📁 Ingest Repository
🔒 100% In-Memory HBM
Drop any local codebase, folder, or .zip archive (up to 3M tokens). Code is processed in volatile GPU memory and evaporates upon disconnect.
⚡
No files handy? Try Benchmark Repo
SQLite3 (2.67M Tokens · 241K Lines)
Scanning repository files...
Reading contents into memory
Repository Active
0
Files
0
Lines
0
Est. Tokens
0 KB
Size
Inference Cost:
🪙
1 Credit
(1M tok rule)
Memory Pipeline:
📦
1 Macro-Shard
⚡ Suggested Whole-Repo Inquiries:
Server Configuration
Endpoint Live
In-Memory Attention Field
Tier: A10G (FP8)
Latency: --
Bypass: --
Tool Calls: 0
🚀 Whole-Repository Zero-Tool Inference
Drop or select your project folder on the left. Unlike traditional vector RAG or tool-calling agents that search 5 chunks at a time, WholeRepo holds the entire repository in memory via Riemannian Frustum attention.
✓ 2M+ Token Context
✓ Zero Data Persistence
✓ 1.4s Single-Pass
Executing single forward pass across in-memory repository...
0.0s