AiSafe API Documentation

General Information

• Title: AI Safe API
• Version: 1.0
• Base URL: /mdp/ai-safe/

Authentication

• Type: API Key
• Key Name: Authorization
• Location: Header
• Description: JWT token for authorization

Endpoints

1. Fetch Chat History

• Method: GET
• URL: /history
• Description: Retrieves the list of chat names and their ids of the user.
• Responses:
• 200:

  • Description: List of chat names and ids retrieved successfully.

  • Example:

    {
      "success": true,
      "message": "History fetched successfully",
      "data": {
      "chat id 1": "chat name 1",
      "chat id 2": "chat name 2",
      "chat id 3": "chat name 3",
      "chat id 4": "chat name 4"
      }
    }

2. Fetch Chat History by Chat ID

• Method: GET
• URL: /history/{chat_id}
• Description: Retrieves the prompts for a specific chat for a user.
• Parameters:
  • chat_id (path, string, required): Chat ID to fetch history for.
• Responses:
• 200:

  • Description: Chat history retrieved successfully.

  • Example:

    {
        "success": true,
        "message": "History fetched successfully",
        "response_code": 200,
        "data": {
            "anonymized_values": {
                "Emails": ["Email1", "Email2"],
                "Names": ["Name1", "Name2"],
                "Phone": ["ph number 1", "ph number 2"]
            },
            "prompts": [
                {
                    "responseType": "text",
                    "originalPrompt": "Original prompt 1",
                    "anonymizedPrompt": "Anonymized prompt 1",
                    "response": "Anonymized response 1"
                }
                {
                    "responseType": "text",
                    "originalPrompt": "Original prompt 1",
                    "anonymizedPrompt": "Anonymized prompt 1",
                    "response": "Anonymized response 1"
                }
            ]
        }
    }

3. Upload a File

• Method: POST
• URL: /upload
• Header should contain the name of the file in the key File-Name
• Description: Endpoint to upload files to the LLM.
• Responses:
  • 200:

    • Description: File uploaded successfully.

    • Example:

      {
        "success": true,
        "message": "File uploaded successfully."
        "data" : "(Sample file id)"
      }
  400: Invalid file.

4. Anonymize Text

• Method: POST
• URL: /anonymize
• Description: Anonymizes sensitive entities in a given prompt.
• Parameters:

  • lang (body, string, required): Language for anonymization.

  • prompt (body, string, required): Input text.

  • choices (body, array of strings, required): Entities to anonymize.

  • model (body, string, required): Model to use.

  • anonymized (body, boolean, required): Flag to determine anonymization type.

    {
    	"lang" : "eng",
    	"prompt" : "(Name1 Sample prompt 1)",
    	"choices" : ["Names", "Emails"],
    	"model" : "llama3.1",
    	"anonymized" : false
    }
• Responses:

  • 200: Text anonymized successfully.

  • Data:

    {
    	"anonymized_prompt": "(Prompt after replacement of detected PII)",
    	"anonymized_values": {
    											"Emails": ["anon_email1", "anon_email2"],
    											"Names": ["Anonymized Name1"],
    											"Phone": ["Anonymized phone number1"]
    										}
    	"detected_values": {
    											"Emails": ["anon_email1", "anon_email2"],
    											"Names": ["Detected Name1"],
    											"Phone": ["Detected phone number1"]
    										}
    }

5. Send Chat

• Method: POST
• URL: /chat
• Description: Sends the anonymized prompt to the LLM and returns the response after replacement from the LLM.
• Parameters:

  • lang (body, string, required): Language for Name detection (eng, ger).

  • prompt (body, string, required): Prompt that is meant to be sent.

  • choices (body, array of strings, required): Entities to anonymize (Name, Email, Phone).

  • model (body, string, required): Model to use.

  • chat_id(body, string, optional): Chat_id for the prompt, assumed a new chat if empty.

  • doc (body, string, optional): Document id of the uploaded file.

  • anonymized (body, boolean, required): Flag to determine if anonymization is required.

    {
    	"lang" : "eng",
    	"prompt" : "(Sample prompt 1)",
    	"choices" : ["Names", "Emails"],
    	"model" : "llama3.1",
    	"chat_id" : "", // Considered a new chat
    	"doc" : "(Sample doc id)", //empty if document is uploaded
    	"anonymize" : false //boolean
    }
• Responses:

  • 200: Text anonymized successfully.

  • Data:

    {
    	"session_id": "(session id or chat id)",
    	"anonymized_prompt": "(Prompt by the user after anonymization)",
    	"replaced_response": "(Response from LLM after replacement)",
    	"anonymized_values": {
    											"Emails": ["anon_email1", "anon_email2"],
    											"Names": ["Anonymized Name1"],
    											"Phone": ["Anonymized phone number1"]
    									},
    	"total_tokens": "34" //tokens consumed for the prompt
    }

6. Delete a Session

• Method: DELETE
• URL: /delete-session/{session_id}
• Description: Delete a session/chat by its ID.
• Parameters:
  • session_id (path, string, required): ID of the session/chat to be deleted.
• Responses:
  • 200: Session deleted successfully.
  • 400: Invalid session ID.

7. Rename a Session

• Method: POST
• URL: /rename-session
• Description: Rename an existing session/chat.
• Parameters:

  • session_id (body, string, required): ID of the session to rename.

  • new_name (body, string, required): New name for the session.

    {
    	"session_id" : "(session id to be renamed)",
    	"new_name" : "(new name for this session)"
    }
• Responses:
  • 200: Session renamed successfully.
  • 400: Invalid session ID or name.

8. Get Token Usage

• Method: POST
• URL: /get-token-usage
• Description: Retrieve token usage data based on a date range for an organization.
• Parameters:

  • start_date (body, string, required): Start date for the usage data (format: Unix timestamp in milliseconds).

  • end_date (body, string, required): End date for the usage data (format: Unix timestamp in milliseconds).

    {
        "start_date" : "2024-12-18T12:52:00Z",
        "end_date" : "2024-12-18T12:52:00Z"
    }
• Responses:
  • 200: Token usage count retrieved successfully.
  • 400: Invalid date range.

9. Generate Image

• Method: POST
• URL: /generate-image
• Description: Generate image based on prompt provided by user.
• Parameters:

  • chat_id (body, string, optional): Chat id for image generation.

  • prompt (body, string, required): Prompt for image generation.

    {
        "chat_id" : "(Sample Chat id)",
        "prompt" : "Sample prompt for img gen"
    }
• Responses:

  • 200: Token usage count retrieved successfully.

  • 400: Invalid date range.

    {     "chat_id" : "(Chat id provided)",
          "image_url" : "(Link to generated image)"
    }

Additional Notes:

• session id and chat id are referring to the same entity and can be used interchangeably.
• User specific data like email id and organization id are extracted from the authentication token itself.