Docuchat

Created a RAG using 3 free options:

• Mistral (full online)
• HuggingFace (LLM online, embeddings local)
• Ollama (local) It uses a python backend with an html/js/css frontend to interact with it.

Quick start

You will need Docker for all versions. If you don't have it, you can download it here.

Docker (recommended):

Open the compose folder inside the docker folder and follow the instructions

Local:

• You need python installed. We recommend creating a venv for the app, so it doesn't have problems with interdependencies with other python you might have.
• Make sure you have all the requirements needed.

Preparation

1. Download the repository

Download it manually or using the terminal:

git clone https://github.com/MiguelGP-13/Projects-RAG.git

2. Create .env file

Important

The env file must be in the RAG folder, not elsewhere

DB_PASSWORD=<DB_PASSWORD IN .env>
API_MISTRAL=<Mistral API>
API_HUGGINGFACE=<HuggingFace API>

Note

The only api key needed is the one you are going to use To get a Mistral API:

1. Create an account or login if you alredy have one at Mistral.
2. Go to the api key section and click on Create a new Key.
3. Copy the key and put it in the .env file.

3. Confirm settings.env content

Make sure it has this information if you choose Mistral or HuggingFace as MODE.

MODE=Mistral
CHUNK_SIZE=512
CONTEXT_SIZE=3
DOCUMENT_FOLDER=Backend/documents
QUESTIONNAIRES_FOLDER=Backend/questionnaires
CHATS_FOLDER=Backend/chats

If you want to use local LLM:

MODE=Local
MODEL=<Model you want to use (for example: mistral)> 
EMBEDDINGS=<Model you want to use (for example: jina/jina-embeddings-v2-base-en)> 
DIMENSION=<Dimension of embeddings generated (768 for jina embeddings)>
CHUNK_SIZE=512
CONTEXT_SIZE=3
DOCUMENT_FOLDER=Backend/documents
QUESTIONNAIRES_FOLDER=Backend/questionnaires
CHATS_FOLDER=Backend/chats

Important

Create the folders you are indicating in settings.env (documents, questionnaires and chats)

4. Create Redis database in docker

docker run -d --name redis-stack -p 6379:6379 -p 8361:8001 -e REDIS_ARGS="--requirepass <DB_PASSWORD IN .env>" redis/redis-stack:latest

Tip

If the command is not working, make sure Docker is alredy running

Steps

1. Start Redis in docker

Start the docker engine and the redis db

docker desktop start
docker start redis-stack

2. Run main.py file

Start the backend API

cd Backend
python main.py

3. Open the main.html file

Open the web so you can use the app Linux:

cd ../Frontend
xdg-open file.htm

Windows:

cd ../Frontend
start file.htm

Note

You can also open it clicking on the file

Frontend:

main.html: Chat where you ask questions about the documents uploaded.

files.html: Page to upload files for the RAG to use.

API endpoints:

Error response:

{"success": false, "error_code": "Code (int)", "description": "Error description"}

1. POST /query/ [ RAG / chat]

Payload:

{"prompt": "QUERY", "chat": "Chat selected"}

Answer:

Chat:

{"success": true, "answer": "LLM answer"}

RAG:

{"success": true, "answer": "LLM answer", "references": "Id of context peices", "reference_text": "Context pieces text"}

2. POST /remove_doc

Payload:

{"file_name": "filename", "clear_database": true}

Answer:

{
  "success": true,
  "description": "The file filename was deleted",
  "database_cleared": "n_chunks chunks deleted"
}

3. POST /update

Payload:

{"pdfs_array": ["PDF NAME 1", "PDF NAME 2", "..."]}

Answer:

{
  "success": true,
  "pdfs_created": ["PDF NAME 1", "PDF NAME 2", "..."],
  "number_of_chunks": "n_chunks new chunks were created"
}

4. POST /upload

Payload:
files[]: Array of .pdf files
Answer:

{
  "success": true,
  "pdfs_created": ["PDF NAME 1", "PDF NAME 2", "..."],
  "number_of_chunks": "n_chunks new chunks were created"
}

5. GET /files

Answer:

{"success": true, "files": ["PDF NAME 1", "PDF NAME 2", "..."]}

6. GET /chats

Answer:

{
  "success": true,
  "chats": [
    {
      "realName": "My Chat",
      "id": "hash",
      "creationDate": "timestamp"
    },
    ...
  ]
}

7. GET /chats/

Answer:

{
  "success": true,
  "chat": [ "Conversation saved" ]
}

8. POST /createChat/

Answer:

{"success": true, "id": "hash"}

9. POST /deleteChat/

Answer:

{"success": true, "deleted": "id.json"}

10. POST /createQuestionnaire

Payload:

{"pdfs_array":["PDF NAME 1", "PDF NAME 2", "..."],
"level":2, // From 1 to 4
"number_of_questions":5}

Answer:

{"success": true, "questionnaireId": 1234}

11. POST /deleteQuestionnaire/

Answer:

{"success": true, "deleted": "id.json"}

12. GET /questionnaires/

Answer:

{"success": true, 
"questions": [{"question":"",
                "options": ["..."],
                "correctOption":1}]
}

13. GET /questionnaires

Answer:

{"success": true, 
"questions": [{"question":"",
                "options": ["..."],
                "correctOption":1}]
}

14. POST /renameQuestionnaire/

Answer:

{"success": true, "newName": "newName", "oldName":"oldName"}

15. GET /

Answer: Serves main.html from frontend

16. /healt

Answer:

{"success": true}
{"success": false, "description":"What's not up running"}

Error codes:

Code	Description
0	Unexpected error
100	A compulsory value wasn't found in the JSON provided with the POST
101	File trying to add already exists
102	The path of the file provided doesn't exist
103	Error occurred while fetching files
104	No files arrived to the backend
105	The file you are trying to find does not exist
106	Conversation with that name already exists
107	Conversation name can't have dots (".")
108	Conversation name can't start with a number
109	Conversation name can't be empty
110	Incorrect query mode
111	The questionnaire you are trying to find does not exist