Scolring - Forum

I tested AI chat using GPT models on v1.96 beta1.
Some GPT models work but the response is too slow and the answers are incomplete.
Some GPT models don't work at all.

In v1.95, Gemini model works better than GPT model.
The answer given by Gemini is faster and not truncated and always smooth.
The answer given by GPT is too slow and sometimes the answer does not come out.

Responses API has not yet been implemented in the OpenAI compatible endpoint for Gemini.

I've tried but it has no response.

Better to include Gemini AI as well. Gemini is on the rise.

Let's download & try the latest version!

Let's download!!

I think in the custom interface plugit there is an option to create a new interface or default chat interface that has been created so that all users don't have to create it repeatedly but users can still add other elements they want.

I think we need a new chat GUI for phone. please look at the image below.
https://drive.google.com/file/d/1WTnwSF … drive_link

Then, Is it possible to use this TTS and STT in openspace3d?
https://huggingface.co/collections/meso … 4c6673ed13
https://huggingface.co/collections/meso … 44f0cfae79

I forgot to inform you.

Chatgpt ask to make sure:

You don't allow ChatGPT to "hallucinate" or fabricate facts

Use specific prompts such as:

“Answer this question based only on the attached document. If there is no information in the document, answer: ‘There is no information in the document to answer this question.’”

I just want to make a simple AI agent app. I prepare a document then AI analyzes the file. So, the end user can just ask questions and AI just answers based on the document.

Chatgpt suggest like below.

Self-Hosted Lightweight AI Agent with Open-Source Tools + OpenAI API Minimum Use

? Focus:
-Documents uploaded & provided by you (developer)

-AI answers based only on documents

-Minimize API calls (save costs)

-Use free/open-source tools as much as possible

?️ CHEAP OPTIONS: SAVING TECHNOLOGY COMBINATION
Cheap / Free Optional Components
Vector Search ? ChromaDB (local) or FAISS
Backend ? Flask (Python) or FastAPI
AI Model ? OpenAI gpt-3.5-turbo (cheap + power)
Mobile App ? Flutter or React Native
Document Hosting ? Store on your own server (or free Firebase)
Embedding ? OpenAI Embedding API (text-embedding-3-small) – very cheap

? HOW IT WORKS (Simple + Save)
Step 1: Prepare Developer Documents
Save all your content (articles, notes, textbooks) in .txt or .md

Split into topic/chapter

Step 2: Embedding Once

Use text-embedding-3-small (OpenAI)

Save in ChromaDB (local vector db)

? Embedding is cheap and only needs to be done ONCE — not every time a user asks

Step 3: When a User Asks a Question
Find the most relevant content snippet from vector DB (top-k = 3)

Recombine as prompt + question

Send to gpt-3.5-turbo

Step 4: Answers sent to app
? ESTIMATED COST (MONTHLY)
Cost Items
Embedding (100k tokens) ± $0.01
GPT-3.5 API (1,000 answers) ± $1.50 – $3
Hosting + VectorDB (local) $0 (if self-hosted)
Firebase / Render (backend) $0 – $5 (free tier available)

Thank you so much.
Is it possible to upload document then chatgpt will respond based on the document/topic?

Hi Arkeon, I get this error while exporting app bundle (java). maybe missing file. take a look this link
https://drive.google.com/file/d/1k1YUf8 … drive_link