AI RAG Application Website Development
Budget: ₹12,500 – ₹37,500 INR
I'm looking to develop a python flask based RAG application website designed for AI LLM. The core function of this application will be question answering similar to ChatGPT.com
Key Features:
- A web-based GUI, as opposed to a command-line interface or just an API. This GUI should be user-friendly and intuitive, allowing users to easily navigate through the application.
- A website functionality very similar to ChatGPT.com including the UI and message boxes etc
- 5 free questions for unregistered users
- paid users get xx questions per month (value changable by me in the script)
- a good modern slick polished UI using any tech stack
- I have plain text files, which need to be chunked and uploaded to Zilliz or Qdant (you can either install it on my VPS or use cloud). Need a python script where i have a placeholder to input the folder path and the script must get all text files in it (including subdirectories) and chunk it according to the given number of sentences (example every 5 sentences = 1 chunk) + option to chunk by /n/n paragraph (according to the plain text file i choose) and embed using Google multilingual 002 (latest)
- also for Hybrid search need sparse vector embeddings
- file metadata should also be uploaded
-- the website UI will have a few small modifications which will be talked about in private chat with me.
Process flow: user asks question --> uses LLM 1 (google flash 2) to paraphrase (according to my prompt) ---> gets retrieved chunks from vector DB hybrid search (uses metadata filtering based on user question decided by Instruction prompt) ----> uses google ReRanker ----> sends everything to LLM2 (google flash 2 or google gemini 2 pro) to generate answer (according to system prompt) ---> displays the answer to the user.
- You can use Google Vertex or Google RAG engine (if Gemini 2.0 models are available - please make sure beforehand)
- i already have domain name and VPS ready with python installed and tested a sample flask app which works.
-- Server is Cloudlinux 8 with Cpanel and WHM.
--- It's up to you if you want to use AI tools to generate the code for this project but the final website must be in fully working condition.
--- you can demonstrate the final thing with a sample text file and i will try it myself once it is ready.
***** NO FINETUNING OR TRAINING ETC of models wanted******
Ideal Skills and Experience:
- Proficient in web development, with a strong portfolio of creating interactive, user-friendly web-based applications. Python, Linux, JS, HTML
- Prior experience working with RAG applications or AI LLMs will be highly advantageous.
- Excellent understanding of GUI design principles.
- Strong problem-solving skills, able to deliver a solution that meets the project requirements.
-- WE WILL NOT BE USING CHATGPT ASSISTANTS wrappers etc
Key Features:
- A web-based GUI, as opposed to a command-line interface or just an API. This GUI should be user-friendly and intuitive, allowing users to easily navigate through the application.
- A website functionality very similar to ChatGPT.com including the UI and message boxes etc
- 5 free questions for unregistered users
- paid users get xx questions per month (value changable by me in the script)
- a good modern slick polished UI using any tech stack
- I have plain text files, which need to be chunked and uploaded to Zilliz or Qdant (you can either install it on my VPS or use cloud). Need a python script where i have a placeholder to input the folder path and the script must get all text files in it (including subdirectories) and chunk it according to the given number of sentences (example every 5 sentences = 1 chunk) + option to chunk by /n/n paragraph (according to the plain text file i choose) and embed using Google multilingual 002 (latest)
- also for Hybrid search need sparse vector embeddings
- file metadata should also be uploaded
-- the website UI will have a few small modifications which will be talked about in private chat with me.
Process flow: user asks question --> uses LLM 1 (google flash 2) to paraphrase (according to my prompt) ---> gets retrieved chunks from vector DB hybrid search (uses metadata filtering based on user question decided by Instruction prompt) ----> uses google ReRanker ----> sends everything to LLM2 (google flash 2 or google gemini 2 pro) to generate answer (according to system prompt) ---> displays the answer to the user.
- You can use Google Vertex or Google RAG engine (if Gemini 2.0 models are available - please make sure beforehand)
- i already have domain name and VPS ready with python installed and tested a sample flask app which works.
-- Server is Cloudlinux 8 with Cpanel and WHM.
--- It's up to you if you want to use AI tools to generate the code for this project but the final website must be in fully working condition.
--- you can demonstrate the final thing with a sample text file and i will try it myself once it is ready.
***** NO FINETUNING OR TRAINING ETC of models wanted******
Ideal Skills and Experience:
- Proficient in web development, with a strong portfolio of creating interactive, user-friendly web-based applications. Python, Linux, JS, HTML
- Prior experience working with RAG applications or AI LLMs will be highly advantageous.
- Excellent understanding of GUI design principles.
- Strong problem-solving skills, able to deliver a solution that meets the project requirements.
-- WE WILL NOT BE USING CHATGPT ASSISTANTS wrappers etc