Starting friction
Students receive a brief but struggle to turn vague interests into an actionable direction.
Back to selected projects
AI product practice · V1
DesignSpark
An AI creative-start agent and inspiration archive for design students
It helps design students move from an ambiguous brief to an actionable creative state, then preserves useful sparks as reusable personal inspiration assets.
DESIGNSPARK / V1- 01Ambiguous brief
- 02Guided dialogue
- 03Creative spark
- 04Inspiration map
01
Problem framing
The early-design challenge is not a lack of ideas
Tasks, interests, references, and visual preferences are fragmented. A generic chatbot may answer questions, but rarely preserves task state, progression, or long-term inspiration.
Lost process
Useful keywords and visual cues disappear across chats and assets, forcing every project to restart.
AI overreach
One-shot output can replace user judgment instead of supporting clarification and deliberate choices.
02
Core loop
Grow one conversation into a long-term creative asset
V1 focuses on two connected modules: helping the user start, then preserving the signals worth revisiting.
Creative Spark
Sparkie asks about the task, interests, and constraints to surface keywords, visual language, and next actions.
Output: structured sparks and session progressInspiration Map
Conversation sparks and captured text, images, or voice become a personal creative galaxy for review and further exploration.
Output: reusable personal inspiration nodesAI proposes directions while the user keeps control over selecting, saving, editing, and deleting.
03
Product decisions
Protect the core hypothesis through deliberate scope
Personal tool over teaching management
Validate the individual creative-start and inspiration loop first.
Teacher, class, course, and assignment management are excluded.Open dialogue before fixed forms
Early design input is ambiguous, so the agent decides when to clarify or advance.
Users do not need to complete a complex task form first.Preserve process, not final-output automation
The product records how users form decisions rather than delivering only a final answer.
It is not positioned as a final artwork generator.Structured PRD · V1
Product Requirements Document
This web edition is structured from the source PRD and technical notes. It preserves factual boundaries and does not invent research samples, production metrics, or unverified capabilities.Context and positioning
DesignSpark is a personalized AI creative-start agent and inspiration archive for design students. Sparkie is not a generic chatbot or one-shot generator; it advances early design thinking through task state, user input, historical sparks, and saveable inspiration.
Target users and needs
Design students
Understand assignments, competitions, or portfolio briefs and find an entry point.
Portfolio builders
Accumulate themes, keywords, visual language, and reviewable process evidence.
Independent learners
Turn a vague interest into a small executable project and next action.
V1 goals
- Complete the Creative Spark conversation loop
- Deliver the personal Inspiration Map
- Support text, image, and voice-related input
- Define AI boundaries and failure handling
Non-goals
- No teacher, course, or assignment management
- No multiplayer or team space
- No final artwork generation
- No claim of a complete style profile or deployed long-term RAG
V1 scope and status
ModuleStatusScope
Account and authImplemented
Registration, login, logout, protected pages, and settings.
Creative SparkImplemented
Sparkie dialogue, persistence, image attachments, stage progress, and spark parsing.
History and spark cardsImplemented
Restore, rename, archive, delete, and continue from a saved spark.
Inspiration MapImplemented
Graph view of themes, sparks, keywords, and visual language.
Independent capturePartial
Text, image, and voice upload, analysis, and save depend on external AI configuration.
Style profile and RAGPlanned
Concepts and schemas exist but are not presented as a shipped core capability.
Core user flow
- 01Choose a module
After login, enter Creative Spark or the Inspiration Map.
- 02Describe the task
Share a brief, interest, constraint, or reference; Sparkie starts clarifying.
- 03Form a spark
The response is parsed into a structured spark and saved with session progress.
- 04Preserve and revisit
Sparks enter the map and can seed a new exploration later.
AI and data flow
The main path turns open-ended agent dialogue into user-controlled, persistent product data.
- Next.js UI
- Route Handler
- Coze Bot SSE
- Hidden-marker parsing
- Supabase persistence
- Inspiration Map
Planned metrics and factual boundary
No public real-world user count, retention, conversion, agent accuracy, or production performance data is available. These are validation metrics only.
- Activation
- Registration-to-first-message conversion
- Start
- First sessions completing at least one spark capture
- Archive
- Saved conversation sparks and independent inspirations
- Return
- Map visits and new explorations started from a spark
- Reliability
- /api/chat success, Coze timeout, and upload failure rates
Current boundaries
- Coze workflow topology, knowledge-base bindings, and published versions require separate confirmation.
- Three-direction and starter-card concepts are historical or partial and are not presented as the V1 main path.
- Image and voice behavior depends on external configuration and browser support and still requires real testing.
Live product
Experience the DesignSpark loop
The live product opens in a new tab. This portfolio does not embed the external site, avoiding cross-origin restrictions and preserving a reliable reading experience.Open live demoExternal site · yiyi01011.onlineSuggested path
- 01Enter Creative Spark
Start with a real design task or a vague interest.
- 02Talk with Sparkie
Observe how the agent clarifies context, advances stages, and forms sparks.
- 03Open the Inspiration Map
Review how conversation sparks and personal inspiration nodes are organized.