OpenCode Setup for Multi-Model Streaming Sessions

§1 Build a Streaming Workspace with Two Models

OpenCode lets you work with AI models from a terminal-based coding interface. Connecting it to an API relay can simplify multi-model workflows: you configure one provider, register the models you want, and switch between them as your task changes.

Fluxio fits this setup because it is a streaming-first, pay-per-token AI API relay. One key provides access to multiple model families, with responses delivered token by token. This walkthrough uses Claude Sonnet 5.5 and Gemini 3.1 Pro as example display names. Confirm their availability and exact API identifiers in your Fluxio account before configuring them.

§2 Step 1: Install OpenCode

Install OpenCode using a supported installation method. If Node.js and npm are available, run npm install -g opencode-ai. Then run opencode --version to confirm the command is accessible.

Open a terminal in a small project directory for your first test. Choose a repository without sensitive files, and review permission prompts before allowing the assistant to run commands or modify code.

§3 Step 2: Prepare Your Fluxio Credentials

Create an API key in your Fluxio account and check that your account has usable credit. Keep the key outside your repository rather than embedding it in configuration files.

On macOS or Linux, set it for the current shell with export FLUXIO_API_KEY='your-key'. In PowerShell, use $env:FLUXIO_API_KEY='your-key'. Start OpenCode from that same terminal so it inherits the variable.

Fluxio’s API base is https://api.fluxioapi.com. For the OpenAI-compatible connection below, use https://api.fluxioapi.com/v1. Avoid appending a second version suffix.

§4 Step 3: Add an OpenAI-Compatible Provider

Create an opencode.json file in your project directory, or merge the following provider into your existing configuration. The placeholder model keys must be replaced before use.

{ "provider": { "fluxio": { "npm": "@ai-sdk/openai-compatible", "name": "Fluxio", "options": { "baseURL": "https://api.fluxioapi.com/v1", "apiKey": "{env:FLUXIO_API_KEY}" }, "models": { "MODEL_ID_A": { "name": "Claude Sonnet 5.5" }, "MODEL_ID_B": { "name": "Gemini 3.1 Pro" } } } } }

Replace MODEL_ID_A and MODEL_ID_B with the exact model identifiers provided by Fluxio. Display names are labels for humans; they do not establish valid API identifiers. Do not guess identifiers from product names.

This configuration uses OpenCode’s OpenAI-compatible provider package. Basic streaming compatibility does not guarantee support for every coding feature, so verify tool calling and any additional capabilities you need with your selected models.

§5 Step 4: Start and Verify Streaming

Run opencode, enter /models, and select your first model under Fluxio. Send a small request such as, “Explain the purpose of this project without changing files.”

Look for incremental output rather than a single completed response. OpenCode manages streaming through its provider integration; you do not need to add a separate streaming flag to the configuration above.

• – Authentication errors: Check that the environment variable exists in the terminal launching OpenCode.
• – Model errors: Confirm the configured identifier and your account’s access to that model.
• – Delayed output: Check network conditions and intermediary buffering before assuming streaming is disabled.

§6 Step 5: Switch Models Within Your Workflow

Use /models again to choose the second model. For a practical comparison, ask each model to review the same small function, then compare correctness, latency, and suggested changes.

Switching models in an existing session is convenient, but it does not transfer hidden provider-side state. Conversation history, context limits, and supported message features can affect continuity. Use /new when you want an independent comparison, and supply the same instructions and files.

§7 Step 6: Check Reliability and Cost

Before using the setup for larger tasks, test a harmless file edit and a permission-approved command. If ordinary text streams correctly but tool actions fail, investigate tool compatibility separately from authentication or transport.

Monitor token usage in Fluxio, especially during long sessions. A shared relay simplifies credentials, but each model can still have different pricing, latency, and capability trade-offs.

Sign up for Fluxio to try your streaming setup.