📊 voyage-4-large newest flagship; voyage-3.5 default; voyage-code-3 / voyage-code-4 for code
🌍 voyage-multilingual-2 + domain-tuned (finance, law)
🔑 API key from dash.voyageai.com/api-keys

Setup Guide →

Jina AI

Embeddings + reranking

📊 jina-embeddings-v3 multilingual flagship
🔄 jina-reranker-v2 for retrieval reranking
🔍 jina-colbert-v2 late-interaction retrieval
🔑 API key from jina.ai

🎨 Direct Image Generation

Specialised image-gen providers (in addition to Vertex Imagen / OpenAI DALL-E / Anthropic / Bedrock):

Stability AI

Stable Image Ultra/Core + SD 3.5 family

🎨 Stable Image Ultra (flagship), Core (fast), SD 3.5 Large/Large-Turbo/Medium
🖼️ PNG output, aspect-ratio + negative-prompt + seed support
🔑 API key from platform.stability.ai/account/keys
🔄 Aliases: stability-ai, sd

Setup Guide →

Ideogram

Strong typography + design-focused image generation

📝 V3 default; V2/V2-Turbo/V1 also supported
🎨 magic_prompt + style + aspect_ratio controls
🔑 API key from developer.ideogram.ai

Recraft

Vector / illustration-focused image generation

🎨 recraftv3 (raster), recraftv3-svg (vector), recraftv2
📐 OpenAI-compat shape + style + size controls
🔑 API token from recraft.ai/api

💻 Local Providers

Run models entirely on your own hardware — no API key or internet required for inference:

Ollama

Run open-source models locally with full privacy

🖥️ 100% local inference — no data leaves your machine
🦙 70+ models: Llama, Mistral, Qwen, DeepSeek, Gemma, Phi, CodeLlama
🌐 Native Ollama API and OpenAI-compatible mode
🆓 No API key required

Setup Guide →

LM Studio

Run any supported model locally with a GUI app

🖥️ Download and run models via the LM Studio desktop application
🔍 Auto-discovers the loaded model from /v1/models (no model name required)
🌐 OpenAI-compatible API at http://localhost:1234/v1 by default
🆓 No API key needed for local use (key optional for reverse-proxy setups)
🔄 Aliases: lmstudio, lms

Setup Guide →

llama.cpp

High-performance local inference via llama-server

⚡ Run GGUF models with llama-server at http://localhost:8080/v1 by default
🔍 Auto-discovers the loaded model from /v1/models
🛠️ Tool support requires --jinja flag when starting llama-server
🆓 No API key needed for local use (key optional for reverse-proxy setups)
🔄 Aliases: llama.cpp

Setup Guide →

🔌 Aggregators & Proxies

Access multiple providers through unified interfaces:

OpenRouter

300+ models from 60+ providers

🌐 Single API across many AI providers (Anthropic, OpenAI, Google, Meta, etc.)
⚡ Automatic failover and routing
💰 Competitive pricing with cost optimization
🎯 Zero lock-in - switch models instantly
📊 Usage tracking dashboard
🆓 Free models available

Setup Guide →

OpenAI Compatible

OpenRouter, vLLM, LocalAI, and more

🌐 100+ models through OpenRouter
💻 Local deployment with vLLM
🔓 Self-hosted with LocalAI
🔄 Drop-in OpenAI replacement

Setup Guide →

LiteLLM

100+ providers through proxy

🔄 Unified API for 100+ providers
📊 Load balancing and fallbacks
💰 Cost tracking
🎯 Model routing

Setup Guide →

🧠 Decision-Only Providers

The providers that serve decide rather than generate/stream: TypeSafe, Laya, XOR, SubQ, Perplexity, Cloudflare Clef, OpenAI Decisions and ScaleDown. Each returns typed boolean/choice/score answers and emits no text, so none appears in generation fallback chains or the health sweep.

