local-model-spend-router v1.0

Price: $4.99

Category: analytics

Status: APPROVED_FOR_RELEASE — Viper QA 6/6, Security Audit CLEARED by Warlock on 2026-04-28

What it does

local-model-spend-router helps OpenClaw users choose the cheapest sensible model route for each task without guessing. It weighs task complexity, privacy sensitivity, context size, latency tolerance, and remaining daily budget, then recommends whether to use local/free models, low-cost cloud models, premium models, or a hybrid fallback chain.

It is built for operators who want practical spend control: use local models when quality risk is low, preserve paid budget for high-impact work, and keep a clear fallback plan when a route fails or privacy requirements change.

Use cases

• Cost-conscious OpenClaw operators who want to keep daily model spend under control while still escalating important work to stronger models when needed.
• Local-first AI builders who run Ollama or Hermes-style free workers and need a simple routing policy for routine summaries, coding support, and internal analysis.
• Privacy-sensitive teams that need sensitive tasks routed locally first, with explicit fallback guardrails before anything reaches a paid cloud provider.

Requirements

• OpenClaw v2026.3.23 or later recommended.
• Python 3.10+.
• No third-party Python packages required; uses Python standard library only.
• No API keys required for routing recommendations.
• Optional: installed local models such as Ollama llama3.2:3b, qwen2.5-coder:7b, or a local Hermes worker if you want to execute the recommended local routes.

Example usage

From the skill directory:

python3 scripts/route_models.py \
  --task "summarize overnight logs" \
  --complexity low \
  --privacy internal \
  --context-tokens 8000 \
  --budget-remaining 0.35

Expected output:

Route: local-first
Fallback: ollama/llama3.2:3b -> ollama/qwen2.5-coder:7b -> hermes-local/free -> openrouter/haiku-or-flash -> gemini-flash -> claude-haiku
Budget: Tight budget: reserve premium models for failures or high-impact tasks.
Why: complexity=low, privacy=internal, context_tokens=8000, budget=$0.35

Machine-readable mode:

python3 scripts/route_models.py --task "review contract" --complexity high --privacy sensitive --budget-remaining 2.00 --json

Expected output shape:

{
  "task": "review contract",
  "primary_route": "local",
  "fallback_chain": [
    "ollama/llama3.2:3b",
    "ollama/qwen2.5-coder:7b",
    "hermes-local/free",
    "openrouter/haiku-or-flash"
  ],
  "budget_note": "Budget available: route by complexity with low-cost defaults.",
  "rationale": "complexity=high, privacy=sensitive, context_tokens=8000, budget=$2.00"
}

Marketing copy

local-model-spend-router gives OpenClaw users a simple, evidence-based way to stop overspending on routine AI work. Tell it the task, complexity, privacy level, context size, and remaining daily budget, and it returns a practical model route: local-first, balanced, premium, or local-only when budget is gone. It is perfect for teams mixing Ollama, free Hermes workers, and paid cloud models who want clear fallback chains and budget guardrails without building a full routing layer from scratch.