overengineer it, use a workflow! Check out this example that builds a scheduled workflow that pushes out notifications.
https://lnkd.in/dvSkSepJ
| https://www.linkedin.com/feed/update/urn:li:activity:7265822308466839552 | Organic | David Mezzetti | 11/22/2024 | All followers | 235 | 3 | 0.012766 | 1 | 0 | 1 | 0.021277 |
The NeuML Thanksgiving 🦃 🍂 🍻 🏈 newsletter is out!
https://lnkd.in/e-veEUhS
| https://www.linkedin.com/feed/update/urn:li:activity:7265772152988078080 | Organic | David Mezzetti | 11/22/2024 | All followers | 136 | 4 | 0.029412 | 1 | 0 | 0 | 0.036765 |
Analyzing Hugging Face Posts with Graphs and Agents! Check out this article that explores graph analysis and agent execution.
Learn what and who's popular and trending in the world of AI!
https://lnkd.in/eFmdz6bF | https://www.linkedin.com/feed/update/urn:li:activity:7265544322547171329 | Organic | David Mezzetti | 11/22/2024 | All followers | 722 | 47 | 0.065097 | 12 | 3 | 1 | 0.087258 |
↪️️ Workflows vs 🤖 Agents in txtai - there's a place for both! | https://www.linkedin.com/feed/update/urn:li:activity:7265138396157648896 | Organic | David Mezzetti | 11/20/2024 | All followers | 358 | 17 | 0.047486 | 3 | 0 | 1 | 0.058659 |
txtai has a lot of features. Sometimes we miss the easy use cases!
https://lnkd.in/eP3ppFaY | https://www.linkedin.com/feed/update/urn:li:activity:7265055770658881537 | Organic | David Mezzetti | 11/20/2024 | All followers | 258 | 6 | 0.023256 | 3 | 0 | 1 | 0.03876 |
Our PubMedBERT Embeddings model has over 100K downloads and 100 likes on the Hugging Face Hub. It also has a growing number of citations, albeit sadly it's often paired with LangChain 🙁.
Model: https://lnkd.in/egnEKcqd
Google Scholar Search: https://lnkd.in/exFepgQV
| https://www.linkedin.com/feed/update/urn:li:activity:7265021014332293120 | Organic | David Mezzetti | 11/20/2024 | All followers | 544 | 19 | 0.034926 | 8 | 0 | 1 | 0.051471 |
That was fast! Check out this excellent video on how to build AI Agents with txtai from the community!
https://lnkd.in/evRmtZBT | https://www.linkedin.com/feed/update/urn:li:activity:7265018068584558593 | Organic | David Mezzetti | 11/20/2024 | All followers | 310 | 8 | 0.025806 | 3 | 0 | 1 | 0.03871 |
Check out this video on RAG with txtai from the community!
https://lnkd.in/eQEw6bDi | https://www.linkedin.com/feed/update/urn:li:activity:7265017977404624896 | Organic | David Mezzetti | 11/20/2024 | All followers | 179 | 3 | 0.01676 | 1 | 0 | 1 | 0.027933 |
🥁 Alright....so here we go. We're proud to announce the release of txtai 8.0! This release adds an agent framework to txtai (built on top of Transformers Agents 🤗 with all LLMs supported).
txtai is now one of the most straightforward ways to add real-world agents to production without the bloat. LFG to the 🚀🌕!
GitHub: https://lnkd.in/dxWDeey
Release Notes: https://lnkd.in/e47jPNKZ
PyPI: https://lnkd.in/eE_Jvft
Docker Hub: https://lnkd.in/e598zTHb
Article: https://lnkd.in/e7XPs6Ub | https://www.linkedin.com/feed/update/urn:li:activity:7264489976346636288 | Organic | David Mezzetti | 11/19/2024 | All followers | 1,593 | 109 | 0.068424 | 28 | 1 | 3 | 0.088512 |
💥 The txtai 8.0 release is coming very soon! The big leap forward is a full agent framework. Simplicity, easy of use and integration with the txtai ecosystem inbound.
