Tencent/WeSmartFlow

Every question can open a new path. WeSmartFlow turns learning into conversation, exploration, stories, and hands-on discovery.

What it solves

WeSmartFlow is an agent-native adaptive learning framework designed to move beyond simple Q&A interactions. It addresses the fragmented nature of learning by providing a system that understands a user's learning goals, organizes personalized paths, provides interactive practice, and maintains a long-term record of progress through a knowledge graph.

How it works

The system employs a multi-layered architecture combining a ReAct-based tutoring agent with a specialized educational toolset. It uses a "Graph Memory" to track a user's mastery level of concepts and their relationships (prerequisites, extensions, etc.), utilizing SM-2 parameters for spaced repetition. For comprehensive topics, a multi-agent workflow (Planning, Research, Writing, Illustration, Voice, and Quiz agents) collaborates to generate full courses including PDFs and audio. Additionally, the EduViz SDK allows agents to create interactive visualizations in a sandbox for exploring abstract concepts.

Who it’s for

  • Learners seeking a personalized, interactive AI tutor that remembers their progress across different sessions.
  • Educators and Content Creators who want to build and publish independent, interactive "learning worlds" (e.g., simulations, story-based lessons) using their own frontend technologies while leveraging the framework's agent and backend capabilities.
  • Developers looking for a reusable educational agent engineering framework.

Highlights

  • Personalized Knowledge Graph: Tracks mastery levels and concept relationships to prevent starting from zero every time.
  • Explore Mode: A curated space for immersive, scenario-based learning (e.g., "Magic English Town" or "Chemistry Lab").
  • EduViz SDK: Enables the generation of interactive, parameter-driven visualizations for complex concepts.
  • Multi-Agent Course Generation: Automates the creation of structured courses with research, content, and assessment.
  • WeClaw Integration: Allows the learning assistant to be accessed via WeChat, rendering interactive cards as images for seamless mobile learning.

Related

  • Project
    SterlingChin/marvin-template

    An AI Chief of Staff that extends Claude Code to provide session continuity, goal tracking, and integrations with productivity tools like Google Workspace and Slack.

  • Project
    HanyuanWang/LiveStream-Agent-Studio

    A local-first AI workspace for livestream commerce teams that automates creator discovery, livestream transcription, performance review, and video scripting.

  • Project
    Tiger3807861189/GLM-5.3-Flash-J-Space-Capability-Realization-Report

    A model-agnostic inference-time control system that enhances AI agent reasoning and long-horizon task performance through selective workspace management and state externalization.

  • Project
    agiwhitelist/auteur

    An AI agent skill that directs the creation of cinematic, high-performance websites using a disciplined pipeline of art direction, asset generation, and strict anti-slop linting.