{"hub":"BBE Sprachmodell-Hub","version":"0.3.0","updated":"2026-07-19","note":"Kanonische Single Source of Truth aller BBE-Modell-Schienen (code-verifiziert). Forward-compatible fuer Auto-Routing. Ehrliche Status inkl. 'gebaut aber kein Key' und 'Key aber kein Wrapper'.","status_legend":{"live":"laeuft, direkt nutzbar","key-ready":"Integration + gueltiger Key -> aufrufbar","module-ready":"Modul gebaut (Mock/Verdrahtung offen)","key-gated":"Key vorhanden, aber Tier/Guthaben blockiert echte Nutzung","key-missing":"API-Pfad/OpenAI-kompatibel, aber KEIN Key (Operator-Gate, Kosten)","key-orphan":"Key liegt im Vault, aber KEIN Code-Wrapper gebaut","stub":"Backend-Geruest im Code, aber nicht implementiert","eval":"Kandidat, noch kein Code/Installation","degraded":"aktuell defekt","session-gated":"Subscription-Bridge, aktiv erst wenn Operator Session/Cookies ablegt"},"capabilities":["chat","code","stt","tts","voice-agent"],"cost_policy":"eur:0-Quellen (self-hosted / Subscription / Bestands-Guthaben) beim Routing bevorzugt; kostenpflichtige API nur als Fallback.","integration_layers":{"bbe-ai-gateway":"/home/dev/projects/bbe-ai-gateway (178:5420) — Router claude/gpt/grok/together, Modi mock|cli|api","chatgpt-bridge-mcp":"/home/dev/chatgpt-bridge-mcp — Subscription-Bridges (chatgpt/codex/grok/gemini) via SSH 73er","bbe-voice":"/home/dev/projects/bbe-voice — Voice-Gateway, Adapter ElevenLabs/Deepgram/Voicebox","bbe-deepgram":"/home/dev/bbe-deepgram — umbrandbares STT/TTS/Voice-Agent-Modul","cloud-ai-agent":"178:5604 /agent — Rollen trainee/junior/senior/experte -> Gateway + Mistral-Fallback"},"models":[{"id":"claude-gateway","name":"Claude (Opus 4.8 / Sonnet 5 / Haiku 4.5 / Fable 5)","capability":"chat","provider":"Anthropic","hosting":"subscription","access":"bbe-ai-gateway","endpoint":"http://127.0.0.1:5420/v1/agent","health":"http://127.0.0.1:5420/v1/ready","cost":{"model":"subscription","unit":"flat","eur":0},"status":"live","priority":1,"quality_tier":"frontier","latency_class":"medium","params":{"streaming":true,"models":["claude-opus-4-8","claude-sonnet-5","claude-haiku-4-5","claude-fable-5"]},"test":"curl -H 'Authorization: Bearer $GATEWAY_STATIC_TOKEN' -d '{\"prompt\":\"PING\"}' http://127.0.0.1:5420/v1/agent","source":"bbe-ai-gateway cli-backend (Claude-Max OAuth); bbe-llm SDK-Pfad","notes":"eur:0 Subscription. Fable 5 = schreib/kreativ. api-backend callAnthropic existiert, aber kein raw ANTHROPIC_API_KEY (subscription-first)."},{"id":"grok-brydge","name":"Grok","capability":"chat","provider":"xAI / SpaceXAI","hosting":"subscription","access":"chatgpt-bridge-mcp","endpoint":"mcp grok_consult · ssh grok-ask (73er)","health":"grok-token-refresh.service","cost":{"model":"subscription","unit":"flat","eur":0},"status":"live","priority":2,"quality_tier":"high","latency_class":"slow","params":{"streaming":false},"test":"MCP grok_consult(prompt)","source":"chatgpt-bridge-mcp grok-ask.mjs","notes":"Subscription eur:0. Gateway grok-api-backend = STUB (kein XAI_API_KEY)."