LLM Provider Integration

yCrash AI can work with following Large Language Model (LLM). Pick the LLM provider that suits your organization’s need.

1. Google Gemini

The shortest path to production. A single key from Google AI Studio powers chat, embeddings, and file analysis, and the Flash models are a strong default for RCA Q&A. Start here unless your organization already standardizes on another vendor.

To configure Gemini, follow the instructions provided here.

2. OpenAI

Best when OpenAI is already your corporate standard for billing, keys, and model governance. You get the GPT-5.6 family — Terra for balanced RCA chat, Sol when the analysis is harder — plus explicit controls for output length and reasoning effort.

To configure OpenAI, follow the instructions provided here.

3. Azure OpenAI

The enterprise-hosted path: same OpenAI-class models, but they run on your Azure resource. Use this when data residency, an existing Azure agreement, or private networking matters. Configuration points at your Azure endpoint and uses Azure deployment names, not public model IDs.

To configure Azure OpenAI, follow the instructions provided here.

4. Anthropic Claude

Choose Claude when you want Anthropic's models for long-form RCA reasoning. Semantic search can run on a local embedding model, so you do not need a second vendor key. On Sonnet 4.6+ you can raise reasoning effort when depth matters more than first-token latency.

To configure Claude, follow the instructions provided here.

5. GitHub Copilot

Reuse a Copilot subscription you already pay for instead of buying a separate LLM API. All yCrash AI chat is billed against that one shared seat. Embeddings stay local; chat attachments stay on the yCrash server because Copilot has no Files API.

To configure GitHub Copilot, follow the instructions provided here.

6. LiteLLM Proxy

Use LiteLLM when your organization already runs (or wants) a central LLM gateway. yCrash points at one proxy URL and virtual key; LiteLLM routes aliases such as rca-chat or rca-claude to OpenAI, Anthropic, Bedrock, Gemini, vLLM, and other backends.

To configure LiteLLM, follow the instructions provided here.