AingDesk × AI Code With
Use AI Code With in AingDesk: Third-Party API / Relay Setup
This page shows how to connect a third-party API in AingDesk, an open-source desktop assistant, so you can use models and knowledge bases locally with a custom supplier.
Download and install AingDesk first, then add a custom supplier in the app. Enter the key only in the client, never on this page.
- 供应商名称
- AI Code With
- 接口地址
- https://api.aicodewith.ai/v1
- API Key
- sk-xxxxxxxx
- Model ID
- claude-sonnet-4-6
Quick setup
In AingDesk, add a model supplier: fill 「供应商名称」, 「接口地址」, and 「密钥」, then pick a model. About 3 minutes.
AingDesk is a desktop client. Open the welcome page or the sidebar 「第三方模型API」, add a model supplier, fill 「供应商名称」, 「接口地址」, and 「密钥」, then sync models and send a test message.
Choose 「使用第三方模型」
The first time you open AingDesk, a welcome page appears. Click 「使用第三方模型」 in the red box; you should land on 「第三方API配置」. If the welcome page is already closed, use the sidebar 「第三方模型API」 instead. If you cannot get in, see #errors. Next: click 「添加模型服务商」.
On first launch of AingDesk, click 「使用第三方模型」 on the welcome page. The sidebar 「第三方模型API」 is another way in. Click 「添加模型服务商」
At the bottom of the 「第三方API配置」 window, click 「添加模型服务商」 (do not click built-in tiles such as 硅基流动). The 「添加模型服务商」 form should open. If you cannot see the button, close the window and reopen 第三方 API. If that still fails, see #errors. Next: fill the three fields and click 「确认」.
At the bottom of 「第三方API配置」, click 「添加模型服务商」. The built-in supplier on the right (硅基流动) is not this page's path. Fill 「供应商名称」, 「接口地址」, and 「密钥」
In the 「添加模型服务商」 dialog, fill 「供应商名称」 (AI Code With is fine), 「接口地址」 https://api.aicodewith.ai/v1 (helper text: 需要兼容openAI格式的接口), and 「密钥」, then click 「确认」. After success you should land on the saved supplier and it should start fetching /models. If 「确认」 hangs or the list is empty, see #errors. Next: check the model list or manually 「添加模型」.
Fill 供应商名称, 接口地址 https://api.aicodewith.ai/v1, and 密钥, then click 「确认」. Paste the AI Code With key into the 密钥 field. Sync models or 「添加模型」
After save, the list auto-fills from 「默认从/models获取所有模型」. IDs in the screenshot such as chatgpt-4o-latest are a relay catalog, not this page's Model ID. Turn on the models you need; if this page's ID is missing, click 「添加模型」, paste the Model ID, set 模型别名 yourself, and set 模型功能 to LLM. If the list is empty or you cannot find the ID, see #errors. Next: return to chat and send a test message.
Copy a Model ID
After save, AingDesk fills the list via 「默认从/models获取所有模型」. Copy a Model ID below; if it is missing, click 「添加模型」 and paste it, with 模型功能 set to LLM.
Pick a family, then copy a Model ID. 接口地址 stays the same for every model.
Selected
- Provider
- Claude
- Model ID
- claude-sonnet-4-6
The full catalog follows the AI Code With docs and pricing page. Names like chatgpt-4o-latest in screenshots only show auto-sync; use an ID from this page.
View the full model listIDs in the list such as chatgpt-4o-latest are a relay catalog, not Model IDs from this page. When you enable a model or click 「添加模型」, paste an ID copied here. Use 「请选择模型」 and send a test
Return to chat, click 「请选择模型」 in the top bar, pick the model you just added, and send a test message. The screenshot stops with no model selected and no reply yet. Success = a real reply, then match time and model in AI Code With usage records. Do not treat 「检查」 as connected. If there is no reply, see #errors.
Return to chat. The top bar still shows 「请选择模型」 (the screenshot has not sent a message). Click 「请选择模型」, pick the model you just added, and send a test. Success = a real reply + a usage record. Do not treat 「检查」 as success.
How does AingDesk connect to AI Code With?
