OctOpus in your terminal and code.
The same AI data scientist as the app, from your terminal, your scripts and your apps. Same account, same projects, same credits. OctOpus does the work on our servers; all you need is Python.
One command
Paste into your terminal. If OctOpus is open, type /exit first.
python -m pip install --upgrade https://www.octoopus.dev/developers/sdk/octopus_ml-0.5.6-py3-none-any.whl
Need help with Python?
Python 3.10 or newer is recommended. On Mac/Linux, use python3 if python is unavailable. If pip asks for a virtual environment, create one with python3 -m venv .venv, activate it with source .venv/bin/activate, then run the install command.
Open OctOpus
Run this. Your browser opens to connect your account; next time the terminal remembers you.
octopus
Ask in plain words
Mention a file with @ and say what you want.
Analyze @customers.csv
Type /help for commands, /exit to leave.
What you can do from the terminal
❯ what drives churn in @customers.csv? ✓ customers.csv · 7,043 rows · 21 columns Contract type is the strongest driver … ❯ train a model to predict it, optimise AUC — go ★ champion AUC 0.861 · hold-out 0.858 ❯ build a dashboard of revenue by region ❯ what does our refund procedure say? answered from Descartes, your company brain ❯ code my last call and send it to the CRM ❯ /calls follow_up
Shortcuts
/files /load 3pick a local file
/data corpus.zipdata, or a speech corpus
/plan /train /goplan and train
/watch /statusfollow the live run
/dashboard /viewbuild it, read it here
/calls /recordingsCRM tickets, recorded calls
/proofthe signed certificate
/calls and /recordings list them. Full CLI guide →More ways in
Use OctOpus in ChatGPT
Connect your OctOpus account to analyze datasets and manage training from a ChatGPT conversation. Your subscription and credits apply.
Advanced: API tokens and Python scripts
Get your token
Create a token and copy it. It grants your account permissions, expires after 30 days, and uses your account credits for paid operations.
Sign in to OctOpus, then return here to manage your tokens.
Copy this token now. It is shown only once.
Starter script
Download the starter and run python octopus-start.py. Sign in once with octopus login first. The script uses that saved account and prints “Connected!” without starting a paid AI run.
The example uses your saved terminal sign-in. No secret is included in the file.
Next: analyze a CSV
After connecting, add these lines to the starter. Analysis uses your account credits.
session.attach("customers.csv")
result = session.run("Summarize this dataset").wait()
print(result["result"]["response"])Advanced: API reference and execution limits
The API contract
POST /api/v1/sessions
GET /api/v1/sessions/{session_id}
POST /api/v1/sessions/{session_id}/datasets
POST /api/v1/sessions/{session_id}/messages
GET /api/v1/sessions/{session_id}/jobs/{job_id}
Authenticate with Authorization: Bearer YOUR_TOKEN. Message requests require X-Octopus-Idempotency-Key; reuse the same key and body when retrying a disconnected request. Each session accepts one active conversation turn at a time.
CSV attachments support up to 5 million characters. Job status is available through polling. Python job.events() yields changed status snapshots, not a live token stream.
Execution and approvals
Chat can return proposed training or dashboard actions. A completed chat job means its response is saved; it does not mean a proposed model or dashboard was executed. Your application must review the response and explicitly invoke the corresponding operation. Existing SDK client.run(..., wait=False) starts model training and returns a run handle.
Completed results are durable. This beta does not automatically resume unfinished computation after a worker restart, and does not provide a durable task queue or replayable tool-event log.