Skip to main content

AI Provider Connectors

DgiDgi supports 25+ AI providers for LLM inference, giving you flexibility to choose the best model for your use case, budget, and compliance requirements.

Supported Providers​

Tier 1: Major Cloud Providers​

ProviderModelsKey FeaturesPricing
OpenAIGPT-4o, GPT-4o-mini, o1-previewIndustry standard, vision, toolsPay-per-token
AnthropicClaude 3.5 Sonnet, Claude 3 Haiku200K context, safety-focusedPay-per-token
Google GeminiGemini 1.5 Pro, Gemini 1.5 Flash2M context, multimodalPay-per-token
Azure OpenAIGPT-4o (Azure), GPT-4o-mini (Azure)Enterprise compliance, SLAPay-per-token
AWS BedrockClaude, Llama, Titan, CohereAWS integration, regulatedPay-per-token

Tier 2: High-Performance Inference​

ProviderModelsKey FeaturesPricing
GroqLlama 3.3 70B, Mixtral 8x7BUltra-fast LPU inferencePay-per-token
CerebrasLlama 3.3 70B, Llama 3.1 8B2000+ tokens/secPay-per-token
SambaNovaLlama 3.3 70B, DeepSeek R1, QwenEnterprise-gradePay-per-token
FireworksLlama 3.3 70B, DeepSeek V3Fast, cost-effectivePay-per-token
LeptonLlama 3.1 405B, Llama 3.3 70BServerless GPUPay-per-token

Tier 3: Specialized Providers​

ProviderModelsKey FeaturesPricing
MistralMistral Large, CodestralEuropean, coding-focusedPay-per-token
CohereCommand R+Enterprise embeddings, RAGPay-per-token
PerplexitySonar LargeSearch-augmented, real-timePay-per-token
xAIGrok 2, Grok VisionMultimodal, reasoningPay-per-token
DeepSeekDeepSeek V3, DeepSeek CoderCost-effective, codingPay-per-token

Tier 4: Aggregators & Open Source​

ProviderModelsKey FeaturesPricing
OpenRouter100+ modelsUnified API, fallbackPay-per-token
Together AILlama, Qwen, MixtralTraining + inferencePay-per-token
ReplicateLlama, Mistral, customModel marketplacePay-per-use
Hugging FaceOpen-source modelsCommunity modelsFree tier
GitHub ModelsCopilot Enterprise/ProGitHub integrationSubscription

Tier 5: Self-Hosted​

ProviderModelsKey FeaturesPricing
OllamaLlama, Qwen, DeepSeekLocal, private, FREEFree
DgiDgi Self-HostedAny OpenAI-compatibleYour infrastructureYour cost

Model Catalog​

Fast Tier (Low Latency)​

Best for interactive chat, quick responses:

  • GPT-4o-mini (OpenAI)
  • Claude 3 Haiku (Anthropic)
  • Gemini 1.5 Flash (Google)
  • Llama 3.3 70B (Groq, Cerebras)
  • Mixtral 8x7B (Groq)

Balanced Tier (Cost/Performance)​

Best for general tasks:

  • GPT-4o (OpenAI)
  • Claude 3.5 Sonnet (Anthropic)
  • Gemini 1.5 Pro (Google)
  • Mistral Large (Mistral)
  • DeepSeek V3 (DeepSeek)

Powerful Tier (Maximum Capability)​

Best for complex reasoning, code generation:

  • o1-preview (OpenAI)
  • Claude 3.5 Sonnet (Anthropic)
  • Llama 3.1 405B (Together, Replicate, Lepton)
  • Grok 2 (xAI)

Reasoning Tier (Extended Thinking)​

Best for multi-step problems:

  • o1-preview (OpenAI)
  • DeepSeek R1 Distill (SambaNova)

Provider Configuration​

Platform-Level (Admin)​

Configure default providers for all tenants:

# Environment variables
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GROQ_API_KEY=gsk_...
GEMINI_API_KEY=...

Tenant-Level (User)​

Users can add their own API keys via Settings > Connectors > AI.

Routing & Fallback​

DgiDgi's LLM Gateway automatically routes requests based on:

  1. Task Type: Code generation, chat, analysis, etc.
  2. Required Capabilities: Vision, tools, reasoning
  3. Cost Optimization: Tenant budgets, token limits
  4. Availability: Provider health, rate limits

Fallback chain example:

Primary: OpenAI GPT-4o
  -> Fallback 1: Anthropic Claude 3.5 Sonnet
    -> Fallback 2: Groq Llama 3.3 70B
      -> Fallback 3: Gemini 1.5 Pro

Cost Management​

Per-Tenant Budgets​

Set daily/monthly limits per tenant to control costs.

Model Tiers​

  • Fast: ~$0.0001-0.001/1K tokens
  • Balanced: ~$0.001-0.005/1K tokens
  • Powerful: ~$0.005-0.015/1K tokens
  • Reasoning: ~$0.015-0.06/1K tokens

Free Options​

  • Ollama (local, requires hardware)
  • Hugging Face free tier
  • DgiDgi self-hosted (your infrastructure)

Self-Hosted LLMs (DgiDgi Provider)​

Deploy your own LLMs and connect them to DgiDgi:

  1. Deploy an OpenAI-compatible API server (vLLM, text-generation-inference, etc.)
  2. Configure the DgiDgi provider:
    • Base URL: Your LLM server endpoint (e.g., https://your-llm.example.com/v1)
    • API Key: Optional authentication key
  3. Models will be available to all users in your tenant

Supported frameworks:

  • vLLM
  • text-generation-inference (TGI)
  • llama.cpp server
  • Ollama (remote)
  • Any OpenAI-compatible API