AingDesk talks to an OpenAI-compatible endpoint through a custom supplier. Connecting to AI Code With means adding a model supplier, filling 接口地址 and 密钥, then picking a model:
- Choose 「使用第三方模型」
- First launch shows a welcome page. Click 「使用第三方模型」 to open 「第三方API配置」. If the welcome page is gone, use sidebar 「第三方模型API」.
- Click 「添加模型服务商」
- At the bottom of 「第三方API配置」, click 「添加模型服务商」 (not the built-in 硅基流动 tiles). The 「添加模型服务商」 form should open.
- Fill 「供应商名称」, 「接口地址」, and 「密钥」
- 供应商名称 can be AI Code With. Set 接口地址 to https://api.aicodewith.ai/v1 (helper text: 需要兼容openAI格式的接口). Paste the AI Code With key into 密钥, then click 「确认」. Do not append /chat/completions.
- Sync models or 「添加模型」
- After save, the list fills via 「默认从/models获取所有模型」. Names like chatgpt-4o-latest are a relay catalog, not Model IDs from this page. If this page's ID is missing, click 「添加模型」, paste it, and set 模型功能 to LLM.
- Use 「请选择模型」 and send a test
- Return to chat, click 「请选择模型」, pick the model you just added, and send a test message. Success = a real reply, then match usage records. Do not treat 「检查」 as connected.
Flow: AingDesk → custom supplier (供应商名称 + 接口地址 /v1 + 密钥) → AI Code With API. Everything is done in the GUI. No files to edit.
Frequently asked questions
New to AingDesk?
AingDesk is an open-source desktop AI assistant for Windows and macOS.
AingDesk download page
Download the installer for your OS from the official download page. After install, open AingDesk and start third-party model setup from the welcome page or the sidebar.
Open the AingDesk download pageIntel Mac is no longer supported; Docker :7071 is the server, not this page's default path. Full install guidance follows the AingDesk official site; this site does not host the installer.
Common error troubleshooting
The most frequent problems fall into 4 categories.
Empty model list / Confirm hangs
Symptom: 「确认」 hangs, or after save 「默认从/models获取所有模型」 is still empty.
Cause: AingDesk GETs {baseUrl}/models with the 接口地址 as-is and does not auto-append /v1. Missing /v1, appending /chat/completions, a bad key, or an empty upstream catalog all yield an empty list. Confirm can hang while fetching the full catalog (#105).
- Set 接口地址 to https://api.aicodewith.ai/v1 only — do not add /chat/completions
- If the list is empty, click 「添加模型」 and paste a Model ID from this page
- Set 模型功能 to LLM
Chat picker missing the new model / only Ollama
Symptom: After adding the supplier, 「请选择模型」 does not list it, or only Ollama appears.
Cause: The supplier switch is off; the list does not refresh after adding a model (#50); imported models with title==="" are filtered; other suppliers flood the picker (#92).
- Enable this supplier and disable the others
- Close and reopen 「第三方模型API」 so the list refreshes
- If needed, fill 模型ID and 模型别名 by hand
Send fails even though 「检查」 says success
Symptom: On the saved-supplier panel, 「检查」 says the config is correct, but sending returns 401, no reply, or an unexpected interrupt.
Cause: 「检查」 and 「保存API」 only appear on an already-saved supplier panel. 「检查」 is a stub: it does not hit the network and always returns success. Real verification is a chat reply plus AI Code With usage records.
- Re-paste the AI Code With key; never treat 检查成功 as 接入成功
- Confirm 接口地址 includes /v1
- Send a test message, then match the call by time and model in usage records
Synced list is a relay catalog / model not found
Symptom: After save you see a long list, but this page's Model ID is missing, or sending says the model does not exist.
Cause: /models returns the upstream catalog, not this key's IDs. Toggles do not remove models from chat (#92); you must delete them. Shot 4 illustrates this: names like chatgpt-4o-latest are a relay catalog.
- Click 「添加模型」 and paste a Model ID from this page
- Do not append /chat/completions to 接口地址
- Confirm the key can access that model
Need the full guide?
Every config parameter, the complete model list, and more troubleshooting live in the AI Code With docs.
Open the AI Code With docsSources
- AingDesk official website
- AingDesk download page
- AingDesk custom third-party API docs
- AingDesk GitHub
- AI Code With docs