LLM providers

Meridian routes all AI calls through a pluggable layer in src/lib/llm/. Choose a provider in config.llm.provider or override via environment variables.

OpenAI-compatible (default)

Works with OpenAI, Azure OpenAI, Groq, Together, and local Ollama.

{
  "llm": {
    "provider": "openai-compatible",
    "baseUrl": "https://api.openai.com/v1",
    "model": "gpt-4.1-mini",
    "apiKey": null
  }
}

Set OPENAI_API_KEY in .env instead of storing the key in config.

Ollama example

OPENAI_BASE_URL=http://localhost:11434/v1
OPENAI_MODEL=llama3.1

Webhook

POST JSON to your endpoint; expect JSON back within 60 seconds.

{
  "llm": {
    "provider": "webhook",
    "webhookUrl": "http://localhost:8080/meridian-llm"
  }
}

Request body shape:

{
  "systemPrompt": "...",
  "userPayload": { },
  "schema": "json"
}

Keyword-only

No LLM calls — deterministic scoring and resume keyword tailoring. Zero API cost.

{
  "llm": {
    "provider": "keyword-only"
  }
}

Test connection

Use Test LLM connection in Settings, or:

curl -X POST http://127.0.0.1:3030/api/llm/test