Now available
A chess engine for context.
Vaner predicts your next move and prepares useful context before you ask. When the real prompt arrives, it serves the best package fast.
Ponder loop
Prepare likely context before the prompt lands.
Vaner spends idle compute predicting likely next requests and prebuilding artifacts so the hot path starts ahead.
Answer loop
Route the best context package fast.
At prompt time, Vaner ranks candidate artifacts under privacy and token constraints, then injects the highest-value package.
Model-agnostic
Use local and cloud backends without lock-in.
Run with Ollama, vLLM, OpenAI-compatible providers, or mixed backend profiles per client and workflow.
Local-first
Keep control of scope and data boundaries.
Vaner is designed for explicit repository scope, redaction-aware packaging, and inspectable context decisions.
Security
Local-first defaults, explicit path scope, no content logging, and inspectable context decisions. How Vaner handles security →