Date: January 1, 2026 Status: Active / Stable (Hybrid AI Architecture)
The Poshmark Assistant application has undergone a major architectural overhaul and migration. The codebase has been consolidated, optimized for local resource usage, and moved to a Linux-native environment for improved Docker performance.
- Hybrid AI Engine:
- Vision: Migrated to Google Gemini 1.5 Pro for superior OCR and tag recognition.
- Logic & Drafting: Migrated to Ollama (Llama 3.2) running locally on the host machine to eliminate API costs for text generation.
- Research Integration:
- Implemented Google Custom Search API to replace manual link generation. The app now performs real-time market pricing research.
- Automation:
- Deployed a dedicated Obsidian Agent (
obsidian-watcher) to synchronize documentation and code blueprints to an external vault (E:\WSL\Poshmark-Final-Vault).
- Deployed a dedicated Obsidian Agent (
To ensure a clean development environment, the following actions were taken:
- Filesystem Migration: The active project was moved from the Windows mount (
/mnt/c/...) to the WSL Linux filesystem (/home/ubuntuog/Poshmark-Assistant-LATEST) to resolve permission issues and improve IO performance. - Legacy Cleanup: Old project iterations (
poshmark,poshmark-ollama,Poshmark-Assistant-GitHub) were archived or removed. - Docker Optimization: Unused Docker images and build caches were pruned, reclaiming ~26GB of disk space.
- Active Directory:
~/Poshmark-Assistant-LATEST - App URL:
http://127.0.0.1:3000 - Documentation Vault:
E:\WSL\Poshmark-Final-Vault
- User Acceptance Testing (UAT): Verify the full listing flow (Upload -> Details -> Pricing -> Draft) in the browser.
- Env Management: Ensure
.envkeys are rotated if necessary.