Release Models

JD.Writer is intentionally shipped in multiple shapes so teams can choose the right tradeoff for development, demos, and production.

Which Model Should You Use?

  • Use client-server when you want full AI routing and API-backed behavior.
  • Use client-only when you want offline-friendly local editing with no backend dependency.
  • Use docker when you want reproducible environment packaging.
  • Use github-pages when you want a zero-backend browser demo.

Models at a Glance

Model Components Best For
client-server JD.Writer.Web + JD.Writer.ApiService (+ optional JD.Writer.AppHost) