https://github.com/omarmdiab/chatgpt-clone
Chat with your local Documents (PDF, CSV, DOCX, TXT, PPTX)
https://github.com/omarmdiab/chatgpt-clone
huggingface langchain llm
Last synced: 3 months ago
JSON representation
Chat with your local Documents (PDF, CSV, DOCX, TXT, PPTX)
- Host: GitHub
- URL: https://github.com/omarmdiab/chatgpt-clone
- Owner: OmarMDiab
- Created: 2024-05-31T08:02:31.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2024-09-20T15:28:19.000Z (9 months ago)
- Last Synced: 2025-01-18T06:44:18.139Z (5 months ago)
- Topics: huggingface, langchain, llm
- Language: Python
- Homepage:
- Size: 99.6 KB
- Stars: 0
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Private ChatGPT Clone
Using huggingface and langchain this app offers: -
- **Model Switching:** Easily switch between different Huggingface models as per your requirements.
- **Document Chat:** Chat with your documents, including `(CSV, DOCX, PPTX, PDF)` files. Upload your documents and start querying them.
**Note:** You need to add an `.env` file with your huggingface api token with `(Permission: write)` or you can just run with your local model but make sure to uncomment the LlamaCpp code block in `get_conversation_chain` and pass your Model path
## Sample Run Videos
- [PDF Sample Run](https://drive.google.com/file/d/1pV-0HFy6f2Bj5J7dVN_MSCCbyjsbkS6i/view?usp=sharing)
- [CSV Sample Run](https://drive.google.com/file/d/11Zrc1USczjzVT6I5MQ9HQr68aot1Qr5i/view?usp=sharing)
## Easy Run ^^
In this Streamlit App you only just have to run the python script **`without`** the need of this code: `streamlit run script.py`
Just run: -
```bash
python Chatgbt_Ghost.py
```
## Models Used
- hkunlp/instructor-xl (For Embeddings)
- google/flan-t5-xxl
- google/flan-t5-small
- google/flan-t5-base
- mistral-7b-instruct-v0.2.Q4_K_S (For local runs using LlamaCpp)
## Some Rules
- Switching models mid conversation **resets** the chat_history (your past questions)
- Files can be embedded as **text_based files** or **CSV files** not both. This resitriction was made to improve Response accuracy.
- Chat history is Disabled with CSV files, As we are only retreiving Data So, To ensure that the retreival is at its best accuracy and not confusing the model with past prompts values.