},{"id":"chatgpt-brydge","name":"ChatGPT (GPT) — chat/cowork/code/design","capability":"chat","provider":"OpenAI","hosting":"subscription","access":"chatgpt-bridge-mcp","endpoint":"mcp chatgpt_consult · ssh chatgpt-ask (73er) · Web-UI :8124 (178)","health":"chatgpt-cdp-bridge.service","cost":{"model":"subscription","unit":"flat","eur":0},"status":"degraded","priority":3,"quality_tier":"high","latency_class":"slow","params":{"streaming":false},"test":"MCP chatgpt_consult(prompt)","source":"chatgpt-bridge-mcp chatgpt-ask.mjs","notes":"2026-07-19 stream_incomplete (WS-Conduit 73er). Fix = ChatGPT-Browser/CDP auf 178 (:8124/:8125). Personas chat/cowork/code/design = Kanaele EINES Boards. Gateway gpt-api = STUB."},{"id":"gemini-brydge","name":"Google Gemini","capability":"chat","provider":"Google","hosting":"subscription","access":"chatgpt-bridge-mcp","endpoint":"mcp gemini_consult · ssh gemini-ask (73er)","health":null,"cost":{"model":"subscription","unit":"flat","eur":0},"status":"session-gated","priority":4,"quality_tier":"high","latency_class":"slow","params":{"streaming":false,"multimodal":true,"long_context":true},"test":"MCP gemini_consult(prompt) — nach Cookie-Ablage","source":"chatgpt-bridge-mcp gemini-ask.mjs (2026-07-19 ergaenzt)","notes":"NEU (4. Chat-Bridge). Aktiv erst wenn Operator gemini.google.com-Cookies auf 73er ablegt (sonst session_missing)."},{"id":"mistral-agent","name":"Mistral (large-latest)","capability":"chat","provider":"Mistral AI","hosting":"api","access":"cloud-ai-agent","endpoint":"cloud-ai-agent /agent ask=mistral (178:5604)","health":null,"cost":{"model":"per-token","unit":"call","eur":0.004},"status":"live","priority":6,"quality_tier":"mid","latency_class":"fast","params":{"streaming":true},"test":"cloud_ai_agent /agent model=mistral_fallback","source":"cloud-ai-agent; Key secrets/llm/Mistral","notes":"Key vorhanden, verifiziert real. Direkter API-Fallback (~0,004 EUR/Call, rate-limited)."},{"id":"together-chat","name":"Together AI (Aggregator-Roster)","capability":"chat","provider":"Together AI","hosting":"api","access":"bbe-ai-gateway","endpoint":"https://api.together.xyz/v1/chat/completions (OpenAI-kompatibel)","health":null,"cost":{"model":"per-token","unit":"1M tok","eur":0},"status":"key-ready","priority":5,"quality_tier":"high","latency_class":"fast","params":{"streaming":true,"roster":["gpt-oss-20b","Qwen3-235B","MiniMax-M3","Kimi-K2.7-Code","DeepSeek-V4-Pro"]},"test":"curl -H 'Authorization: Bearer $TOGETHER_API_KEY' https://api.together.xyz/v1/models","source":"bbe-ai-gateway callTogether + together-roster.ts; Key secrets/llm","notes":"EIN Key -> 272 Modelle, kommerziell erlaubt, Bestands-Guthaben. callTogether implementiert, ABER Gateway laeuft GATEWAY_MODE=cli -> api-Lane noch nicht scharf gewired. Enthaelt DeepSeek-V4-Pro (reasoning)."},{"id":"openrouter","name":"OpenRouter (Aggregator)","capability":"chat","provider":"OpenRouter","hosting":"api","access":"geplant","endpoint":"https://openrouter.ai/api/v1/chat/completions (OpenAI-kompatibel)","health":null,"cost":{"model":"per-token","unit":"1M tok","eur":null},"status":"key-missing","priority":7,"quality_tier":"high","latency_class":"fast","params":{"streaming":true,"audio":"12 TTS / 10 STT","models":"300+"},"test":"n/a — noch nicht integriert","source":"voice_provider_landscape_state","notes":"EHRLICH: KEINE Code-Integration, KEIN Key im System. Unser funktionierendes Aequivalent ist Together (Key da). OpenRouter nur wenn zusaetzlicher Multi-Provider-Zugang gewuenscht (Operator-Gate)."},{"id":"codex-brydge","name":"Codex (GPT-Codex)","capability":"code","provider":"OpenAI","hosting":"subscription","access":"chatgpt-bridge-mcp","endpoint":"mcp codex_consult · ssh codex-ask (73er) · gateway cli 'codex exec'","health":"teilt ChatGPT-Kontingent","cost":{"model":"subscription","unit":"flat","eur":0},"status":"live","priority":1,"quality_tier":"high","latency_class":"slow","params":{"streaming":false,"model":"gpt-5.1-codex"},"test":"MCP codex_consult(prompt)","source":"chatgpt-bridge-mcp + bbe-ai-gateway cli 'codex exec'","notes":"Code-Partner. usage_limit bei Kontingent-Erschoepfung (heute erschoepft)."},{"id":"parakeet-cpp","name":"Parakeet (parakeet.cpp)","capability":"stt","provider":"NVIDIA / ggml","hosting":"self-hosted","access":"geplant","endpoint":"self-hosted geplant · ODER via Together sofort","health":null,"cost":{"model":"self-hosted","unit":"compute","eur":0},"status":"eval","priority":1,"quality_tier":"high","latency_class":"fast","params":{"streaming":true,"languages":"25","gpu":true,"cpu":true},"test":"self-hosted noch nicht installiert","source":"mudler/parakeet.cpp v0.4.0 (MIT)","notes":"NEU, eur:0-Primaerkandidat STT. Kein Code/Installation bisher — bis dahin via Together-Key nutzbar."},{"id":"whisper-cpp","name":"Whisper (whisper.cpp)","capability":"stt","provider":"OpenAI / ggml","hosting":"self-hosted","access":"geplant","endpoint":"self-hosted geplant · ODER via Together sofort","health":null,"cost":{"model":"self-hosted","unit":"compute","eur":0},"status":"eval","priority":2,"quality_tier":"high","latency_class":"medium","params":{"streaming":false,"languages":"99","gpu":true},"test":"via Together whisper-large-v3 sofort","source":"whisper.cpp ggml; Together-Roster","notes":"Self-hosted eur:0 geplant. Sofort ueber Together (openai/whisper-large-v3) nutzbar."},{"id":"together-stt","name":"Together STT (Whisper-large-v3)","capability":"stt","provider":"Together AI","hosting":"api","access":"bbe-ai-gateway","endpoint":"https://api.together.xyz/v1/audio/transcriptions","health":null,"cost":{"model":"per-minute","unit":"audio-min","eur":0},"status":"key-ready","priority":3,"quality_tier":"high","latency_class":"fast","params":{"streaming":false,"model":"whisper-large-v3"},"test":"curl -H 'Authorization: Bearer $TOGETHER_API_KEY' --data-binary @audio.wav https://api.together.xyz/v1/audio/transcriptions","source":"together-roster transcribe; Key da","notes":"Whisper ohne eigene Installation. Roster-Eintrag, per-Endpoint noch zu verdrahten."},{"id":"deepgram-stt","name":"Deepgram STT (nova-3)","capability":"stt","provider":"Deepgram","hosting":"api","access":"bbe-deepgram","endpoint":"POST /api/deepgram/transcribe","health":"GET /api/deepgram/health","cost":{"model":"per-minute","unit":"audio-min","eur":null},"status":"module-ready","priority":4,"quality_tier":"high","latency_class":"fast","params":{"streaming":true,"languages":"36","model":"nova-3"},"test":"curl --data-binary @audio.wav http://127.0.0.1:4610/api/deepgram/transcribe","source":"@bbe/deepgram + bbe-voice deepgram-Adapter","notes":"Modul fertig. Key-Datei nonempty ABER INDEX.md sagt 'noch leer' -> VERIFIZIEREN (bestimmt live vs. mock)."