drukscode/drukscode-v2
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
drukscode-v2/docs/guide/more-features/ai-settings.md
OpenCode AI e9899d2e6d
Some checks failed
Android CI Build / check (push) Has been cancelled
Details
Android CI Build / package-apks (push) Has been cancelled
Details
Docs Deploy / build (push) Has been cancelled
Details
Module Market Check / Validate modules/ (push) Has been cancelled
Details
Module Market Publish / Generate submissions.json (push) Has been cancelled
Details
Docs Deploy / deploy (push) Has been cancelled
Details
Initial DruksCode V2: fresh fork of web-to-app (f3b488e) with DruksCode identity 
- Package rename com.webtoapp -> br.com.drukstech.codeapp
- Product rename WebToApp/web-to-app -> DruksCode/drukscode
- Design system Wta* -> Dkc*
- applicationId br.com.drukstech.codeapp (.play for gplay flavor)
- rootProject.name DruksCode
- Fresh git history (upstream kept as remote for sync)
2026-09-12 02:42:02 +02:00

1.6 KiB
Raw Permalink Blame History

AI Settings

Configures the AI backend used by Agent. Open it from ⋮ → AI Settings.

Provider catalog

The built-in catalog keeps well-known providers only, at most three per category:

  • Recommended — Google Gemini, OpenRouter.
  • International — OpenAI, Anthropic, Grok.
  • Aggregator — Together, Perplexity, Fireworks.
  • Chinese — DeepSeek, Qwen, GLM.
  • Self-hosted — Ollama, LM Studio, vLLM.

Anything else fits Custom: any provider not in the catalog (or removed by a past update) is handled as a custom endpoint, and legacy configs referencing a removed provider migrate to Custom automatically with their base URL preserved.

API keys

  • Add provider API keys, each with an optional alias.
  • Custom endpoints choose an API format — Chat Completions (/chat/completions), Anthropic Messages (/v1/messages), OpenAI Responses (/responses), or Google Gemini — and the gateway routes the request by the declared format instead of assuming OpenAI-compat. The chat endpoint can be overridden per key.
  • Connection test — verify a key works (connection OK / fail).
  • Keys are stored securely on-device.

Saved models

  • Add models individually or batch-select from the available list.
  • Configure model capabilities and what each model is available for.
  • A pricing reference helps you pick models.

Advanced

  • Context capacity — the context window size used for Agent sessions.

Notes

Agent won't work until at least one valid API key and model are configured here.