A self-hosted Flask AI assistant with RAG, vision, multi-tool execution, and canvas document editing. Full workflow automation in one open-source platform.

Flask ChatBot: Multi-Provider + Tools + RAG + OCR + Multimodal + Canvas + Memory + Workspace

This is a single-page Flask chat application built around DeepSeek plus optional OpenRouter models, multi-step tool use, registry-driven composer slash commands, local RAG, dedicated local OCR, configurable helper/direct image analysis, conversation summarization, pruning, user-configurable entropy-aware context selection, persistent conversation memory, editable canvas documents, page-aware canvas navigation, and a per-conversation workspace sandbox.

It is not a minimal prompt/response demo. The app keeps conversation history in SQLite, restores assistant metadata when a conversation is reopened, supports editing earlier user messages, streams tool progress and reasoning, can enrich a user turn with local OCR or extracted document text before the model sees it, and can compact older content with summaries and pruning.

Screenshots

Tool execution

Long-term memory (RAG)

Canvas view

Settings page

Contents

• What the app does
• Screenshots
• Architecture overview
• Project structure
• Installation
• Configuration
• Using the app
• Available tools
• HTTP endpoints
• Data storage
• Development
• Security and operational notes
• Troubleshooting
• FAQ
• License

What the app does

Chat and conversation workflow

• Create, open, rename, and delete conversations
• Stream assistant output to the bro