Continue.dev + Gonka AI - AI for VS Code/JetBrains

Continue uses ~/.continue/config.yaml (the old config.json format is deprecated). Open it via the command palette: Ctrl+Shift+P (on macOS, Cmd+Shift+P) → "Continue: Open config.yaml".

Add JoinGonka Gateway as a provider:

# ~/.continue/config.yaml
models:
  - name: Gonka DeepSeek V4 Flash
    provider: openai
    model: deepseek-ai/DeepSeek-V4-Flash-0731
    apiBase: https://gate.joingonka.ai/v1
    apiKey: jg-your-key
    roles: [chat, edit, apply]
    capabilities: [tool_use]
  - name: Gonka Kimi K2.6
    provider: openai
    model: moonshotai/Kimi-K2.6
    apiBase: https://gate.joingonka.ai/v1
    apiKey: jg-your-key
    roles: [chat, edit, apply]
    capabilities: [tool_use]
  - name: Gonka MiniMax M2.7
    provider: openai
    model: MiniMaxAI/MiniMax-M2.7
    apiBase: https://gate.joingonka.ai/v1
    apiKey: jg-your-key
    roles: [chat, edit, apply]
    capabilities: [tool_use]
  - name: Gonka Autocomplete
    provider: openai
    model: deepseek-ai/DeepSeek-V4-Flash-0731
    apiBase: https://gate.joingonka.ai/v1
    apiKey: jg-your-key
    roles: [autocomplete]

Continue displays models with chat, edit, apply roles in a list—you can switch directly in the chat: Gonka DeepSeek V4 Flash handles a 380K context and up to 32768 output tokens (good for large files and long edits), Kimi K2.6 is stronger at reasoning, while MiniMax M2.7 is faster for short responses. DeepSeek V4 Flash is set for autocomplete, and this is not just preference: Continue selects the prompt format based on the model NAME, and among the network models, only DeepSeek understands it. MiniMax and Kimi will respond to the same formatting with reasoning instead of code, resulting in garbage suggestions.

Autocomplete via a decentralized network is slower than Copilot: about two seconds per suggestion versus a fraction of a second. Chat, edits, and agent mode are not affected by this.

The line capabilities: [tool_use] is mandatory for agent mode. Continue only enables tools for models in its own list (gpt-*, o*, codex*)—Gonka models are not there, and without this line, the agent remains tool-less: it cannot read files or edit code, and the chat cannot see the project. The gateway itself supports tool-calling.

Project indexing (@codebase) requires an embedding model—there are currently none in the Gonka network. Add context via @file and @folder or work in agent mode, where it reads files using tools.

To avoid keeping the key in plain text, Continue supports apiKey: ${{ secrets.GONKA_API_KEY }} substitution with a variable in ~/.continue/.env.

Verification: open any code file and start typing—autocomplete should work. Or open the Continue chat (Ctrl+L) and ask a question.

Important: ensure apiBase ends with /v1 at the end of the URL. Without it, requests will return 404.