Context optimization
Removes repeated fragments, trims irrelevant retrieval results, and controls long-conversation token budgets.
ir.ngjoo.com · ONLINE
IntelligentRouting
Proxies LLM API requests, reduces wasted tokens through deduplication, RAG trimming, and history budgets, then makes usage, cost, and policy effects observable.
Core capabilities
CORECost observability
Tracks tokens and spend by model and request to reveal the call paths consuming the budget.
Safety policies
Supports sensitive-data masking, policy versioning, and quality checks before new rules go live.
Who it is for
Useful for teams using an OpenAI-style API that want to observe tokens, cost, and redundant context before enabling routing policies gradually.
Read the boundary first
Results depend on request shape, retrieved content, and quality thresholds. This page does not repeat a fixed saving percentage without a current benchmark.
Ways to use it
- Record tokens, models, and cost for current requests to establish a real unoptimized baseline.
- Enable one policy on duplicated context or irrelevant retrieval and compare output quality on a small sample.
- Expand to more call paths only after metrics and reviewed examples show no material regression.
How to review the result
- Token reduction is not the only goal; check omissions, citation shifts, and changed tool arguments.
- Request logs may contain sensitive data, so verify masking, access, and retention first.
- Policies need versions and rollback so one configuration change does not affect every production request at once.
Current status and evidence
2026-09-23The console redirected to unified sign-in on 2026-09-23. Deployment records keep the /v1/* machine API on API-key authentication.
The available evidence is first-party: deployment records, product documentation, and endpoint status at the review time. No independent evaluation was available. Reachability confirms access at that moment, not performance, accuracy, savings, or long-term availability.