YanCheng-go/my-focal-ai
Personal news intelligence — aggregate AI content from curated sources, score with LLM, serve your own dashboard
What's novel
Personal news intelligence — aggregate AI content from curated sources, score with LLM, serve your own dashboard
Code Analysis
14 files read · 4 roundsA personal AI news aggregation system that fetches content from RSS, Twitter, GitHub trending, events, and AI tool sites, scores relevance using LLM-based principled evaluation (signal/noise, mechanism/opinion, builders/commentators), and serves a web dashboard with multiple deployment modes (local
Strengths
The scoring framework is genuinely thoughtful — a principled three-axis evaluation with an information flow model rather than generic 'rate this 1-10', and the multi-pipeline export system that merges local and cloud outputs while preventing cross-contamination shows real engineering depth. The codebase is well-organized with clean separation between ingestion, storage, scoring, and presentation layers, and the Twitter/GitHub scrapers contain real non-trivial parsing logic.
Weaknesses
The Twitter scraper relies on hardcoded GraphQL endpoint IDs and a public bearer token that will break when Twitter rotates them, and the Supabase backend's per-item RPC calls in ingest_items will be slow at scale. Some source fetchers (events, aitmpl) are tightly coupled to specific website structures with no abstraction layer for resilience.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
262
Contributors
5
Files
158
Active weeks
2
Repository
Language
Python
Stars
2
Forks
0
License
NOASSERTION