ragdoll-studio VS llama_index

- Basic internet search (I start ollama CLI faster than I can start a browser - https://ollama.com)

- Formatting/changing text

- Troubleshooting code, esp. new frameworks/libs

- Recipes

- Data entry

- Organizing thoughts: High-level lists, comparison, classification, synonyms, jargon & nomenclature

- Learning esp. by analogy and example

RAG for:

- Website assistants (https://github.com/bennyschmidt/ragdoll-studio/tree/master/e...)

- Game NPCs (https://github.com/bennyschmidt/ragdoll-studio/tree/master/e...)

- Discord/Slack/forum bots (https://github.com/bennyschmidt/ragdoll-studio/tree/master/e...)

- Character-driven storytelling and creating art in a specific style for video game loading screens, background images, avatars, website art, etc. (https://github.com/bennyschmidt/ragdoll-studio/tree/master/r...)

> I always say I want AI to work offline

I'm with you, I'm most excited about this too.

Currently building an AI creative studio (make stories, art, music, videos, etc.) that runs locally/offline (https://github.com/bennyschmidt/ragdoll-studio). There is a lot of focus on cloud with LLMs but I can't see how the cost will make much sense for involved creative apps like video creation, etc. Present day users might not have high-end machines, but I think they all will pretty soon - this will make them buy them the way MMORPGs made everyone buy more RAM. Especially the artists and creators. Remember, Photoshop was once pretty difficult to run, you needed a great machine.

I can imagine an offline music/movies apps, offline search engines, back office software, etc.

I built a JavaScript library for creating LLM-based game NPCs, here's a video demo: https://github.com/bennyschmidt/ragdoll-studio/tree/master/e...

You drop all their knowledge into a file, like .txt. or .html and it uses RAG to learn everything in the file, so if it's a vendor things like shop inventory, prices, or if it's a storyteller/questgiver you can give it knowledge the player must uncover, etc.

It uses the LLM for natural sounding conversations and instructions ("speak in the style of..." etc.)

> You would also need to load an enormous amount of precedential case law

Very easily done. Is that it?

> lack of common sense, false conclusions

The AI tool doesn't replace the judge/DA/etc. it's just a very useful tool for them to use. Checkout the "RAG-based learning" section of this app I built (https://github.com/bennyschmidt/ragdoll-studio) there's a video that shows how you can effectively load new knowledge into it (I use LlamaIndex for RAG). For example, past cases that set legal precedents, and other information you want to be considered. It creates a database of the files you load in, so it's not making those assumptions like an LLM without RAG would. I think a human would be more error-prone than an LLM with vector DB of specific data + querying engine.

> I don't see the utility

Then you are not paying attention or haven't used LLMs that much.

> actual work involved in practicing law

This is what it's best at, and what people are already using RAG for: Reading patient medical docs, technical documentation, etc. this is precisely what humans are bad at and will offload to technology.

> actual research is required

You have not tried RAG.

> LLMs struggle to produce useful outputs

You have not tried RAG.

> LLMs are unlikely to identify issues

You have not tried RAG.

> the LLM by definition is creative analysis

You have not tried RAG.

I've made AI assistants that are perfectly accurate with products, pricing, etc. yet still maintain a human quality: https://github.com/bennyschmidt/ragdoll-studio/tree/master/e...

You can accomplish this with RAG.

Your overall point is taken though, the LLM itself is not enough, fine-tuning is not always feasible, and I think no matter how good an AI persona gets at, say, teaching yoga - for some yoga students it will never replace an in-person instructor.

However for a game NPC, online agent, Discord bot, etc. not to mention research, translation, tutorials, summarizing, etc. there is a lot of present day utility for LLMs.

Very cool app, if anyone wants to try the FOSS alternative to this check out Ragdoll Studio - https://github.com/bennyschmidt/ragdoll-studio

Create anyone you want based on a list of documents (URLs), chat with them, have them create images in their style, tell stories, etc.

(runs open models with Ollama, Stable Diffusion, etc. not OpenAI)

I would be curious what Sam Altmans people make, you can share characters here: https://ragdoll-studio.vercel.app

It's a FOSS alternative because instead of using their online platform to make an account in order to interact with their characters, you instead download and run this source code without needing an account to interact with yours, and those in the community - even offline if you want.

There's more to it than just chatting with AI models - these "ragdolls" are based on RAG so they are able to have distinct personalities and scoped knowledge. They can't leak info they don't know about, can't be censored from talking about things they do know about, and can be exported and shared with others on the community site. The model itself is more like the underlying engine, where the ragdoll is a specific persona that you can deploy to a variety of tasks (starting with just chat).

I want to go beyond what character.ai is even dreaming of doing by making this into a creative suite for all things AI: Chat, storytelling, art & visual design, cinematography, CGI, music & sound effects, and so-on. It would be great to have an open-source character-focused tool for making stories, cinematics, films, music, etc. Here's an example that uses a ragdoll as an NPC in a game that has specific knowledge that the player can uncover in order to advance in the game: https://github.com/bennyschmidt/ragdoll-studio/tree/master/e...

Thanks for the question, I will add a section on how these are fundamentally different from just chatting with ChatGPT.

For anyone curious llamaindex's "prompt mixins", they're actually dead simple: https://github.com/run-llama/llama_index/blob/8a8324008764a7... - and maybe no longer supported.

I basically reinvented this wheel in ragdoll but made it more dynamic: https://github.com/bennyschmidt/ragdoll/blob/master/src/util...

Being able to plug third party frameworks (Langchain, LlamaIndex) so you can build complex projects

Mistral Instruct does use a system prompt.

You can see the raw format here: https://www.promptingguide.ai/models/mistral-7b#chat-templat... and you can see how LllamaIndex uses it here (as an example): https://github.com/run-llama/llama_index/blob/1d861a9440cdc9...

Learn how to use Milvus as persistent vector storage with LlamaIndex in under 5 minutes.

▶️ Read Blog 📷 Watch Demo 🦙 Notebook using Pipelines inside LlamaIndex

15. LlamaIndex | Github | tutorial