Workflow Guides

Route 1: Keep using your subscription

1. App Desktop: Detect or install the client you normally use.
2. Accounts: Add your official account and check its identity and authorization.
3. App Desktop: Select the client and account, then apply the configuration and launch.
4. The actual client: Check the identity and make a request. Return to Accounts to refresh quota when needed.

Result: Manage account status centrally while using familiar tools. Adding an account and switching a client remain clearly separate steps.

Route 2: Reuse a model across tools

1. Models: Save the endpoint, key, and model ID, then click Test Latency to check connectivity and latency.
2. App Desktop: Select this model for a supported tool and apply its configuration.
3. A second tool: Reuse the same model entry, applying it with a protocol that tool supports.
4. Verify: Make a request in each tool to confirm the configuration has taken effect beyond simply being saved in EchoBird.

Result: Less repeated data entry and a clear record of your model sources. Protocol and feature support still depend on each tool and model.

Route 3: Primary models and fallback sources

1. Verify your primary and fallback APIs in Models. You can also prepare supported account sources.
2. Create a group in Smart Routes, configure the actual endpoint and key, and arrange sources by priority.
3. Connect your coding client to that group and make a real request.
4. Review completion and failover results in Statistics, and add your usual usage sources to Usage Island.

Result: Keep the client's connection relatively stable while adjusting sources centrally. When quotas or failures change, status information helps you decide what to do next instead of blindly repeating requests.

Route 4: From an error back to work

1. Check installation status in App Desktop and keep the exact error if a tool cannot launch.
2. Open AI Assistant and describe the tool, operating system, error, and desired result.
3. Watch the environment checks and tool output, handling permissions through system prompts when needed.
4. Return to App Desktop, refresh detection, and launch the tool to confirm the issue is resolved.

Result: Connect installation troubleshooting with a real acceptance check. The assistant's summary explains the process; a successful launch confirms the outcome.

Route 5: Speak your request and stay focused

Configure a channel in Voice Input and test the microphone, then place the cursor in AI Assistant or your usual tool. Dictate the task, check the recognized text, and send it. Monitor status with Usage Island as you work. Tokens recorded through Smart Routes can also contribute to pet growth on the Pets page.

Route 6: Your hardware, your model service

Prepare model files and an engine in Local LLM, start the service, and then let supported model entries and AI Assistant use it. If something fails, distinguish a service that has not started, an unavailable model, and a failed task. This lets you verify “running the model” separately from “having the model do something.”