API reference

Chat completions

POSThttps://api.offmem.app/v1/chat/completions

Generates a model response in the OpenAI Chat Completions format. Works with the official OpenAI SDKs.

curl https://api.offmem.app/v1/chat/completions \
  -H "Authorization: Bearer $OFFMEM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "offmem-glm-5.3-flash-abliterated",
    "messages": [{"role": "user", "content": "Explain what an API is in one sentence."}]
  }'

Headers

AuthorizationBearer om_...required: API key. x-api-key also works.
Content-Typeapplication/jsonrequired: JSON body.

Body

modelstringrequired: Model ID. See Models.
messagesarrayrequired: Conversation in the OpenAI format (system, user, assistant, tool). Cannot be empty. image_url parts only on models with vision.
streamboolean: Returns the answer in parts, as server-sent events. Default false.
stream_options.include_usageboolean: With stream, adds a last event with usage.
max_completion_tokens, max_tokensinteger: Output token limit. Also used to reserve balance before the call; without it, the reserve assumes 4,096 output tokens.
ninteger: Only 1. Other values return 400 invalido.
tools, tool_choicearray, string | object: Function calling in the OpenAI format, forwarded to the model.
temperature, top_p, stop, …: Other OpenAI parameters are forwarded to the model as sent. Support depends on the model.

Removed fields

store, metadata, user and safety_identifier are discarded before the request reaches the model.

An OpenAI chat.completion object. model is the offmem ID, and usage has only token counts: prompt_tokens, completion_tokens, total_tokens, prompt_tokens_details.cached_tokens and completion_tokens_details.reasoning_tokens, when the model reports them.

200 OK

{
  "id": "...",
  "object": "chat.completion",
  "created": 1762560000,
  "model": "offmem-glm-5.3-flash-abliterated",
  "choices": [
    {
      "index": 0,
      "message": { "role": "assistant", "content": "..." },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 18,
    "completion_tokens": 32,
    "total_tokens": 50,
    "prompt_tokens_details": { "cached_tokens": 0 }
  }
}

x-offmem-request-idheader: Request ID, the same one shown in the console.

Streaming

With "stream": true the response is text/event-stream: chat.completion.chunk events and a final data: [DONE]. The usage event only comes when stream_options.include_usage is true.

curl -N https://api.offmem.app/v1/chat/completions \
  -H "Authorization: Bearer $OFFMEM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "offmem-glm-5.3-flash-abliterated",
    "stream": true,
    "stream_options": {"include_usage": true},
    "messages": [{"role": "user", "content": "Explain what an API is in one sentence."}]
  }'

Events

text/event-stream

data: {"id":"...","object":"chat.completion.chunk","model":"offmem-glm-5.3-flash-abliterated","choices":[{"index":0,"delta":{"content":"..."}}]}

data: {"id":"...","object":"chat.completion.chunk","model":"offmem-glm-5.3-flash-abliterated","choices":[],"usage":{"prompt_tokens":18,"completion_tokens":32,"total_tokens":50}}

data: [DONE]

If the connection is closed before the end, the request is billed by estimate: the input plus what was already delivered.

Duration