Skip to main content

LiteLLM

LiteLLM clients should use UnifiedMemory's OpenAI-compatible gateway when they want memory injected into chat/completions or responses-style calls.

Pattern​

  • Route through https://edge-api.jithendranara.dev/v1/chat/completions or /v1/responses.
  • Include a scoped key in Authorization.
  • Pass memory options such as container_tag and trust policy when supported.
  • Use direct retain/writeback for post-call capture.

Safety​

Gateway calls must still obey scoped-key identity. Do not use a broad admin key as a LiteLLM provider credential.