2025 is going to be 🔥 | https://www.linkedin.com/feed/update/urn:li:activity:7263523258203795456 | Organic | David Mezzetti | 11/16/2024 | All followers | 612 | 21 | 0.034314 | 12 | 2 | 1 | 0.058824 |
Bored? Want to read something mildly interesting? Then check out this article on how to run txtai in assembly 😀
https://lnkd.in/emHFyU98 | https://www.linkedin.com/feed/update/urn:li:activity:7263365469854744576 | Organic | David Mezzetti | 11/16/2024 | All followers | 335 | 8 | 0.023881 | 3 | 0 | 1 | 0.035821 |
Looking to "hire" an "Agent Team"? Then check out the link below. Work smarter not harder 🔥
https://lnkd.in/eaAwEyyq | https://www.linkedin.com/feed/update/urn:li:activity:7260662463963058177 | Organic | David Mezzetti | 11/08/2024 | All followers | 425 | 22 | 0.051765 | 4 | 1 | 1 | 0.065882 |
🚀 We're thrilled to share a preview version of txtai agents. Inspired by the simplicity of frameworks like OpenAI Swarm, txtai agents are built on top of the Transformers Agent framework.
This supports all LLMs txtai supports (Hugging Face, llama.cpp, OpenAI + Claude + AWS Bedrock via LiteLLM).
Available in GitHub now, will be released soon!
Example code: https://lnkd.in/eapczzk3 | https://www.linkedin.com/feed/update/urn:li:activity:7259584341209485312 | Organic | David Mezzetti | 11/05/2024 | All followers | 608 | 29 | 0.047697 | 7 | 0 | 2 | 0.0625 |
🎃👻 Happy Halloween! Recently txtai generated this "spooky"🕸️ audio of "The Raven" using a speech + audio workflow.
Hope everyone has a fun evening! | https://www.linkedin.com/feed/update/urn:li:activity:7257765442440556544 | Organic | David Mezzetti | 10/31/2024 | All followers | 213 | 10 | 0.046948 | 2 | 0 | 1 | 0.061033 |
The future of AI will be driven by your knowledge and your data. Those with the best knowledge bases win the day.
That's why txtai makes it easy to build all sorts of knowledge bases and store them in all sorts of places. Postgres, SQLite, S3, Hugging Face Hub and more!
https://lnkd.in/eMGY7uRB
| https://www.linkedin.com/feed/update/urn:li:activity:7257094886057742336 | Organic | David Mezzetti | 10/29/2024 | All followers | 317 | 8 | 0.025237 | 3 | 0 | 2 | 0.041009 |
✨ Did you know that each txtai embeddings, pipeline and/or workflow can easily be served with a built-in API?
txtai has a scaffolding framework to automatically build FastAPI endpoints. In addition to that, any custom API endpoint can be added as shown in the article below.
https://lnkd.in/ei-u7grV
| https://www.linkedin.com/feed/update/urn:li:activity:7256623503003521024 | Organic | David Mezzetti | 10/28/2024 | All followers | 372 | 8 | 0.021505 | 5 | 2 | 1 | 0.043011 |
🚀 The next release of txtai will have easy-to-use agents support. It's going to be 🔥!
https://lnkd.in/dxWDeey
| https://www.linkedin.com/feed/update/urn:li:activity:7255893422522208256 | Organic | David Mezzetti | 10/26/2024 | All followers | 451 | 17 | 0.037694 | 4 | 0 | 1 | 0.04878 |
The NeuML Halloween 🎃 👻 🦇 🕷️ newsletter is out!
https://lnkd.in/efTYfZwD | https://www.linkedin.com/feed/update/urn:li:activity:7255544570149572608 | Organic | David Mezzetti | 10/25/2024 | All followers | 180 | 3 | 0.016667 | 1 | 0 | 1 | 0.027778 |
A hotfix release of txtai (7.5.1) is out! This addresses a breaking change from the huggingface-hub library.
https://lnkd.in/ekz77NG8
| https://www.linkedin.com/feed/update/urn:li:activity:7255535693475598336 | Organic | David Mezzetti | 10/25/2024 | All followers | 243 | 10 | 0.041152 | 2 | 0 | 1 | 0.053498 |
The AI space is growing fast but it's often hard to know how to apply AI to your business.