},{"id":"aqua-voice","name":"Aqua Voice (STT)","capability":"stt","provider":"Aqua Voice","hosting":"api","access":"kein-wrapper","endpoint":"kein Wrapper gebaut","health":null,"cost":{"model":"per-minute","unit":"audio-min","eur":null},"status":"key-orphan","priority":5,"quality_tier":"high","latency_class":"fast","params":{},"test":"n/a — Integration fehlt","source":"secrets/voice-ai (Key vorhanden)","notes":"KEY LIEGT, aber KEIN Code-Wrapper irgendwo. Klarer Modul-Kandidat — bezahlter Zugang ungenutzt."},{"id":"voicebox-stt","name":"Voicebox (self-hosted STT)","capability":"stt","provider":"self-hosted","hosting":"self-hosted","access":"bbe-voice","endpoint":"127.0.0.1:17493 /transcribe","health":null,"cost":{"model":"self-hosted","unit":"compute","eur":0},"status":"live","priority":6,"quality_tier":"mid","latency_class":"fast","params":{"languages":"23"},"test":"bbe-voice /v1/transcribe (adapter voicebox)","source":"bbe-voice src/adapters/voicebox.ts","notes":"Lokaler eur:0-Dienst, MIT. Teil des bbe-voice Voice-Gateways."},{"id":"together-cartesia-tts","name":"Together · Cartesia Sonic-2 (TTS)","capability":"tts","provider":"Together / Cartesia","hosting":"api","access":"api-key","endpoint":"https://api.together.xyz/v1/audio/speech (OpenAI-kompatibel)","health":null,"cost":{"model":"per-character","unit":"1k chars","eur":0},"status":"key-ready","priority":1,"quality_tier":"frontier","latency_class":"fast","params":{"streaming":true,"languages":"DE+EN verifiziert","voices":["german reporter man","german conversational woman","friendly german man"]},"test":"curl -H 'Authorization: Bearer $TOGETHER_API_KEY' -d '{\"model\":\"cartesia/sonic-2\",\"input\":\"Hallo\",\"voice\":\"german conversational woman\"}' https://api.together.xyz/v1/audio/speech","source":"voice_provider_landscape_state (echtes DE verifiziert 2026-07-17)","notes":"BESTE sofort nutzbare DE-TTS: Key gueltig, kommerziell, eur:0-Guthaben. Auch sonic-3/Rime/MiniMax/Orpheus. (Gateway-Roster-tts=Kokoro hat KEIN DE -> Sonic-2 direkt nutzen.)"},{"id":"piper-tts","name":"Piper (self-hosted TTS)","capability":"tts","provider":"OHF-Voice / self-hosted","hosting":"self-hosted","access":"self-hosted","endpoint":"178 /lab/assistant/voice","health":"/lab/assistant/voice-status","cost":{"model":"self-hosted","unit":"compute","eur":0},"status":"live","priority":2,"quality_tier":"mid","latency_class":"fast","params":{"streaming":false,"languages":"DE"},"test":"GET /lab/assistant/voice-status (178)","source":"voice_provider_landscape_state","notes":"Traegt aktuell den Sprach-Kanal. GPL-3.0 (eigener Prozess ok, nicht ins Binary linken)."},{"id":"voicebox-tts","name":"Voicebox (7 self-hosted TTS-Engines + Cloning)","capability":"tts","provider":"self-hosted","hosting":"self-hosted","access":"bbe-voice","endpoint":"127.0.0.1:17493 /speak","health":null,"cost":{"model":"self-hosted","unit":"compute","eur":0},"status":"live","priority":3,"quality_tier":"high","latency_class":"medium","params":{"engines":["Qwen3-TTS","Qwen CustomVoice","LuxTTS","Chatterbox","HumeAI TADA","Kokoro"],"languages":"23","voice_cloning":true},"test":"bbe-voice /v1/speak (adapter voicebox)","source":"bbe-voice src/adapters/voicebox.ts","notes":"eur:0, MIT, DE + Zero-Shot-Cloning. Quality-Gate human_score>=0.95 vor Live."