Quickstart

Set up billing, sign in with the CLI, and make a first robot call and a first LLM call.

1. Set up billing

   Open abstract.inc/start?setup=billing and sign in. Add a payment method and set your monthly spend limit. Abstract creates no API key and runs no model until the account has a payment method or credit.

2. Install the CLI

   The abstract command signs you in and saves an API key on your machine.

   curl -fsSL https://abstract.inc/install.sh | bash
   

   The installer supports macOS and Linux on arm64 and x64.

3. Sign in

   abstract login
   

   The command prints a link and a code. Open the link in a browser and confirm the code. The command then saves an API key and exits. Check the key against the API:

   abstract status
   
4. Make a first robot call

   This call sends one observation to allenai/molmoact2:yam and prints the action chunk it returns. Save one JPEG from each camera as top.jpg, left.jpg and right.jpg. Replace robot_state with your robot’s state when you took the frames.

   first_robot_call.py

   import base64
   import json
   import os
   
   import requests
   
   API = "https://api.abstract.inc/v1"
   MODEL = "allenai/molmoact2:yam"
   CAMERAS = ["top", "left", "right"]  # the order this model expects
   
   
   def image_part(path):
       with open(path, "rb") as file:
           data = base64.b64encode(file.read()).decode()
       return {"type": "image_url", "image_url": {"url": f"data:image/jpeg;base64,{data}"}}
   
   
   body = {
       "model": MODEL,
       "messages": [
           {
               "role": "user",
               "content": [
                   {"type": "text", "text": "put the red cup in the bin"},
                   *[image_part(f"{camera}.jpg") for camera in CAMERAS],
               ],
           }
       ],
       "robot_state": {
           # Left joints 1-6, then right joints 1-6, in radians.
           "joint_position": [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0],
           # Left gripper, then right gripper: 0 is closed, 1 is open.
           "gripper_position": [1.0, 1.0],
       },
   }
   
   response = requests.post(
       f"{API}/chat/completions",
       headers={"Authorization": f"Bearer {os.environ['ABSTRACT_API_KEY']}"},
       json=body,
       timeout=300,
   )
   if not response.ok:
       raise SystemExit(f"HTTP {response.status_code}: {response.text}")
   answer = response.json()
   chunk = json.loads(answer["choices"][0]["message"]["content"])
   print(f"{len(chunk['actions'])} actions of {chunk['action_dim']} values, dt {chunk['dt']:.4f} s")
   print("first action:", chunk["actions"][0])
   print("usage:", answer["usage"])
   

   Run it with the key in the program’s environment only:

   pip install requests
   ABSTRACT_API_KEY="$(abstract token)" python first_robot_call.py
   

   The answer holds 30 actions of 14 values each, one every 1/30 s. Robot models lists the shapes of every model.

5. Make a first LLM call

   The LLM models speak the OpenAI Responses API at POST /v1/responses. A :superfast model answers inline:

   cURL

   curl -s https://api.abstract.inc/v1/responses \
     -H "Authorization: Bearer $(abstract token)" \
     -H "Content-Type: application/json" \
     -H "Idempotency-Key: $(uuidgen)" \
     -d '{"model":"glm-5.3:superfast","input":"Summarise RFC 8628 in three sentences."}'
   

   Run a Python or TypeScript program the same way, with ABSTRACT_API_KEY="$(abstract token)" in its environment. An :async model answers at once with a queued response that you poll. Responses API shows both.

Next steps

Authentication and keys

Keep the key out of logs and handle 401.

Chat completions

Run a whole episode over HTTP.

Sessions over WebSocket

Run an episode on one socket.

Errors and limits

What each error code asks you to do.