Did you know that in addition to our open-source development that NeuML also provides advisory and strategic support services (i.e. Fractional CTO)?
Reach out to learn more!
https://neuml.com | https://www.linkedin.com/feed/update/urn:li:activity:7255193102636781569 | Organic | David Mezzetti | 10/24/2024 | All followers | 444 | 16 | 0.036036 | 6 | 0 | 1 | 0.051802 |
🚀 Thank you to the 9K people who have given txtai a ⭐!
A goal set from the beginning was to reach 10K ⭐'s on GitHub which will happen at some point over the coming months. While some projects race out to a fast start and are fortunate to trend, txtai has been more of a steady incline.
It's been a journey full of ups and downs, thank you to all those following! If you haven't given txtai a star yet and like it, what are you waiting for?
https://lnkd.in/dxWDeey | https://www.linkedin.com/feed/update/urn:li:activity:7254998940314427393 | Organic | David Mezzetti | 10/23/2024 | All followers | 353 | 11 | 0.031161 | 7 | 1 | 1 | 0.056657 |
🚀 txtai is built on the shoulders of open-source giants. Transformers, SQLite, NetworkX, Postgres, Faiss, NumPy are all major players.
Small projects just work without any configuration. Larger projects can integrate with Postgres, which is built on almost 30 years of production experience.
Don't settle for less just because "AI Influencers" say so. We're not paying anyone to do our bidding - we work hard to build a solid project.
https://lnkd.in/eh4-rVEr | https://www.linkedin.com/feed/update/urn:li:activity:7254854842664312832 | Organic | David Mezzetti | 10/23/2024 | All followers | 387 | 4 | 0.010336 | 3 | 0 | 1 | 0.020672 |
txtai 7.5 added a number of features to add Speech to Speech RAG. One of the more complicated tasks was voice activity detection.
If you like FFTs and Butterworth filters, then click through to read more!
https://lnkd.in/e3EB743g | https://www.linkedin.com/feed/update/urn:li:activity:7254515978724691969 | Organic | David Mezzetti | 10/22/2024 | All followers | 271 | 5 | 0.01845 | 1 | 0 | 1 | 0.02583 |
Running an embeddings database with limited compute? Model2Vec is a technique to turn any sentence transformer into a really small static model.
The next release of txtai will have support for this new vectorization method!
Example txtai code: https://lnkd.in/eFSdT7B5
GitHub Project: https://lnkd.in/eEpWxZx8
Blogpost: https://lnkd.in/e4iPg4wi | https://www.linkedin.com/feed/update/urn:li:activity:7254193006306635777 | Organic | David Mezzetti | 10/21/2024 | All followers | 4,042 | 135 | 0.033399 | 65 | 4 | 4 | 0.05146 |
Think Python always has to be slow? Well think again. Check out this article on how txtai was able to achieve near native performance with it's sparse indexes.
https://lnkd.in/eA3ui6cQ | https://www.linkedin.com/feed/update/urn:li:activity:7253716963309375488 | Organic | David Mezzetti | 10/20/2024 | All followers | 461 | 8 | 0.017354 | 3 | 0 | 2 | 0.0282 |
PSA from txtai: RAG doesn't have to be hard. LLM frameworks don't have to be convoluted and poorly written.
💯 Have a great weekend.
https://lnkd.in/eExBX_3A
| https://www.linkedin.com/feed/update/urn:li:activity:7253077188688695297 | Organic | David Mezzetti | 10/18/2024 | All followers | 555 | 21 | 0.037838 | 10 | 0 | 1 | 0.057658 |
Want to use ONNX models through Sentence Transformers with txtai? No problem!
https://lnkd.in/ehsEizjW | https://www.linkedin.com/feed/update/urn:li:activity:7252696467696422913 | Organic | David Mezzetti | 10/17/2024 | All followers | 287 | 11 | 0.038328 | 5 | 0 | 1 | 0.059233 |
There are more RAG frameworks than ever. If you want to see how txtai compares, then check out this article.