},{"id":"elevenlabs-tts","name":"ElevenLabs TTS","capability":"tts","provider":"ElevenLabs","hosting":"api","access":"bbe-voice","endpoint":"bbe-voice elevenlabs-Adapter · /v1/text-to-speech","health":null,"cost":{"model":"per-character","unit":"1k chars","eur":null},"status":"key-gated","priority":4,"quality_tier":"frontier","latency_class":"medium","params":{"streaming":true,"voices":"20 premade (DE: 11 verifiziert)","languages":"29"},"test":"bbe-voice ElevenLabsAdapter (Key da)","source":"bbe-voice src/adapters/elevenlabs.ts; secrets/_elevenlabs","notes":"Adapter existiert + Key da, ABER Free-Tier: pro/library-Stimmen 402, nur 20 premade, NICHT kommerziell. Kommerziell ab $6/Mon (Operator-Gate)."},{"id":"deepgram-tts","name":"Deepgram TTS (aura-2)","capability":"tts","provider":"Deepgram","hosting":"api","access":"bbe-deepgram","endpoint":"POST /api/deepgram/speak","health":"GET /api/deepgram/health","cost":{"model":"per-character","unit":"1k chars","eur":null},"status":"module-ready","priority":5,"quality_tier":"high","latency_class":"fast","params":{"streaming":true,"model":"aura-2"},"test":"curl -d '{\"text\":\"Hallo\"}' http://127.0.0.1:4610/api/deepgram/speak","source":"@bbe/deepgram + bbe-voice","notes":"aura-2 default EN; DE-Stimme in brand.config. Key-Status wie Deepgram-STT (zu verifizieren)."},{"id":"google-chirp3-tts","name":"Google Chirp 3 HD (TTS)","capability":"tts","provider":"Google Cloud","hosting":"api","access":"geplant","endpoint":"Google Cloud Text-to-Speech (Chirp 3 HD)","health":null,"cost":{"model":"per-character","unit":"1M chars","eur":0},"status":"key-missing","priority":6,"quality_tier":"frontier","latency_class":"medium","params":{"voices":"30 de-DE (= en-US Auswahl)","free":"1M chars/Mon inkl. kommerziell"},"test":"Google Cloud TTS (Service-Account noetig)","source":"voice_provider_landscape_state","notes":"Bester DE-Kandidat (DE vollwertig), 1 Mio Zeichen/Mon frei. Braucht GCP-Key (Operator-Gate)."},{"id":"grok-voice","name":"Grok Voice (SpaceXAI Realtime)","capability":"voice-agent","provider":"xAI / SpaceXAI","hosting":"api","access":"geplant","endpoint":"/v1/tts · /v1/stt · grok-voice-latest (OpenAI-Realtime)","health":null,"cost":{"model":"per-minute","unit":"conv-min","eur":null},"status":"key-missing","priority":2,"quality_tier":"high","latency_class":"fast","params":{"duplex":true,"stt_cost":"$0,10/h","tts_cost":"$15/1M chars","agent_cost":"$0,05/Min"},"test":"OpenAI-Realtime-Client gegen SpaceXAI (Key noetig)","source":"voice_provider_landscape_state","notes":"GA+oeffentlich, DE, wire-kompatibel OpenAI-Realtime -> passt direkt in Slice-2-Gateway. Paid-Key (Operator-Gate)."},{"id":"deepgram-voice-agent","name":"Deepgram Voice-Agent","capability":"voice-agent","provider":"Deepgram","hosting":"api","access":"bbe-deepgram","endpoint":"WS /api/deepgram/agent","health":"GET /api/deepgram/health","cost":{"model":"per-minute","unit":"conv-min","eur":null},"status":"module-ready","priority":1,"quality_tier":"high","latency_class":"fast","params":{"streaming":true,"duplex":true},"test":"WS-Client an /api/deepgram/agent","source":"@bbe/deepgram config.js + agent-proxy.js","notes":"STT+LLM+TTS Full-Duplex. think-LLM = BYO server-seitig (default gpt-4o-mini). Auch bbe-llm voiceCommand-Pipeline (STT->ask->TTS)."}]}