TypeSafe (Jev)

Typed, calibrated judgments instead of text

🎯 boolean / choice / score answers, each with a calibrated confidence
⚡ Latency flat in question count — 1 question ~393 ms, 400 questions ~465 ms
💰 ~~$0.00002 per decision (~~$0.042/M input, output billed at zero)
🔌 Two transports: TypeSafe direct, or the Vercel AI Gateway
🛡️ Fails open — with no decision provider configured, every consumer behaves exactly as before
🔑 API key from console.typesafe.ai/keys
🔄 Aliases: jev, typesafe-ai

Setup Guide →

Laya

Open-weights decision provider — the same typed boolean/choice/score answers as Jev, from Convai Innovations' Apache-2.0 checkpoints, at a Laya server or LiteLLM proxy route you configure

🧭 Serves decide only; built-in features use it when its key and base URL are set and neither TYPESAFE_API_KEY nor AI_GATEWAY_API_KEY is
📏 Refuses more than ~768 tokens of state (320 on english) before any network call, since its encoders read only 1,024 (512)
🔌 No built-in endpoint: LAYA_BASE_URL (or credentials.laya.baseURL) names a Laya server or a LiteLLM pass-through route to one
🔑 LAYA_API_KEY is the key that endpoint accepts; on LiteLLM, the route must be in the key's Allowed Routes

Setup Guide →

XOR

Open-weights decision provider — the same typed boolean/choice/score answers as Jev, plus value extraction, from xor-latest (text) and xor-omni (an image and audio), Juspay's Apache-2.0 model, at a deployment or LiteLLM proxy route you configure

🧭 Serves decide only; built-in features use it when its key and base URL are set and no TypeSafe key, no AI_GATEWAY_API_KEY and no Laya key and base URL is
🖼️ Takes one image and one audio clip (the first 30 seconds are heard) with a decision on xor-omni, as a Buffer, a local file path or a data: URL; a video is refused, because XOR's gateway drops it
📏 Refuses more than 120,000 estimated tokens of state on xor-latest (28,000 on xor-omni) or more than 128 questions before any network call
🔌 No built-in endpoint: XOR_BASE_URL (or credentials.xor.baseURL) is the origin of a deployment or of a LiteLLM route to one; requests go to <base URL>/v1/decisions
🔑 XOR_API_KEY is the key that endpoint accepts; on LiteLLM, the key's team must allow xor-latest (and xor-omni for media)
📦 Weights and setup: huggingface.co/juspay/xor

Setup Guide →

SubQ

Hosted decision provider — the same typed boolean/choice/score answers as Jev, from Subquadratic's subq-snap-preview (public preview), at https://api.subq.ai

🧭 Serves decide only (provider id subq; the aliases subquadratic and subq-snap work too); built-in features use it when it is configured and none of TypeSafe, Laya or XOR is — it comes ahead of Perplexity and Cloudflare Clef, because its key is shared with no other provider
🔑 SUBQ_API_KEY alone configures it (keys come from console.subq.ai, shown once); SUBQ_MODEL is optional
🖼️ No image or video input yet: SubQ's image format is not one the decide catalog can describe, so images are refused before any request
📏 NeuroLink reports an advisory state window (376,000 estimated tokens; SubQ's own 400 response states a maximum model length of 393,216) and enforces none; no question cap is declared
🔌 Hosted endpoint, so no base URL is needed; SUBQ_BASE_URL (or credentials.subq.baseURL) can name another origin

Setup Guide →

Perplexity Decisions

Hosted decision provider — the same typed boolean/choice/score answers as Jev, from Perplexity's pplx-decider-v1-27b, at https://api.perplexity.ai

🧭 Serves decide only (provider id perplexity-decider, not the Sonar text provider perplexity); built-in features use it when it is configured and no TypeSafe key, no AI_GATEWAY_API_KEY, no Laya key and base URL, no XOR key and base URL and no SUBQ_API_KEY is
🔑 PERPLEXITY_API_KEY alone configures it, and it is the same key the Perplexity text provider reads, so a key set for Sonar also lets built-in features use it when none of TypeSafe, Laya, XOR or SubQ is configured; the guide lists what each consumer then sends
🖼️ Takes up to 8 PNG, JPEG or WebP images with a decision; no video
💰 $0.04 per million input tokens (image tokens included), output free, as Perplexity documents it
📏 Refuses more than about 100,000 estimated tokens of state (NeuroLink's own window; the server's ceiling is 262,144 input tokens), or more than 128 questions in a decide() call (tryDecide() splits them), before any network call. The 128-question cap, the server's token ceiling and the latency were measured on a real account in October 2026
🔌 Hosted endpoint, so no base URL is needed; PERPLEXITY_DECIDER_BASE_URL (or credentials.perplexityDecider.baseURL) can name another origin

Setup Guide →

Cloudflare Clef

Hosted decision provider on Workers AI — the same typed boolean/choice/score answers as Jev, from Cloudflare's clef (27B) and clef-flash (9B), at https://api.cloudflare.com/client/v4

🧭 Serves decide only (provider id cloudflare-clef, not the Workers AI text provider cloudflare); built-in features use it only when it is configured and none of TypeSafe, Laya, XOR, SubQ or Perplexity is — it comes after those five
🔑 CLOUDFLARE_API_KEY (Workers AI permission) and CLOUDFLARE_ACCOUNT_ID configure it, and they are the same two variables the Cloudflare text provider reads, so a pair set for that provider also lets built-in features use it when none of TypeSafe, Laya, XOR, SubQ or Perplexity is configured
🖼️ Takes up to 4 PNG, JPEG or WebP images with a decision; no video
💰 $0.24 per million input tokens for clef, $0.09 for clef-flash; no output price is listed on Cloudflare's pricing page
📏 The state window changed in October 2026 (hosted service or model: unknown): until 2026-10-04 the endpoint ignored text past about 2,048 tokens without an error, and on 2026-10-07 it read states of at least 190,153 tokens (clef-flash) and 52,154 (clef); NeuroLink refuses more than 16,000 estimated tokens (digits count a token each, punctuation 0.75, emoji 3), more than 64 questions in a decide() call (tryDecide() splits them) and a request over 256,000 bytes, base64 image data included, before any network call. The state cut and the request ceiling were measured on a real account in October 2026, with the follow-up text cuts and API limits checked on both models; see the guide for the tested shapes and dates
🔌 Hosted endpoint, so no base URL is needed; CLOUDFLARE_CLEF_BASE_URL (or credentials.cloudflareClef.baseURL) can name another base

Setup Guide →

OpenAI Decisions

Hosted decision provider (public beta) — the same typed boolean/choice/score answers as Jev, from OpenAI's Decisions API (gpt-6-luna), at https://api.openai.com/v1/decisions

🧭 Serves decide only (provider id openai-decisions, not the OpenAI chat provider openai); built-in features use it only when it is configured and none of TypeSafe, Laya, XOR, SubQ, Perplexity or Cloudflare Clef is — it comes seventh, ahead of only ScaleDown
🔑 OPENAI_API_KEY alone configures it, and it is the same key the OpenAI chat provider reads, so a key set for chat also lets built-in features (and a transcription dictionary's correction check) use it, and make paid decision calls, when none of the six providers ahead of it is configured, unless OPENAI_BASE_URL, OPENAI_API_BASE or credentials.openai.baseURL names a host other than https://api.openai.com (the key is then taken to be a gateway's and is not used by default; a call that names the provider still uses it and sends it to api.openai.com); the guide lists what each consumer then sends
🖼️ Takes up to 8 images as inline data URLs with a decision (NeuroLink's own allowance, not OpenAI's); no video
💰 $0.10 per million input tokens, input only, per OpenAI's Decisions guide as read on 2026-10-07; regional and long-context multipliers are not modelled
📏 Questions per request are capped at 200, the figure in OpenAI's Python SDK, and tryDecide() splits a longer set at it; going over 200 was not tried against the live endpoint, and option limits, image formats, rate limits and latency are not established (none measured). NeuroLink reports 922,000 state tokens as an advisory figure and never refuses on it
🔌 Hosted endpoint, so no base URL is needed; OPENAI_DECISIONS_BASE_URL (or credentials.openaiDecisions.baseURL) can name another origin

Setup Guide →

ScaleDown

Hosted decision provider — the same typed boolean/choice/score answers as Jev, from ScaleDown's Classify models (classify-1 by default), at https://api.scaledown.xyz/v1/scaledown

🧭 Serves decide only (provider id scaledown; the alias scaledown-classify works too); built-in features use it only when it is configured and none of TypeSafe, Laya, XOR, SubQ, Perplexity, Cloudflare Clef or OpenAI Decisions is — it comes last, by choice and not because its key is shared
🔑 SCALEDOWN_API_KEY alone configures it, and no other provider reads that variable; the key is sent in an x-api-key header, never as a Bearer token
🖼️ Text only: images and video are refused before any request, and the state is always sent as a string (a string unchanged, anything else as its JSON text; a blank string is refused locally)
💰 USD 0.05 per million input tokens, output not charged: a list price reported by the marketing session on 2026-10-15, not measured here. A choice question is scored once per option, so input tokens grow with options times state length
📏 Context size, latency, rate limits and the option, question and level maximums are not measured; no state window is declared and nothing is refused on size
🔌 Hosted endpoint, so no base URL is needed; SCALEDOWN_BASE_URL (or credentials.scaledown.baseURL) can name an origin or an origin with a path prefix

