Couchdb Notes

Ensure you have set the following configs in the configurations database in couchdb:

  "keys": {
    "openai": "sk-mm",
    "perplexity": "pplx-21",
    "gemini": "AIza"
  },
  "models": {
    "openai": "gpt-3.5-turbo",
    "perplexity": "llama-3-sonar-small-32k-online",
    "gemini": "gemini-pro"
  },
  "assistant": {
    "name": "Planet Context",
    "instructions": "You are a brainstorming manager for Open Learning Exchange (OLE) - https://ole.org/, you have specialised knowledge in Planet(web app) and myPlanet(mobile app) applications developed by OLE. You are designed to generate innovative ideas and provide suggestions and help the community members so as to ensure OLE's mission of empowering communities. Emphasize on terms like 'learning,' 'learner,' 'coach,' 'leader,' 'community,' 'power,' 'team,' and 'enterprises,' and avoids overly technical jargon. You are to embody OLE's ethos of self-reliance, mentoring, and community leadership, steering clear of concepts that contradict these values. Communicates in a formal tone, treating users with respect and professionalism, and maintaining a supportive, solution-oriented approach. Ask for clarifications when necessary to ensure contributions are accurate and relevant, and always encourages community-focused, empowering brainstorming."
  }

Note: This applies for both production and development environments.

Development Notes

For development environment add a .env file in the chatapi directory

Add the following configs in the .env file:

  SERVE_PORT=5000
  COUCHDB_HOST=http://localhost:2200
  COUCHDB_USER=username
  COUCHDB_PASS=password

In the production environment these configs are set in the planet.yml file.

API Overview

HTTP requests

GET /

Description: Returns a welcome message from the API. Response:

• Status Code: 200 OK
• Body: json

{
  "status": "Success",
  "message": "OLE Chat API Service"
}

POST /

Description: Sends a chat message with an option to save. Input: JSON object in the request body with fields data (object) and save (boolean). Sample Input

{
  "data": {
    "user": "admin",
    "content": "Hello",
    "assistant": true,
    "aiProvider": {
      "name": "openai",
      "model"?: "gpt-3.5-turbo",
    },
    "_id"?: "0157cd109e2a6852a82e2b5da402e64b",
    "_rev"?: "2-ad3d40af8abdead9eef8a6aa6a8c6f87"
  },
  "save": true
}

Additional info on data:

• user: string(required) -> Provide the planet/myPlanet username
• content: string(required) -> The latest prompt for the AI to answer
• assistant: boolean(required) -> Set to true if you want to use the assistants endpoint
• aiProvider: Object(required)
  • name: string(required) -> Name of the API provider to choose from i.e openai, perplexity or gemini.
  • model: string(optional) -> Name of the specific provider model to use. Defaults to gpt-3.5-turbo for openai, llama-3-sonar-small-32k-online for peplexity and gemini-pro for google gemini
• _id: couchdb document id
• _rev: couchdb revision id The couchdb id and rev are optional fields(only optional when starting a new chat), however to update an existing chat the id and rev (required) be provided. Take note that the rev changes each time the db document is updated.

Response: Varies depending on the outcome. Success response example:

• Status Code: 201 Created
• Body: json

{
  "status": "Success",
  "chat": "How may I assist you today?",
  "couchDBResponse": {
      "ok": true,
      "id": "aee429dffc8e50ddeca9addf5a00beb8",
      "rev": "2-b97b95e0b32de4f0c622b0e5559c1671"
  }
}

GET /checkproviders

Description: Checks the availability of AI providers supported. Response:

• Status Code: 200 OK
• Body: json

{
  "openai": true,
  "perplexity": false,
  "gemini": true
}

---

Websocket requests

Description: Establishes a WebSocket connection for real-time chat. Usage: Connect via WebSocket and send JSON messages containing chat data. Responses will be provided through the WebSocket connection. aiProviders supported: openai, perplexity, and gemini. any provider model supported by the provider can be used.

• Request json sample

  {
      "user": "admin",
      "time": {},
      "content": "Hola",
      "aiProvider": {
          "name": "openai"
      }
  }
  

• Response json sample if response is not final

  {
      "type": "partial",
      "response": "En"
  }
  

if response is final

{
    "type": "final",
    "completionText": "¡Hola! ¿En qué puedo ayudarte hoy?",
    "couchDBResponse": {
        "ok": true,
        "id": "aee429dffc8e50ddeca9addf5a001402",
        "rev": "2-22a89c4b4e6a25c27cb41ac1ea042c3e"
    }
}