Uminal – Large language model anyone can extend with new apps/skills

Hacker News

Hi HN! I’ve been working on a project called Uminal, which lets anyone extend GPT-3 with new apps & capabilities that auto-compose with each other. Short demo video: https://www.loom.com/share/211327fd8e854513b909b0f69eadd2f8 Online prototype: https://uminal.com These “apps” are similar to “Alexa skills” but they’re auto-composable (and a bit different in other ways), and enable GPT-3 to interact w/ the real world. These apps could be anything (as long as it’s accessible via an API endpoint). For example, these apps could be other LLMs (e.g., Galactica), apps built w/ LLMs (e.g., using something like LangChain [0] or Dust.tt [1]), or just plain old APIs (such as web search). Anyone can add new apps to Uminal by telling it how & when it should use the app’s API endpoint, and users can browse & enable apps for their account. Then, when processing a user’s input prompt, Uminal will automatically call any of the user’s enabled apps as needed, & compose their inputs/outputs together. It will fallback to GPT-3 whenever needed, as part of processing a prompt. It currently uses MRKL + GPT-3’s zero-shot reasoning to accomplish this, but I’m working to let users also teach it how to process sample input prompts, which would hopefully improve Uminal’s reasoning ability for everyone, using few-shot examples in a scalable manner. Can’t wait to try this on top of GPT-4! The project’s vision is to build an “intelligent computer” by letting anyone teach/extend LLMs to interact w/ the real world, similar to how iOS apps really expanded the capabilities of the iPhone. Would really love to hear any ideas/thoughts/honest feedback, etc.! My Twitter DMs are also open: https://twitter.com/subby_tech Feel free to join our Discord, if you have any questions/feedback or just wanna chat, etc: https://discord.gg/qczbX5rzD8 Project docs/readme: https://dent-mirror-a40.notion.site/Uminal-1d6b4b8fe22b42b28... Source code for the sample Uminal apps (web search, etc.): https://github.com/thecooltechguy/uminal-sample-apps Notes on using the online prototype: Once you sign in, go to the Apps page and enable "Web search" and "Work Q&A". The “Web search” app uses SerpApi to give GPT-3 internet access, while the "Work Q&A" app integrates with a sample company's Notion database to answer work-related questions (e.g., can I work from home?, etc.). Due to how SerpAPI works, if any queries involving web searches fail, you can try re-running them and it will usually work. Again, this is a super rough prototype hacked together in a few days, but would love to hear your honest thoughts/feedback/ideas! Thanks! :D [0] https://github.com/hwchase17/langchain/ [1] https://dust.tt — https://github.com/dust-tt/dust

Visit product View on Hacker NewsNov 26, 2022

Share card

Actual performance

3points

2comments

Did not reach leaderboard

Launch Intel predictions

Analyze your own launch →

Product HuntOn track for Day 1 leaderboard · Strong signals: model, apps, user · Missing: mac, agents, macos

98%98% predicted probability of success on Product Hunt, based on ML models trained on real launch data.

best fitHighest predicted score across all platforms for this description.

Indie HackersFits the IH revenue-focused audience · Strong signals: ios · Missing: supports, reddit linkedin, podcasting

96%96% predicted probability of success on Indie Hackers, based on ML models trained on real launch data.

Hacker NewsStrong engagement from HN community · Strong signals: ide, io · Missing: https docs, excited, just released

74%74% predicted probability of success on Hacker News, based on ML models trained on real launch data.

nativeThis product was originally launched on this platform.

TrustMRRFits verified-revenue profile · Strong signals: ios, apps, video · Missing: mobile apps, personal, entrepreneurs

56%56% predicted probability of success on TrustMRR, based on ML models trained on real launch data.

AppSumoMay struggle as an AppSumo deal · Strong signals: users · Missing: plus, platform, intuitive

45%45% predicted probability of success on AppSumo, based on ML models trained on real launch data.

Acquire.comPre-revenue stage for this audience · Missing: arr, mrr, revenue

12%12% predicted probability of success on Acquire.com, based on ML models trained on real launch data.

BetaListMay not resonate with beta-testers · Strong signals: chat, real world · Missing: web3, crypto, cryptocurrency

0%0% predicted probability of success on BetaList, based on ML models trained on real launch data.

Incorrect prediction on native model

Similar products

GEITje-7B – A New Large Open Dutch Language Model58%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

GEITje-7B – A New Large Open Dutch Language Model

Hacker News20

Personalizing Large Language Model Applications56%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

Personalizing Large Language Model Applications

Hacker News1

Refuel-LLM, a large language model for data annotation and enrichment67%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

Refuel-LLM, a large language model for data annotation and enrichment

Hacker News9

Extend ActiveRecord pluck to return array of hashes59%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

Extend ActiveRecord pluck to return array of hashes

Hacker News1

Sustinion - the large opinion collider (in spe)55%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

Sustinion - the large opinion collider (in spe)

Hacker News1

LLM-A* – Large Language Model Enhanced Incremental Heuristic Search60%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

LLM-A* – Large Language Model Enhanced Incremental Heuristic Search

Hacker News3

BrainTripping - language model comedy64%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

BrainTripping - language model comedy

Hacker News283

NCL – New Configuration Language56%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

NCL – New Configuration Language

Hacker News4

Strings62%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

The Large Language Messenger

Indie Hackers2ai

Using Twilio to extend the functionality of Siri54%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

Using Twilio to extend the functionality of Siri

Hacker News3

Browse all products Analyze your launch