These templates show the supported Aegis pattern for OpenAI-compatible providers:
- start from provider-native inputs (
OPENROUTER_API_KEY,AZURE_OPENAI_API_KEY,OLLAMA_MODEL, etc.), - map them into the allowlisted
ANTHROPIC_*session env vars, - launch Aegis with those resolved values.
glm.aegis.config.jsonopenrouter.aegis.config.jsonlm-studio.aegis.config.jsonollama.aegis.config.jsonazure-openai.aegis.config.jsonrun-example.mjs— resolves${VAR}placeholders and starts Aegis
The raw config files are safe to commit because they contain only placeholders, never real credentials.
npm run build
node examples/byo-llm/run-example.mjs --list
OPENROUTER_API_KEY="<token>" node examples/byo-llm/run-example.mjs openrouter
LM_STUDIO_MODEL="your-model-id" node examples/byo-llm/run-example.mjs lm-studio
AZURE_OPENAI_API_KEY="<token>" \
AZURE_OPENAI_BASE_URL="https://your-resource.openai.azure.com/openai/v1" \
AZURE_OPENAI_DEPLOYMENT="your-deployment" \
node examples/byo-llm/run-example.mjs azure-openaiPass extra Aegis CLI arguments after --:
OPENROUTER_API_KEY="<token>" \
node examples/byo-llm/run-example.mjs openrouter -- --port 9200Use --print-config if you only want to inspect the fully resolved
aegis.config.json payload.