API reference — Failure Model Labs

Overview

Method	Path	Auth	Purpose
POST	/v1/chat/completions	Key	Generate a response
GET	/v1/models	Key	List model ids
GET	/v1/usage	Key	This key's month spend, cap and remaining
GET	/v1/pricing	Public	Current prices per model
GET	/v1/status	Public	Live flag and lane states

Base URL https://api.failure.fail/v1 (also https://failure.fail/v1). Auth with Authorization: Bearer <key> or x-api-key: <key>. Bodies are JSON.

POST /v1/chat/completions

OpenAI Chat Completions compatible. Standard Chat Completions fields are accepted; the ones that matter most:

Field	Type	Notes
model	string, required	`failure-tempest-1-flash`, `failure-tempest-1` or `failure-tempest-1-max`
messages	array, required	System, user and assistant messages, as in the OpenAI API
stream	boolean	Server-sent events; working as `reasoning_content` deltas
stream_options	object	Accepted; the final usage chunk is sent regardless
temperature	number	`0` makes exact repeats eligible for the response cache
max_tokens	integer	Ceiling on the answer length
tools, tool_choice	array, object	Tool-call requests go to a single tool-capable model, not fused

Request headers

Header	Notes
x-tempest-cache: on \| off	`on` caches the response at any temperature; `off` bypasses the cache

Example

curl https://api.failure.fail/v1/chat/completions \
  -H "Authorization: Bearer $FAILURE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "failure-tempest-1",
    "messages": [
      {"role": "system", "content": "You are a careful assistant."},
      {"role": "user", "content": "Is 2^61 - 1 prime? Show the check."}
    ]
  }'

Response

{
  "id": "chatcmpl-...",
  "object": "chat.completion",
  "created": 1793750400,
  "model": "failure-tempest-1",
  "choices": [
    {
      "index": 0,
      "message": { "role": "assistant", "content": "Yes. 2^61 - 1 is a Mersenne prime ..." },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 24,
    "completion_tokens": 310,
    "prompt_tokens_details": { "cached_tokens": 0 },
    "completion_tokens_details": { "reasoning_tokens": 240 },
    "cost": 0.00942
  },
  "tempest": {
    "task": "...",
    "difficulty": "...",
    "stage": "...",
    "cache": "...",
    "billed_usd": 0.00942
  }
}

Illustrative values. usage.cost and tempest.billed_usd are in USD.

The tempest object

Field	Meaning
task	What the brief decided the prompt is asking for
difficulty	The difficulty the router and brief settled on
stage	The furthest pipeline stage the request reached
cache	Whether the response cache served this answer
billed_usd	What this request cost you

Streaming chunks follow chat.completion.chunk; see streaming.

GET /v1/models

curl https://api.failure.fail/v1/models -H "Authorization: Bearer $FAILURE_API_KEY"

{
  "object": "list",
  "data": [
    { "id": "failure-tempest-1-flash", "object": "model" },
    { "id": "failure-tempest-1", "object": "model" },
    { "id": "failure-tempest-1-max", "object": "model" }
  ]
}

GET /v1/usage

The calling key's spend this month, its cap, what remains, and a per-model breakdown.

curl https://api.failure.fail/v1/usage -H "Authorization: Bearer $FAILURE_API_KEY"

Illustrative shape:

{
  "month": "2026-11",
  "spend_usd": 12.40,
  "cap_usd": 50,
  "remaining_usd": 37.60,
  "models": { "failure-tempest-1": { "spend_usd": 12.40 } }
}

GET /v1/pricing

Public. Prices per 1M tokens for each model, plus billing notes.

curl https://api.failure.fail/v1/pricing

{
  "models": {
    "failure-tempest-1-flash": { "input": 3, "cached_input": 0.3, "output": 15 },
    "failure-tempest-1": { "input": 5, "cached_input": 0.5, "output": 30 },
    "failure-tempest-1-max": { "input": 10, "cached_input": 1, "output": 60 }
  },
  "notes": ["..."]
}

Launch draft prices; they may change. Read them from this endpoint rather than hardcoding.

GET /v1/status

Public. Whether the API is live and the state of each lane in the panel.

curl https://api.failure.fail/v1/status