https://lnkd.in/eu4frpZ6
| https://www.linkedin.com/feed/update/urn:li:activity:7252284910332166145 | Organic | David Mezzetti | 10/16/2024 | All followers | 343 | 7 | 0.020408 | 5 | 0 | 1 | 0.037901 |
Generative Audio with txtai
Learn how txtai 7.5 supports Speech to Speech RAG and Generative Audio
https://lnkd.in/dPmt-bZm | https://www.linkedin.com/feed/update/urn:li:activity:7252014551863336960 | Organic | David Mezzetti | 10/15/2024 | All followers | 521 | 13 | 0.024952 | 11 | 0 | 1 | 0.047985 |
LLMs are more artist than analyst. What's called a hallucination in business is good in the creative space. Imagine a world where with a few words we can generate stories with automated voice narration and music. We're trending in that direction and it's applications are plentiful.
We're on a path where creativity will be in the hands of many.
https://lnkd.in/e7kd8_TE | https://www.linkedin.com/feed/update/urn:li:activity:7251955738476965888 | Organic | David Mezzetti | 10/15/2024 | All followers | 296 | 2 | 0.006757 | 1 | 0 | 1 | 0.013514 |
🚀 Start your week with the txtai 7.5 release!
txtai 7.5 adds Speech to Speech RAG, new TTS models and Generative Audio features. The latest version of the txtai RAG application is also available on Docker Hub.
See below for more.
GitHub: https://lnkd.in/dxWDeey
Release Notes: https://lnkd.in/eWuMe4FW
PyPI: https://lnkd.in/eE_Jvft
Docker Hub: https://lnkd.in/e598zTHb | https://www.linkedin.com/feed/update/urn:li:activity:7251669798709772289 | Organic | David Mezzetti | 10/14/2024 | All followers | 853 | 24 | 0.028136 | 13 | 0 | 1 | 0.044549 |
Generative storytelling Part 2
Reading "The Night Before Christmas" with generated background music using the story text!
Article: https://lnkd.in/e7kd8_TE | https://www.linkedin.com/feed/update/urn:li:activity:7251203271141289985 | Organic | David Mezzetti | 10/13/2024 | All followers | 400 | 19 | 0.0475 | 3 | 0 | 1 | 0.0575 |
📚 New from txtai - Generative storytelling!
Check out these Generative Audio workflows that takes a story ("The Raven" by Edgar Allan Poe) and joins speech with generated background audio. The audio is generated with a LLM building a prompt for a music generation model.
Article: https://lnkd.in/e7kd8_TE | https://www.linkedin.com/feed/update/urn:li:activity:7251202372645851136 | Organic | David Mezzetti | 10/13/2024 | All followers | 456 | 35 | 0.076754 | 3 | 1 | 1 | 0.087719 |
🤯 My AI generated voice reading a poem with AI generated music in the background. All with a txtai RAG workflow! | https://www.linkedin.com/feed/update/urn:li:activity:7247986217286299651 | Organic | David Mezzetti | 10/04/2024 | All followers | 331 | 14 | 0.042296 | 1 | 0 | 1 | 0.048338 |
See how a RAG process can write a poem about Machine Learning!
This example runs a vector search with Wikipedia, finds the best matching articles to use as context and runs a prompt to generate the poem using that context. The answer is then converted to speech!
Can you guess who the generated voice is? 😂 This is all machine generated!
Video on YouTube: https://lnkd.in/ez_ABNvG | https://www.linkedin.com/feed/update/urn:li:activity:7247959534634201089 | Organic | David Mezzetti | 10/04/2024 | All followers | 518 | 28 | 0.054054 | 5 | 1 | 2 | 0.069498 |
🔥 Check out this snippet of audio from the recently released Speech to Speech RAG workflow with txtai.
This is blending the facts from a knowledge base with the creativity of a LLM!
Full video on YouTube: https://lnkd.in/eEGGsXBQ | https://www.linkedin.com/feed/update/urn:li:activity:7246708419729006592 | Organic | David Mezzetti | 10/01/2024 | All followers | 422 | 41 | 0.097156 | 3 | 0 | 1 | 0.106635 |
If you're running on constrained hardware and/or with no GPU, LLaMA 3.2 1B models with 4-bit quantization work surprisingly well! The model is only 800MB.
Here's an easy-to-try example with txtai!
