AI Assistant for Frappe Framework

Jana is an AI-powered assistant built on the Frappe Framework. Designed for teams who want context-aware AI without leaving their Frappe ecosystem, Jana provides chat, document understanding, and natural language interaction across any Frappe-based application — with support for multiple LLM providers and GDPR-compliant privacy.

A floating assistant available on every Frappe Desk page — ask questions, get explanations, and interact with your data in natural language.

• Floating chat widget accessible from any page
• Real-time streaming responses (token-by-token output)
• Persistent chat sessions across page navigation
• Full conversation history

Jana automatically detects the document you're viewing and uses its data in conversation — no copy-pasting required.

• Automatic DocType and document detection
• Field values injected into conversation context
• Grounded responses based on real document data
• LLM guardrails to prevent hallucination

Bring Your Own Key — use your preferred LLM provider. Each user can connect their own API key, or the system admin can set a shared key.

• OpenAI (GPT-4, GPT-3.5)
• Anthropic (Claude)
• Google (Gemini)
• OpenRouter
• Ollama (local models)
• vLLM (self-hosted)
• Any OpenAI-compatible endpoint

Create AI agents tailored to specific roles with custom system prompts, model settings, and tool configurations.

• Define agents with specific personas and instructions
• Assign tools and knowledge articles per agent
• Configure model and temperature per agent
• Switch between agents in chat

Inject business-specific knowledge into every conversation so Jana understands your company, processes, and terminology.

• Business description injected into all conversations
• Knowledge articles for domain-specific context
• Per-agent knowledge assignments
• Company-specific AI responses

Generate structured content — emails, summaries, reports — using AI with your business context.

• Template-driven content generation
• Business-aware output
• Reusable prompt templates

GDPR-compliant privacy layer that strips personal data before sending to cloud LLM providers and unmasks it in the response.

• Automatic detection and masking of personal data
• Data never leaves your server in identifiable form
• Unmasking on response for seamless user experience
• Configurable masking rules

Conversations are saved and accessible across page navigation — pick up where you left off.

• Session history per user and document
• Resume conversations after navigating away
• Search and browse past sessions

All user-facing strings are translatable — use Jana in your language.

• Translatable UI strings
• Multilingual AI response support depending on provider

When Dock is installed, Jana integrates into the ecosystem navigation and panel system.

• App registered in the Dock app switcher
• Jana panel accessible from the Dock navbar
• Settings accessible from Dock settings panel
• Notification types registered in Dock

Other ecosystem apps can expose tools to Jana for AI-powered insights — property data from Home, project summaries from Orga, and more.

• Tool-based integration system
• Apps register Jana tools via hooks
• No hard dependency — tools available only when apps are installed

• Frappe Framework v16 or higher
• Python 3.14+
• Node.js 24+
• MariaDB 10.6+
• At least one LLM provider account (OpenAI, Anthropic, etc.) or a local Ollama installation

# Get the app
bench get-app jana https://github.com/tonic-6101/jana.git

# Install on your site
bench --site your-site.localhost install-app jana

# Run migrations
bench --site your-site.localhost migrate

# Build assets
bench build --app jana

After installation, access Jana at: https://your-site.localhost/jana

1. Configure a Provider: Go to Jana Settings and create a provider with your LLM credentials
2. Set as Default: Mark the provider as default
3. Open Any Document: Navigate to any document in Frappe Desk
4. Start Chatting: Click the Jana widget and ask a question about the document
5. Explore Agents: Create custom agents for specific workflows

• Backend: Frappe Framework, Python 3.14+
• Frontend: Vue 3, TypeScript, Tailwind CSS
• UI Components: FrappeUI
• Database: MariaDB
• AI Providers: OpenAI, Anthropic, Google Gemini, OpenRouter, Ollama, vLLM
• Build: Vite

Contributions are welcome! This project uses pre-commit for code formatting and linting:

cd apps/jana
pre-commit install

Pre-commit runs the following tools automatically:

• ruff — Python linting and formatting
• eslint — TypeScript/JavaScript linting
• prettier — Code formatting
• pyupgrade — Python syntax modernization

• Issues: GitHub Issues
• Discussions: GitHub Discussions

GNU Affero General Public License v3.0 (AGPL-3.0)

See LICENSE for details.

SPDX-License-Identifier: AGPL-3.0-or-later
Copyright (C) 2024-2026 Tonic

Built with Frappe Framework and FrappeUI.