WobblyJellybean
WobblyJellybean

Update: open-source PerplexityAI

Hey fellow comers, just showing my progress. I will launch by next month hopefully.

I've completed my work for:

  1. ⁠Streaming text
  2. ⁠Formatted responses
  3. ⁠Citations and websites

Currently, I'm working on finishing:

  1. ⁠Chat history
  2. ⁠Documents library
  3. ⁠LLM settings

I'm using the Vercel AI SDK, Next.js, Firebase, and Bing to ensure setting up and running the project is as straightforward as possible. I hope to support more LLMs, like Claude, Mistral, and Gemini, to offer a mix-and-match approach.

Although I've accomplished a lot, there are still a few more weeks of work ahead. Unfortunately, I've failed to raise any funds for my project and am fully dependent on the open-source community for support.

Note: VCs told me I can't build perplexity so simply because I don't have that much skills or high enough pedigree. They are literally blinded by the fact that any average dev can also build such an app.

8mo ago
Find out if you are being paid fairly.Download Grapevine

What's the point of open source here? The llm is closed, the data is proprietary. Why do we care if the outer shell is open or close?

WobblyJellybean
WobblyJellybean

Whatever I can help with. You really think I can build the next version of LLM or something?

I am trying to share as much as possible

PeppyMarshmallow
PeppyMarshmallow

For as much shit they get, VC’s aren’t that stupid you know. They see if you dont posses deep background in AI, you might not able to handle when things go wrong .

Especially in such pure consumer SaaS where you aren’t showing any moat

Regardless, I like what you are trying to do, perplexity has heavy western bias in searches too.

If you would wish to build defensibility in your project, and make it venture friendly, you can DM me, we can brainstorm ideas (ex. grounding answers with Knowledge graphs, agent based search and answer, adding deep multilinguallity and finetuning with datasets like Aya)

Am working as research scientist with 6 YOE in ML

SwirlyTaco
SwirlyTaco

I didn't understand, how is this different from regular chatGPT? Who's the target audience for this product?

WobblyJellybean
WobblyJellybean

Competition for Perplexity. Mix and match any other LLM and get realtime data

WobblyJellybean
WobblyJellybean

Let me know if you guys love it.

Discover more
Curated from across