Example: https://lnkd.in/etd5w-9Q | https://www.linkedin.com/feed/update/urn:li:activity:7246583112908906496 | Organic | David Mezzetti | 09/30/2024 | All followers | 705 | 32 | 0.04539 | 8 | 2 | 1 | 0.060993 |
Heading into October🍂 - it's been quite a year for txtai. Just this year alone, txtai added:
- llama.cpp support
- LLM API support (i.e. OpenAI, Claude, Bedrock etc)
- Graph RAG
- Binary quantization and support for Matryoshka Embeddings
- Postgres integration for all vector database components
- Streaming LLM, Streaming RAG and chat message support
- Significant text extraction improvements
There's much more but these are the most crucial changes. It's hard to remember a time when txtai didn't have these features but it wasn't long ago!
https://lnkd.in/dxWDeey | https://www.linkedin.com/feed/update/urn:li:activity:7246089678221717504 | Organic | David Mezzetti | 09/29/2024 | All followers | 857 | 30 | 0.035006 | 8 | 3 | 1 | 0.049008 |
A great thing about txtai's new Speech to Speech RAG workflow is that it's modular. It's simple to swap out a local LLM for a LLM API, same for transcription and text to speech. Want RAG with Postgres + pgvector? No problem! That all comes with txtai and it's why it's the "all-in-one embeddings database"
https://lnkd.in/eEH-gZz2
| https://www.linkedin.com/feed/update/urn:li:activity:7245734487353880576 | Organic | David Mezzetti | 09/28/2024 | All followers | 359 | 5 | 0.013928 | 5 | 0 | 1 | 0.030641 |
💥 Dropping one of the most powerful and capable workflows txtai has to date: Introducing the Speech to Speech RAG workflow!
Lots of hard work went into this from end-to-end and we're confident txtai is the easiest way to build your own Speech to Speech RAG workflow. With the simplicity of txtai, you can swap in your own Embeddings database and be off the the races 🏇 - enjoy!
Video
https://lnkd.in/eEGGsXBQ
Article
https://lnkd.in/eEH-gZz2 | https://www.linkedin.com/feed/update/urn:li:activity:7245465896457052162 | Organic | David Mezzetti | 09/27/2024 | All followers | 1,230 | 123 | 0.1 | 22 | 11 | 2 | 0.128455 |
Want your TextToSpeech (TTS) pipeline to sound British 🇬🇧? Then check out this model with over 100 variations of English accents!
https://lnkd.in/e_gg4nfD
| https://www.linkedin.com/feed/update/urn:li:activity:7244670426629517312 | Organic | David Mezzetti | 09/25/2024 | All followers | 514 | 21 | 0.040856 | 9 | 0 | 1 | 0.060311 |
Want a Hugging Face dataset with the most recent (September 2024) copy of Wikipedia?
https://lnkd.in/eZsaMb3A
| https://www.linkedin.com/feed/update/urn:li:activity:7242134292942761984 | Organic | David Mezzetti | 09/18/2024 | All followers | 289 | 3 | 0.010381 | 1 | 0 | 1 | 0.017301 |
🔥 Check out this RAG with CoT + Self-Reflection example using the Wikipedia Embeddings index from txtai!
The release of OpenAI's o1 model has many trying to glean how it works without knowing for sure since it's a closed model. There is much speculation that CoT + Self-Reflection is part of the process.
Code: https://lnkd.in/eyFzupDq | https://www.linkedin.com/feed/update/urn:li:activity:7241820975690838017 | Organic | David Mezzetti | 09/17/2024 | All followers | 944 | 40 | 0.042373 | 11 | 0 | 2 | 0.056144 |
Want to build a RAG system at scale with your own data? Then focus on the R in RAG - Retrieval!
Vector storage is one of the major challenges at scale. Vectors tend to take up more space than the data itself. But there are options available such as quantization and concepts like Matryoshka Embeddings.
Check out the following articles to learn more.
