AetherWiki
by Team Pijja
Infinite Ingestion. Zero Noise.
About This Project
AetherWiki is a self-healing knowledge engine that turns unstructured web data into a verified, hallucination-free knowledge graph.
The Core Problems It Solves
Semantic Data Rot: Prevents information bloating by actively distinguishing historical context from current reality, pruning out-of-date or contradictory nodes. Invisible Graph Corruption: Blocks broken or duplicate entity generation by forcing strict JSON structural schemas before data touches the disk. Asynchronous Blindness: Eliminates race conditions where querying a background graph-builder too early pulls half-baked, incomplete answers.
The Technical Stack
Firecrawl: Strips website layouts and ads into clean, pure Markdown text. Groq: Streamlines the pipeline by passing data at ultra-fast speeds to minimize execution lag. NVIDIA Cloud (Kimi-K2.6): Acts as the preprocessor, forcing rigid schema outputs to map human intent, merge entity duplicates, and resolve cross-article conflicts. HydraDB MCP: Natively hosts and maintains the physical long-term graph matrix. Claude Code: Functions as the terminal orchestrator, managing data-flow locks during ingestion, debugging scripts, and deployment.
Built With
Repository
Submitted May 18, 2026 at 9:41 AM