45 lines
2.3 KiB
Markdown
45 lines
2.3 KiB
Markdown
**Vaessl: Project Goals**
|
|
|
|
# Primary Objective
|
|
|
|
The goal of Vaessl is to bridge the gap between unstructured physical data (images) and structured digital systems (Inventory APIs). This project serves as a technical deep-dive into the integration of Generative AI within a traditional Full-Stack Java/Spring Boot environment.
|
|
|
|
# Learning & Development Pillars
|
|
|
|
While the end product is a functional utility, the development process is specifically designed to master the following domains:
|
|
|
|
- Modern Spring Boot & Spring AI: Moving beyond basic CRUD to orchestrate complex AI workflows, utilizing the Spring AI ecosystem to handle prompts and structured outputs.
|
|
|
|
- Vector Databases & Retrieval: Gaining hands-on experience in using vector databases in combination with classical development stacks.
|
|
|
|
- System Architecture: Designing a "Bridge" architecture that is decoupled from the target application, allowing for a flexible, provider-agnostic middleware service.
|
|
|
|
- AI Gateway Implementation: Mastering LiteLLM to manage multiple LLM providers (Gemini, OpenAI, Local LLMs).
|
|
|
|
- Containerized Orchestration: Delivering a production-ready environment via Docker Compose that manages networked services (App, DB, Proxy) with a single command.
|
|
|
|
# Product Goals (Public Value)
|
|
|
|
Vaessl is intended to be a useful tool for the public, specifically those utilizing self-hosted inventory systems like Homebox.
|
|
|
|
- Intuitive Discovery: Replace rigid keyword searches with intent-based search (e.g., finding a "Soldering Iron" when the user searches for "fix electronics").
|
|
|
|
- Improve productivity: Reduce the time it takes to catalog physical items through AI-assisted batch processing.
|
|
|
|
# Short-Term Roadmap & Milestones
|
|
|
|
- Phase 1: Foundation (Current)
|
|
* Deploy the core infrastructure (PostgreSQL + pgvector, LiteLLM, Spring Boot Skeleton, Vite frontend).
|
|
* Establish the bridge between the Vaessl backend and a demo Homebox instance.
|
|
|
|
- Phase 2: The Processing Pipeline
|
|
* Implement the image processing workflow: Upload $\rightarrow$ AI analysis $\rightarrow$ staging Table.
|
|
* Develop the Vite interface for data refinement and verification.
|
|
|
|
- Phase 3: Semantic Search & Demo
|
|
* Integrate Spring AI for embedding generation.
|
|
* Launch a public-facing demo to showcase the full-stack solution.
|
|
|
|
# Long-Term Roadmap & Milestones
|
|
|
|
TBD |