@reaatech/llm-router-cli

Status: Pre-1.0 — APIs may change in minor versions. Pin to a specific version in production.

Command-line interface for llm-router. Provides four commands — route, benchmark, cost-report, and validate-config — for routing prompts, benchmarking model performance, generating cost reports, and validating configuration files from the terminal.

Installation

npm install -g @reaatech/llm-router-cli
# or
pnpm add -g @reaatech/llm-router-cli

Feature Overview

• Route — send a prompt through the router with full strategy selection, cost estimation, and execution
• Benchmark — run a prompt against one or more models with multiple runs, reporting latency and cost stats
• Cost report — generate daily, weekly, or monthly cost summaries per budget
• Config validation — validate a YAML/JSON router configuration file against the RouterConfigSchema

Quick Start

# Route a single request
llm-router route \
  --config llm-router.config.yaml \
  --strategy cost-optimized \
  --prompt "Review this TypeScript module for bugs."
 
# Benchmark two models with 3 runs each
llm-router benchmark \
  --config llm-router.config.yaml \
  --models "glm-edge,kat-coder-pro" \
  --prompt "Write a function to reverse a linked list." \
  --runs 3
 
# Generate a cost report
llm-router cost-report \
  --config llm-router.config.yaml \
  --period today \
  --budgetId default
 
# Validate a config file
llm-router validate-config --config llm-router.config.yaml

Commands

route

Route a prompt through the router and display the decision, cost, and result.

llm-router route [options]

Sample output:

Routing Decision
  Model: glm-edge (zhipu)
  Strategy: cost-optimized
  Confidence: 0.95
  Estimated Cost: $0.0005

Execution Result
  Actual Cost: $0.0004
  Latency: 42ms
  Input Tokens: 120
  Output Tokens: 85
  Content: ...

benchmark

Benchmark one or more models against a prompt with multiple runs.

llm-router benchmark [options]

Sample output:

Benchmarking 2 models with 3 runs each

kat-coder-pro (3 runs)
  Avg Latency: 234ms  |  P50: 220ms  |  P95: 280ms
  Avg Cost: $0.0012

glm-edge (3 runs)
  Avg Latency: 45ms  |  P50: 42ms  |  P95: 52ms
  Avg Cost: $0.0004

cost-report

Generate a cost report for a budget within a time period.

llm-router cost-report [options]

Sample output:

Cost Report
  Budget: default
  Period: today
  Total Cost: $0.0123
  Total Requests: 8
  Remaining Budget: $4.9877

By Model
  glm-edge: $0.0032 (3 requests)
  kat-coder-pro: $0.0091 (5 requests)

validate-config

Validate a YAML/JSON router configuration file.

llm-router validate-config [options]

Sample output:

Configuration is valid

On invalid configs, validation errors are printed with the specific field and reason.

Related Packages

• @reaatech/llm-router-engine — Main routing engine
• @reaatech/llm-router-core — Shared types and schemas

License

MIT