idn-analyzer
Offline prompt profiler for LLM routing decisions
know what a prompt needs before choosing where to serve it
What it is
idn-analyzer is a lightweight prompt profiler for LLM applications. It analyzes natural-language prompts and returns structured metadata about complexity, domain, PII risk, recommended routing tier, and execution requirements before any model call is made.
It is the analytical core of the Intelligence Delivery Network, but it is designed to stand alone inside any LLM app, gateway, or router.
What it returns
{
"token_estimate": 312,
"complexity_score": 0.74,
"reasoning_hops": 2,
"domain_tags": ["coding"],
"pii_risk": "none",
"data_egress_permitted": true,
"recommended_tier": "L2",
"confidence": 0.91
}
Design goals
Offline-capable
Layer 1 analysis is rules-based and does not require a network call or model invocation.
PII-aware
Prompts can be flagged for sensitive data and compliance constraints before cloud egress.
Gateway-agnostic
The output schema is meant to feed any custom router, LiteLLM-style proxy, or app-level policy engine.
Cross-platform
The roadmap includes Python, JavaScript, C/C++, WebAssembly, mobile, and embedded targets.
Status
Development is currently on hold. The repository captures the pre-alpha design, schema, CLI shape, and platform roadmap.
See the GitHub repo for the current source.