Getting started with Aquin CLI

Aquin is a CLI for mechanistic interpretability and model analysis on your own GPU. You install aquin on a CUDA machine, connect a session with a model, and run inspections, simulations, evals, or watch external training runs from the terminal. Every result mirrors live to the web dashboard at aquin.app: same session, same cards, no re-upload.

The CLI is not a hosted GPU service. Your machine runs forward passes locally. The cloud only syncs state: messages, tool results, and session memory stream over SSE to the orchestrator panel. Use aquin chat for a multi-turn agent, or invoke any tool directly as a one-shot command.

Tool mode switches automatically when you start a session with a model. Start with an LLM (GPT-2, Pythia, Llama) and you get inspection, eval, and simulation commands. Start with an embedding encoder (GTE, MiniLM, BGE) and you get the embed analysis suite. Beyond catalog slugs, pass a HuggingFace repo id for supported families — see Supported models. No manual mode toggle; aquin help always shows what is available for your current session. Use aquin status anytime to see your account, API key, and active session (device, model, mode). Generate your CLI token under Profile → Account; view connected engines and usage under Profile → CLI.

Feature-level tools need a sparse autoencoder pulled locally first (aquin load sae gpt2-small-l8). The sidebar groups every command by category (inspection, SAE inspection, simulation, SAE training, training watch, evals), split by LLM and embedding where applicable. Cross-cutting setup lives under Agentic & Basic CLI.

Web profile

Open your avatar on aquin.app to manage CLI auth and activity. Account is for identity and your CLI token — generate or regenerate once, then paste it with aquin login. Each account has one active token; regenerating revokes the previous one. Teammates on a shared GPU use their own tokens and aquin switch.

CLI shows engines online from aquin session start and 14-day usage: new sessions, tool syncs to the web, captures, and all-time totals.

First session

Simulation run management

After aquin simulate, runs are saved locally. Use aquin list simulation to list IDs,aquin load simulation --run_id … to reopen one on the web, and aquin compare simulation --run_id_a … --run_id_b … to diff two forecasts. Legacy aliases list-runs, load-run, and compare-runs still work.

What to read next

• Supported models: LLMs and embedding models Aquin supports today
• Agentic & Basic CLI: login, status, session start (with model), list, switch, delete, load sae, chat, list/load/compare simulation, run-code, mem, configuration, shared flags
• Inspection (LLM): attention, perturbation, weight health
• Inspection SAE (LLM): inspect, steer, feature analysis
• Simulation (LLM): training forecast, then list / load / compare simulation runs
• SAE training: capture activations and train temp SAEs
• Deception features: find-feature on honest vs deceptive probes
• Checkpoint SAE: diff and align on fine-tuned checkpoints
• Training watch: ingest external training metrics JSONL and mirror live charts