Setup Guide →

🧩 Additional Catalog Providers

Every provider below is a Tier-2 catalog entry. Provider-specific behavioural quirks may apply, so check the provider's catalog file. The whole integration is one JSON file under src/lib/providers/catalog/. Each page is generated from that file, which is also what the CI onboarding gate reads.

API Route

Claude Sonnet 4.6

🤖 8 models; default claude-sonnet-4-6 (1M context)
🛠️ Native tool calling + structured output together
💳 Free tier available
✅ Roster verified 2026-09-17 (authenticated GET /v1/models)
🔑 API key from api-route.com
🔄 Aliases: apiroute

Setup Guide →

Baseten

GLM 5.3 Flash

🤖 16 models; default zai-org/GLM-5.3-Flash (1M context)
🛠️ Native tool calling + structured output together
💳 Free tier available
✅ Roster verified 2026-09-03 (authenticated GET /v1/models)
🔑 API key from app.baseten.co

Setup Guide →

Friendli

zai-org/GLM-5.3

🤖 7 models; default zai-org/GLM-5.3 (1M context)
🛠️ Native tool calling — not combined with structured output in one request (HTTP 422 if both are set)
💳 Free tier available (rate limits are tight — pace requests 20s+ apart)
✅ Roster verified 2026-09-06 (authenticated GET /serverless/v1/models)
🔑 API key from suite.friendli.ai

Setup Guide →

GMI Cloud

MiniMaxAI/MiniMax-M3

🤖 1 model; default MiniMaxAI/MiniMax-M3 (1M context)
🛠️ Native tool calling + structured output together
💳 Free tier available
✅ Roster verified 2026-09-03 (authenticated GET /v1/models)
🔑 API key from console.gmicloud.ai
🔄 Aliases: gmi-cloud

Setup Guide →

Inception Labs

Mercury 2, Inception's enterprise diffusion LLM (dLLM)

🤖 1 model; default mercury-2 (125K context)
🛠️ Native tool calling + structured output together
💳 Free tier available
✅ Roster verified 2026-09-03 (authenticated GET /v1/models)
🔑 API key from platform.inceptionlabs.ai/dashboard/api-keys
🔄 Aliases: inception, mercury

Setup Guide →

io.net Intelligence

Meta: Llama 3.3 70B Instruct

🤖 34 models; default meta-llama/Llama-3.3-70B-Instruct (125K context)
🛠️ Native tool calling + structured output together
💳 Free tier available
✅ Roster verified 2026-09-03 (authenticated GET /v1/models)
🔑 API key from ai.io.net
🔄 Aliases: io-net

Setup Guide →

Mancer

DeepSeek V4 Flash

🤖 10 models; default deepseek-v4-flash (1M context)
⚠️ No tool calling — text generation and structured output only
💳 Free tier available
✅ Roster verified 2026-09-03 (authenticated GET /oai/v1/models; full response retained as evidence/mancer-roster-authenticated.json in the campaign scratchpad and every catalog price/limit machine-checked against it (Mancer re-prices — gpt-oss-120b input moved 0.024 → 0.022 within the day))
🔑 API key from mancer.tech/dashboard
🔄 Aliases: mancer-tech

Setup Guide →

Morph

Morph v3 Large

🤖 2 models; default morph-v3-large (262K context)
⚠️ No tool calling — text generation and structured output only
💳 Free tier requires a card on file (rate-limited to 5 req/min until added)
✅ Roster verified 2026-09-05 (authenticated GET /v1/models)
🔑 API key from morphllm.com/dashboard
🔄 Aliases: morphllm

Setup Guide →

Novita AI

zai-org/GLM-5.3 Flash

🤖 5 models; default zai-org/glm-5.3-flash (128K context)
🛠️ Native tool calling + structured output together
💳 Free tier available
✅ Roster verified 2026-09-05 (authenticated GET /openai/v1/models)
🔑 API key from novita.ai/settings/key-management

Setup Guide →

Upstage

Solar Pro 4

🤖 10 models; default solar-pro4 (512K context)
🛠️ Native tool calling + structured output together
💳 Free tier available
✅ Roster verified 2026-09-03 (authenticated GET /v1/models)
🔑 API key from console.upstage.ai/api-keys
🔄 Aliases: solar

Setup Guide →

🎙️ Voice Providers

Synthesize speech, transcribe audio, or run live voice sessions. Voice providers are separate from LLM providers — they handle audio I/O rather than text generation.

