AI Chatbot Implementation in Frappe (Floating Widget + API/Local Model Support)

,
1

Hi everyone,

I would like to check the feasibility and scope of implementing an AI chatbot within Frappe/ERPNext.

Requirement

I am looking for a chatbot that:

  • Appears as a floating chat widget/window across the application.

  • Can interact with users in real time.

  • Supports integration with external AI providers through APIs (OpenAI, Gemini, Claude, Azure OpenAI, etc.).

  • Also supports locally hosted/self-hosted LLMs (Ollama, Llama, Mistral, etc.) for organizations with data privacy requirements.

  • Can be trained or configured using internal documents, knowledge bases, or ERP/Frappe data (RAG/knowledge retrieval approach).

  • Exposes APIs so that other applications can send prompts and receive responses programmatically.

  • Supports role-based access and conversation history if possible.

Queries

  1. Has anyone implemented something similar in Frappe/ERPNext?

  2. Are there existing Frappe apps/plugins that provide this functionality?

  3. What would be the recommended architecture for supporting both cloud AI APIs and local models?

  4. Are there any limitations or performance considerations when implementing a floating chatbot across the framework?

  5. Would you recommend building this as a standalone Frappe app or integrating with an existing chatbot solution?

I would appreciate any guidance, existing implementations, repositories, or best practices from the community.

Thanks in advance!

1 Like
2

You can try the below repos:

  1. GitHub - vyogotech/frappe_ai: Frappe Conversational AI interfaces that leverages ERPNext MCP server · GitHub
  2. GitHub - vyogotech/frappe-mcp-server: AI power for Frappe Framework · GitHub
  3. GitHub - vyogotech/frappe-ai-agent: Frappe AI Agent · GitHub

All these together form a scalable AI Stack for Frappe and totally customizable. DM me if you need a production level support.
Stack is totally model agonistic, respects the RBAC of Frappe.

3

You can check below:

2 Likes