OpenAI-Compatible Integration

Sypha offers a "universal bridge" to any AI service that adheres to the OpenAI-standard API protocol. This empowers you to utilise specialized models from providers beyond the primary registries, including:

• Local Inference Engines: Like Ollama or LM Studio.
• Specialized Cloud Rigs: Such as Perplexity, Together AI, or Anyscale.
• Enterprise Gateways: Custom internal AI proxies or self-hosted deployments.

[!NOTE] This guide focuses on third-party compatible services. For the official OpenAI platform, refer to the Dedicated OpenAI Guide.

Core Configuration Parameters

To bridge a compatible provider, you must define three primary variables:

1. Base Service URL: The unique endpoint provided by your service (e.g., https://api.together.xyz/v1).
2. Service API Key: Your authenticated secret provider key.
3. Model Identifier: The specific engine name recognized by your provider.

Implementation Steps

1. Open Workspace Config: Select the gear icon in the Sypha panel.
2. Set Registry: Choose OpenAI Compatible from the provider menu.
3. Define the Base URL: Critical: Ensure this points to the provider's specific V1 endpoint.
4. Credential Entry: Insert your provider-specific API key.
5. Calibrate Model Specifics: Customize advanced parameters if required:
   • Maximum Token Output
   • Context Window Capacity
   • Multi-modal Support (Image/Vision)
   • Fiscal Metrics (Input/Output rates)

Advanced Endpoint Flexibility

Sypha supports both standard base paths and full terminal URLs for greater flexibility:

• Standard Mapping: https://api.service.com/v1
• Granular Mapping: https://your-custom-proxy.internal/api/v2/chat/completions

Common Resolution Steps

• Credential Mismatch: verify the key format aligns with the provider's expectations.
• Endpoint Reachability: Ensure your development environment can communicate with the provided Base URL (especially for local or VPN-restricted services).
• ID Verification: Reference your provider's documentation to ensure the Model ID is typed exactly as required (e.g., mistral-tiny vs Mistral-Tiny).

---

Sypha allows for an unlimited number of OpenAI-compatible configurations, enabling you to switch between local, private, and specialized cloud engines within seconds.