Text-to-Speech (TTS)

OpenAI TTS

Highest-quality text-to-speech

🎙️ Voices: alloy, echo, fable, onyx, nova, shimmer
🎵 Models: tts-1 (fast) and tts-1-hd (high quality)
🎼 Formats: MP3, WAV, OGG, Opus
🔑 Auth: API Key (OPENAI_API_KEY)

Setup Guide →

ElevenLabs

Best multilingual and voice-cloning TTS

🌍 Supports 30+ languages with natural prosody
🎭 Custom voice cloning from short audio samples
🎼 Formats: MP3, WAV (raw PCM, surfaced as pcm16), Opus (Ogg container)
🔑 Auth: API Key (ELEVENLABS_API_KEY)

Setup Guide →

Google TTS

1M characters/month free tier

💰 Generous free tier for standard voices
🌍 380+ voices across 50+ languages
🎼 Formats: MP3, WAV, OGG
🔑 Auth: Service Account

Setup Guide →

Azure TTS

Enterprise TTS with full SSML support

🏢 Fine-grained prosody control via SSML
🌍 400+ neural voices, 140+ languages
🎼 Formats: MP3, WAV (PCM), Opus (Ogg container)
🔑 Auth: API Key + Region

Setup Guide →

Fish Audio

Low-cost TTS with 15s voice cloning

💰 ~80% cheaper than ElevenLabs
🎭 15-second reference audio → custom voice
🌍 14 languages
🎼 Formats: MP3, WAV, PCM16 (raw)
🔑 Auth: API Key (FISH_AUDIO_API_KEY)

Setup Guide →

60db

Workspace voice catalogs and HTTP text-to-speech.

WAV or raw PCM16 output at 24 kHz
Workspace API key and voice UUID
Setup guide

Cartesia

Low-latency Sonic models — synchronous + streaming

⚡ Sub-second turnaround on the synchronous /tts/bytes endpoint
🌊 Separate WebSocket streaming flow via CartesiaStream (voice server)
🎭 Voice cloning via dashboard upload
🎼 Formats: MP3 (44.1 kHz), WAV (PCM s16le @ 44.1 kHz), PCM16 (raw @ 24 kHz)
🔑 Auth: API Key (CARTESIA_API_KEY)

Setup Guide →

Speech-to-Text (STT)

Whisper (OpenAI)

Highest transcription accuracy

🎯 Best-in-class accuracy on diverse audio
🌍 Multilingual with automatic language detection
🎼 Formats: WAV, MP3, M4A, FLAC, OGG, OPUS, WEBM, MP4, MPEG, MPGA
🔑 Auth: API Key (OPENAI_API_KEY)

Setup Guide →

Deepgram

Real-time streaming transcription via WebSocket

⚡ Sub-300 ms word-level results over WebSocket
🌊 REST batch and WebSocket streaming modes
🎼 Formats: WAV, MP3, OGG, FLAC
🔑 Auth: API Key (DEEPGRAM_API_KEY)

Setup Guide →

Google STT

125+ languages with speaker diarization

🌍 Best fit for existing Google Cloud users
👥 Speaker diarization and multi-channel audio
🎼 Formats: WAV, FLAC, MP3, OGG
🔑 Auth: API Key (GOOGLE_AI_API_KEY / GEMINI_API_KEY) or Service Account (GOOGLE_APPLICATION_CREDENTIALS)

Setup Guide →

Azure STT

Enterprise STT with custom model training

🏢 Batch transcription and custom model support
🔒 Compliance controls for regulated industries
🎼 Formats: WAV (PCM), Ogg/Opus — convert MP3 to WAV first
🔑 Auth: API Key + Region

Setup Guide →

Realtime Voice

Realtime providers maintain a persistent bidirectional WebSocket connection, enabling low-latency spoken conversation with the AI model.

