Custom frontend
A custom frontend is an independently deployed web app that logs users in via the central account site and then operates on a specific workspace on a specific d6e instance through the public HTTP API — the same API the console uses.
There is no privileged integration. The security boundary stays on the instance.
Install skills (recommended: everything)
Section titled “Install skills (recommended: everything)”Installing frontend skills alone is rarely enough. Dependent tables, prompts, workflows, and STFs usually live on the Plugin / Docker STF side — so install all three repositories. The same applies when asking an agent to design or answer questions.
npx skills add d6e-ai/d6e-plugin-skills --skill '*' -ynpx skills add d6e-ai/d6e-docker-stf-skills --skill '*' -ynpx skills add d6e-ai/d6e-custom-frontend-skills --skill '*' -ySee Installing Agent Skills for details.
Frontend-only minimal set (usually not recommended):
npx skills add d6e-ai/d6e-custom-frontend-skills --skill '*' -yRepository: d6e-ai/d6e-custom-frontend-skills
A reference implementation (AI bookkeeping) ships in the same repo.
What the three frontend skills cover
Section titled “What the three frontend skills cover”| Skill | Responsibility |
|---|---|
d6e-auth-integration |
OAuth2 (authorization code), session cookies, refresh, workspace allow-list |
d6e-workspace-api-client |
Server-side proxy (files / SQL / workflows / async jobs, …) |
d6e-prompt-driven-ui |
LLM kind JSON output, Zod parsing, revision flow, UI contracts |
Auth essentials
Section titled “Auth essentials”- Redirect the browser to
https://www.d6e.ai/auth/login - POST the authorization code to the instance
POST /api/v1/auth/token - Call
/api/v1/*with the Bearer token afterwards
Frontends never hold a client secret. Register production callback URLs in d6e-auth (localhost is exempt). See Architecture.
Conceptual docs (canonical)
Section titled “Conceptual docs (canonical)”- frontend-and-instance.md — the three actors
- architecture.md — reference app sequences
- d6e-api-integration.md — request/response shapes
- workspace-setup.md — preparing the dependent workspace
Combining with Plugin / Docker STF
Section titled “Combining with Plugin / Docker STF”Package the tables, prompts, workflows, and STFs your frontend depends on as a Plugin so new workspaces can be provisioned reproducibly. Include Docker STFs in the Plugin when you need heavy compute or external network access. See Choosing a path.