Vector Quantization: https://lnkd.in/db5Jx_fE
Matryoshka Embeddings: https://lnkd.in/gr25HsBF | https://www.linkedin.com/feed/update/urn:li:activity:7240680798092165120 | Organic | David Mezzetti | 09/14/2024 | All followers | 595 | 18 | 0.030252 | 4 | 0 | 2 | 0.040336 |
✅ Check out the NeuML fall newsletter. This issue covers all the developments over the last couple of months!
https://lnkd.in/eqcBMAXj | https://www.linkedin.com/feed/update/urn:li:activity:7240412660146532353 | Organic | David Mezzetti | 09/13/2024 | All followers | 275 | 6 | 0.021818 | 1 | 0 | 1 | 0.029091 |
🔥 Happy to release the September 2024 version of txtai's Wikipedia Embedding indexes!
Link on HF Hub: https://lnkd.in/e4newZeM
| https://www.linkedin.com/feed/update/urn:li:activity:7240380020227997696 | Organic | David Mezzetti | 09/13/2024 | All followers | 407 | 19 | 0.046683 | 4 | 0 | 1 | 0.058968 |
🚀 Want to try the new OpenAI o1 model with txtai? No problem! | https://www.linkedin.com/feed/update/urn:li:activity:7240341524104839168 | Organic | David Mezzetti | 09/13/2024 | All followers | 686 | 14 | 0.020408 | 11 | 0 | 1 | 0.037901 |
From prototyping, small-scale to enterprise production, txtai has you covered. One interface gives easy access to the following:
Vector Search:
✅ In-memory
✅ Local indexes
✅ Postgres (via pgvector)
LLM / RAG Inference:
✅ Local Hugging Face models
✅ llama.cpp
✅ GPT-4, Claude, Bedrock, Cohere etc
https://lnkd.in/emd_5mNr | https://www.linkedin.com/feed/update/urn:li:activity:7239997328827297793 | Organic | David Mezzetti | 09/12/2024 | All followers | 273 | 5 | 0.018315 | 4 | 0 | 1 | 0.03663 |
⚡ Looking to build with txtai on AWS? Then check out this reference architecture.
Code: https://lnkd.in/eWjbci_k | https://www.linkedin.com/feed/update/urn:li:activity:7239656871681306624 | Organic | David Mezzetti | 09/11/2024 | All followers | 530 | 16 | 0.030189 | 4 | 3 | 1 | 0.045283 |
☁️ Want an AWS-hosted version of txtai? Then check out this RAG example! It uses AWS Bedrock for embeddings + LLM calls. Content is stored in Postgres/pgvector via Aurora or RDS.
Other frameworks like LangChain and LlamaIndex require code changes to switch from local to cloud. The same code can handle both with minor configuration changes in txtai!
Code: https://lnkd.in/eWjbci_k | https://www.linkedin.com/feed/update/urn:li:activity:7239623524896690176 | Organic | David Mezzetti | 09/11/2024 | All followers | 456 | 17 | 0.037281 | 8 | 1 | 1 | 0.059211 |
Want a vector embeddings model trained for medical literature?
https://lnkd.in/egnEKcqd
| https://www.linkedin.com/feed/update/urn:li:activity:7239241118700179457 | Organic | David Mezzetti | 09/10/2024 | All followers | 426 | 11 | 0.025822 | 3 | 0 | 1 | 0.035211 |
Curious about Graph RAG? Did you know that txtai has an ready-to-use Embeddings graph database for popular Wikipedia articles on the HF Hub?
https://lnkd.in/e3fPr6fd | https://www.linkedin.com/feed/update/urn:li:activity:7238920785522798594 | Organic | David Mezzetti | 09/09/2024 | All followers | 490 | 10 | 0.020408 | 7 | 0 | 1 | 0.036735 |
🔥 Learn about txtai's embedding index format for open data access.
https://lnkd.in/eh4-rVEr | https://www.linkedin.com/feed/update/urn:li:activity:7237881069146902528 | Organic | David Mezzetti | 09/06/2024 | All followers | 403 | 9 | 0.022333 | 4 | 0 | 1 | 0.034739 |
💯🤯 Big time release with txtai 7.4!
txtai 7.4 adds the SQLite ANN, new text extraction features and a programming language neutral embeddings index format. The latest version of the txtai RAG application is also available on Docker Hub.
