llm-externalizer-or-model-info

LLM Externalizer — OpenRouter Model Info

Overview

Query OpenRouter's /v1/models/{exact_id}/endpoints for a specific model and display context length, pricing, supported request-body parameters, quantization, uptime, latency, and throughput. Uses the LLM Externalizer CLI (not the MCP tool) so it works from subagents — MCP tools from plugins are not available in subagent contexts.

Prerequisites

• llm-externalizer CLI on PATH (bundled with the plugin)
• $OPENROUTER_API_KEY set, OR active profile is OpenRouter-backed

Instructions

Copy this checklist and track your progress:

1. Parse the user's prompt for the exact OpenRouter model id (first arg) — case-sensitive, vendor-prefixed, with any :free / :thinking / :beta suffix. Also scan for optional format flags: - --no-color / --nocolor / --bw / --mono → forward --no-color - --markdown / --plain → forward --markdown - --json / --raw → forward --json
2. If the user gave only a partial name, ask for the full id.
3. Run: npx llm-externalizer model-info "<exact-id>" [flags]
4. Stop. Do not reprint, paraphrase, summarize, or add commentary. The Bash tool output pane renders the ANSI-colored table natively. If Claude Code collapses it behind a "+N lines" fold the user will expand it with ctrl+o themselves. Your response should contain nothing beyond what's necessary to confirm the command ran — ideally just acknowledge completion in a single short sentence.
5. Only add a textual summary if the user explicitly asks a follow-up question beyond "show me the info" (e.g. "which provider is cheapest?", "does it support reasoning?"). Otherwise the table speaks for itself.
6. On error, see references/errors.md.

Output

Per endpoint: context, max_completion, quantization, capability flags (reasoning, tools, structured output, caching), pricing ($/M tokens), uptime (5m/30m/1d), latency + throughput percentiles, supported_parameters. Live data, no cache.

Examples

# Default colored table
npx llm-externalizer model-info "nvidia/nemotron-3-super-120b-a12b:free"

# Compare providers (Llama 3.3 has 17 endpoints)
npx llm-externalizer model-info "meta-llama/llama-3.3-70b-instruct"

# Markdown table — renders in any markdown viewer
npx llm-externalizer model-info "google/gemini-2.5-flash" --markdown

# Raw JSON to stdout (for jq / scripts)
npx llm-externalizer model-info "anthropic/claude-sonnet-4.5" --json

# Raw JSON written to a file
npx llm-externalizer model-info "x-ai/grok-4.1-fast" --json grok-info.json

See references/example-output.md for a full sample:

• Sample response
• Reading the output
• Percentiles explained
• Comparing multiple endpoints

And references/use-cases.md for more scenarios:

• Verify supported parameters before integrating a model
• Compare pricing across providers hosting the same model
• Debug slow or failing calls
• Check quantization for quality trade-offs
• Confirm context length and max tokens
• Check reasoning support

Error Handling

Error	Resolution
OpenRouter returned 404	Wrong model id — check case, vendor prefix, :free / :thinking suffix
No OpenRouter auth token available	Set $OPENROUTER_API_KEY or switch to an openrouter-remote profile
Network error	Retry once; check /llm-externalizer:llm-externalizer-discover for service status
OpenRouter returned no endpoints	Model deprecated — suggest alternative

Full table in references/errors.md:

• Error table
• Debugging tips

Resources

• references/errors.md — all error cases and resolutions
  • Error table
  • Debugging tips
• references/example-output.md — full sample output
  • Sample response
  • Reading the output
  • Percentiles explained
  • Comparing multiple endpoints
• references/use-cases.md — six common scenarios
  • Verify supported parameters before integrating a model
  • Compare pricing across providers hosting the same model
  • Debug slow or failing calls
  • Check quantization for quality trade-offs
  • Confirm context length and max tokens
  • Check reasoning support