openproxy: Skill for Claude Code

openproxy-tts — text to speech

OpenProxy exposes an OpenAI-compatible TTS surface on 127.0.0.1:4623 and routes each request to whichever TTS provider you name in the model field. This skill covers:

1. Registering a TTS provider
2. Synthesizing audio over HTTP, and reading the response
3. Addressing model + voice (they share one field)
4. Provider-specific parity notes (Gemini, Fish Audio, OpenAI, OpenRouter, ElevenLabs, MiniMax, AWS Polly, self-hosted)
5. Combo fallback and account rotation
6. The openproxy media tts CLI

Every step is non-interactive and safe to run unattended. If the server is not up yet, follow .agents/skills/openproxy/SKILL.md first.

0 · Preconditions

curl -sS http://127.0.0.1:4623/health

# A proxy API key. Any key works, including the admin key from `server init`.
export OPENPROXY_API_KEY="$(openproxy key add tts-agent --auto --robot | jq -r '.data.key')"

# Which TTS providers have credentials right now
openproxy media providers list --kind tts

/v1/* is gated by the require_api_key setting (falling back to require_login). When it is on, every call needs Authorization: Bearer $OPENPROXY_API_KEY.

The openproxy media … commands need that key exported too. They talk to the running server rather than reading db.json directly, so an unset OPENPROXY_API_KEY surfaces as a confusing server returned 401 Unauthorized: {"code":"invalid_api_key","message":"Invalid API key"} — the key is usually simply missing, not wrong. Export it in the same shell as the curl calls, and use --api-key (or --url) when talking to a non-default server.

1 · Register a TTS provider

TTS credentials live in the media-provider registry, keyed by a provider id. The id is what you put in front of the model field.

cat > /tmp/tts.json <<'JSON'
{ "apiKey": "sk-...", "isActive": true }
JSON

openproxy media providers add \
  --provider openai --kind tts --name "OpenAI TTS" \
  --from-file /tmp/tts.json

--from-file merges on top of {provider, kind, name}, so it only carries the extra fields. - reads stdin. Edit with openproxy media providers edit <id> --from-file -; remove with openproxy media providers delete <id> --kind tts.