The embeddings index format is a contract to enable open data access in a programmatic and platform independent way. New ways to bind index components will be coming soon!
See below for more.
GitHub: https://lnkd.in/dxWDeey
Release Notes: https://lnkd.in/eDGp3dAi
PyPI: https://lnkd.in/eE_Jvft
Docker Hub: https://lnkd.in/e598zTHb | https://www.linkedin.com/feed/update/urn:li:activity:7237636783453220864 | Organic | David Mezzetti | 09/06/2024 | All followers | 877 | 28 | 0.031927 | 12 | 0 | 1 | 0.04675 |
🔥 Coming soon - txtai stores vectors in SQLite via sqlite-vec!
https://lnkd.in/dDi9_4FF | https://www.linkedin.com/feed/update/urn:li:activity:7237136492538548224 | Organic | David Mezzetti | 09/04/2024 | All followers | 406 | 9 | 0.022167 | 3 | 0 | 1 | 0.03202 |
With LLMs growing more and more powerful⚡, one of the most important tasks is ensuring the best content is provided.
Did you know that txtai has a robust text extraction pipeline that can convert HTML, PDF, DOCX, XLSX and more to LLM-friendly Markdown?
Read more at the links below.
https://lnkd.in/en-wT8zT
https://lnkd.in/eSmxjk7t
| https://www.linkedin.com/feed/update/urn:li:activity:7234997730945691648 | Organic | David Mezzetti | 08/29/2024 | All followers | 505 | 20 | 0.039604 | 3 | 0 | 1 | 0.047525 |
AI/ML/NLP workflows? 🥱 txtai had that back in 2021.
https://lnkd.in/eDj8NZtb | https://www.linkedin.com/feed/update/urn:li:activity:7233815946593726465 | Organic | David Mezzetti | 08/26/2024 | All followers | 370 | 6 | 0.016216 | 3 | 0 | 1 | 0.027027 |
txtai is an all-in-one embeddings database for semantic search, LLM orchestration and language model workflows. It uses a number of index formats to store data with each of it's components.
Each component is designed to ensure open access to the underlying data in a programmatic and platform independent way. The next release is making some backwards-compatible changes to ensure this is the case for all components.
Read more here: https://lnkd.in/eiX-WtQn | https://www.linkedin.com/feed/update/urn:li:activity:7233812131312201729 | Organic | David Mezzetti | 08/26/2024 | All followers | 465 | 8 | 0.017204 | 3 | 0 | 1 | 0.025806 |
If you're evaluating LangChain, LlamaIndex and/or Chroma DB for your projects, then check out this article that contrasts those libraries with txtai.
https://lnkd.in/eu4frpZ6
| https://www.linkedin.com/feed/update/urn:li:activity:7231647335338909696 | Organic | David Mezzetti | 08/20/2024 | All followers | 606 | 21 | 0.034653 | 8 | 0 | 1 | 0.049505 |
txtai RAG v0.4 is out! This release adds a couple new parameters to change the context and max generation sizes.
GitHub: https://lnkd.in/evdB5HgN
Docker Hub: https://lnkd.in/e-3Cx_68
| https://www.linkedin.com/feed/update/urn:li:activity:7231458302478413824 | Organic | David Mezzetti | 08/20/2024 | All followers | 531 | 25 | 0.047081 | 4 | 0 | 1 | 0.056497 |
End of preview. Expand
in Dataset Viewer.
NeuML LinkedIn Company Posts
This dataset is 12 months of NeuML's LinkedIn Company Posts as of January 2025. It contains the post text along with engagement metrics.
It was created as follows:
Export the company posts from the analytics page, see this link for instructions.
Run the following code to create a dataset
import pandas as pd
from datasets import load_dataset
df = pd.read_excel("export_data.xls", sheet_name=1, header=1)
df = df.dropna(axis="columns")
df.to_csv("data/posts.csv", index=False)
This same process can be run to create your own dataset. It can be loaded locally as follows.
ds = load_dataset("data")
- Downloads last month
- 53
Size of downloaded dataset files:
118 kB
Size of the auto-converted Parquet files:
61.4 kB
Number of rows:
258
Models trained or fine-tuned on NeuML/neuml-linkedin-202501