OpenAI Realtime

Low-latency bidirectional voice over WebSocket

⚡ Full-duplex audio stream with GPT-4o
🎵 Voice activity detection (VAD) built-in
🎼 Formats: WAV, Opus
🔑 Auth: API Key (OPENAI_API_KEY)

Setup Guide →

Gemini Live

Google's native realtime voice API

⚡ Native multimodal realtime session with Gemini
🎵 Supports audio + video input simultaneously
🎼 Formats: WAV, Opus
🔑 Auth: API Key (GOOGLE_AI_API_KEY or GEMINI_API_KEY)

Setup Guide →

🎬 Video Generation

Image-to-video and text-to-video providers (use via output: { mode: "video" }):

Vertex Veo 3.1 (default) — --videoProvider vertex
Kling (PiAPI) — --videoProvider kling (details)
Runway (Gen-3 Alpha / Gen-4 Turbo) — --videoProvider runway
Replicate — Wan-Alpha + many others — --videoProvider replicate (guide)

See Video Generation feature page for the full SDK / CLI surface.

👤 Avatar / Lip-Sync Generation

Talking-head video synthesis from a portrait image + audio (use via output: { mode: "avatar" }):

D-ID — --avatarProvider d-id (text-driven via Microsoft voices, or audio-driven)
HeyGen — --avatarProvider heygen (HeyGen avatar catalog id required)
Replicate (MuseTalk) — --avatarProvider replicate or musetalk (guide)

See docs/provider-integration/21-adding-new-modality.md for the architectural pattern.

🎵 Music / Sound Generation

Music + sound-effect generation (use via output: { mode: "music" }):

Beatoven.ai — --musicProvider beatoven (royalty-free background music)
ElevenLabs Music — --musicProvider elevenlabs-music (short SFX / loops up to 22s; same ELEVENLABS_API_KEY as TTS)
Lyria 3 Pro (Google) — --musicProvider lyria
Replicate (MusicGen) — --musicProvider replicate or musicgen (guide)

Quick Comparison

Provider	Free Tier	Enterprise	GDPR	Latency	Best For
Anthropic	Limited	✅	✅	Low	Reasoning, coding, Claude
Hugging Face	✅	❌	✅	Medium	Open source, experimentation
Google AI	✅	✅	✅	Low	Free tier, Gemini
Mistral AI	❌	✅	✅	Low	EU compliance, cost
OpenRouter	✅	✅	Varies	Low	Multi-model, automatic failover
OpenAI Compatible	Varies	✅	Varies	Varies	Flexibility, local deployment
LiteLLM	❌	✅	Varies	Low	Multi-provider, unified API
Azure OpenAI	❌	✅	✅	Low	Enterprise, Microsoft ecosystem
Vertex AI	❌	✅	✅	Low	Enterprise, GCP ecosystem
AWS Bedrock	❌	✅	✅	Low	Enterprise, AWS ecosystem
DeepSeek	❌	✅	❌	Low	Cost-effective reasoning, R1 model
NVIDIA NIM	❌	✅	Varies	Low	NVIDIA-hosted or self-hosted LLMs
LM Studio	✅ (Local)	❌	✅	Varies	Local GUI model management
llama.cpp	✅ (Local)	❌	✅	Varies	High-performance local GGUF inference
OpenAI TTS	❌	✅	✅	Low	High-quality TTS (tts-1-hd)
ElevenLabs	❌	✅	Varies	Low	Multilingual TTS, voice cloning
Google TTS	✅	✅	✅	Low	Cost-effective TTS, 1M chars free
Azure TTS	❌	✅	✅	Low	Enterprise TTS, SSML support
Fish Audio	❌	✅	Varies	Low	Low-cost TTS, voice cloning, 14 langs
Cartesia	❌	✅	Varies	Low	Low-latency Sonic models
Whisper	❌	✅	✅	Low	Best STT accuracy
Deepgram	❌	✅	Varies	Low	Real-time STT streaming (WebSocket)
Google STT	❌	✅	✅	Low	STT for GCP users, 125+ languages
Azure STT	❌	✅	✅	Low	Enterprise STT, custom models
OpenAI Realtime	❌	✅	✅	Low	Realtime bidirectional voice
Gemini Live	❌	✅	✅	Low	Realtime voice + video (Gemini)

Setup Strategies

Strategy 1: Free Tier First (Recommended for Development)

