Full reference

Fusion

Ask a panel of models to produce candidate answers, then use a judge to synthesize the response. Measure whether that extra work helps your task.

Updated

Reviewed against Console v0.2.2. Available actions depend on project access and runtime configuration; check your installed CLI version for command differences.

Choose a Fusion route

The gateway exposes stimulir/fusion and stimulir/fusion-max as virtual model IDs. Their configured panel and judge can change; inspect the current route instead of treating an old provider list as a guarantee. Fusion still uses the chat-completions endpoint.

Try a bounded question

bash
stimulir fusion "Explain the trade-offs in this example" --project <project-id>

The CLI streams by default. --no-stream waits for the complete answer. --max selects the max route. A panel can involve several billed model calls, so inspect cost as well as the returned text.

Override panel and judge

json
{
  "model": "stimulir/fusion",
  "messages": [{"role": "user", "content": "Example task"}],
  "tags": ["fusion-comparison"],
  "fusion": {
    "panel": ["<enabled-panel-model-1>", "<enabled-panel-model-2>"],
    "judge": "<enabled-judge-model>"
  }
}

Choose model identifiers enabled for the workspace. Set the panel and judge explicitly when you need a reproducible comparison.

Respect each route’s limits

The requested content must be compatible with the panel and judge. Do not assume that a large context window or vision support on one member applies to every member. Handle rejected requests and partial failures according to the returned error.

Evaluate the result

Inspect the response and any available route or usage evidence. A fused answer is still a candidate answer, not a proof of correctness. Compare it with a simpler baseline on the same task cases, including latency and cost.

Use the SDK

client.fusion.complete() accepts panel and judge overrides and requires a tag through the call or client. See the method parameters below and the Python SDK.

Command catalogue

Expand a command for its arguments. These commands follow the released CLI source; check stimulir --version and the command’s --help for your installation. Required flags are marked; some operations also validate combinations of inputs.

stimulir fusion

Ask a panel of models and stream the judge-fused answer.

prompt required
Positional argument. The question to answer with the fusion panel.
--panel
Comma-separated panel model ids (overrides the default panel).
--judge
Judge model id (overrides the default judge).
--max
Use stimulir/fusion-max (claude-sonnet-4-6 + claude-opus-4-6, judged by claude-opus-4-6).
--staging
Use the Stimulir staging API base.
--project
Project id to send as X-Project-Id.
--no-stream
Wait for the full answer instead of streaming (SSE).

SDK method catalogue

Expand a client to inspect public methods and parameter names. Read each operation’s task guide before supplying a payload or starting work. The SDK returns platform responses; handle errors and inspect completion state before using a result.

client.fusionFusionClient
complete(prompt, panel, judge, max_tier, temperature, max_tokens, tag)
Run one fusion chat completion and return the completion payload.