SDK Usage
CLI Usage

const ai = new NeuroLink({
providers: [
{
name: 'google-ai',
priority: 1,
config: { apiKey: process.env.GOOGLE_AI_KEY },
quotas: { daily: 1500 }
},
{
name: 'openai',
priority: 2,
config: { apiKey: process.env.OPENAI_API_KEY }
}
],
failoverConfig: { enabled: true, fallbackOnQuota: true }
});

    const result = await ai.generate({
      input: { text: "Hello world" }
    });

# Set up environment variables
export GOOGLE_AI_KEY="your-key"
export OPENAI_API_KEY="your-key"

    # Use with automatic failover
    npx @juspay/neurolink generate "Hello world" \
      --provider google-ai

Strategy 2: Multi-Region Enterprise

const ai = new NeuroLink({
  providers: [
    {
      name: "azure-us",
      region: "us-east",
      config: {
        /* Azure US */
      },
    },
    {
      name: "azure-eu",
      region: "eu-west",
      config: {
        /* Azure EU */
      },
    },
    {
      name: "bedrock-us",
      region: "us-east",
      config: {
        /* Bedrock US */
      },
    },
  ],
  loadBalancing: "latency-based",
});

const ai = new NeuroLink({
  providers: [
    {
      name: "mistral",
      priority: 1,
      config: { apiKey: process.env.MISTRAL_API_KEY },
    },
    {
      name: "azure-eu",
      priority: 2,
      config: {
        /* Azure EU region */
      },
    },
  ],
  compliance: {
    framework: "GDPR",
    dataResidency: "EU",
  },
});

Next Steps

Choose a provider based on your requirements (free tier, compliance, region)
Follow the setup guide to get your API key
Configure NeuroLink with the provider
Test the integration with a simple request
Add failover for production reliability

Multi-Provider Failover - High availability patterns
Cost Optimization - Reduce costs by 80-95%
Compliance & Security - GDPR, SOC2, HIPAA
Load Balancing - Distribution strategies
Voice Providers Comparison - TTS, STT, and Realtime capability matrix
Voice Provider Selection - Choosing the right voice provider

🆓 Free Tier Providers​

🤖 Direct AI Providers​

🏢 Enterprise Providers​

🌍 Compliance-Focused​

🧑‍💻 Hosted Inference Providers​

Fireworks AI​

Perplexity​

Cloudflare Workers AI​

Cohere​

Baseten​

GMI Cloud​

Inception Labs​

io.net Intelligence​

Mancer​

Upstage​

API Route​

🔍 Embedding-Only Providers​

Jina AI​

🎨 Direct Image Generation​

Ideogram​

Recraft​

💻 Local Providers​

🔌 Aggregators & Proxies​

🧠 Decision-Only Providers​

🧩 Additional Catalog Providers​

🎙️ Voice Providers​

Text-to-Speech (TTS)​

Speech-to-Text (STT)​

Realtime Voice​

🎬 Video Generation​

👤 Avatar / Lip-Sync Generation​

🎵 Music / Sound Generation​

Quick Comparison​

Setup Strategies​

Strategy 1: Free Tier First (Recommended for Development)​

Strategy 2: Multi-Region Enterprise​

Strategy 3: GDPR Compliance​

Next Steps​

Related Documentation​

🆓 Free Tier Providers

🤖 Direct AI Providers

🏢 Enterprise Providers

🌍 Compliance-Focused

🧑‍💻 Hosted Inference Providers

Fireworks AI

Perplexity

Cloudflare Workers AI

Cohere

Baseten

GMI Cloud

Inception Labs

io.net Intelligence

Mancer

Upstage

API Route

🔍 Embedding-Only Providers

Jina AI

🎨 Direct Image Generation

Ideogram

Recraft

💻 Local Providers

🔌 Aggregators & Proxies

🧠 Decision-Only Providers

🧩 Additional Catalog Providers

🎙️ Voice Providers

Text-to-Speech (TTS)

Speech-to-Text (STT)

Realtime Voice

🎬 Video Generation

👤 Avatar / Lip-Sync Generation

🎵 Music / Sound Generation

Quick Comparison

Setup Strategies

Strategy 1: Free Tier First (Recommended for Development)

Strategy 2: Multi-Region Enterprise

Strategy 3: GDPR Compliance

Next Steps

Related Documentation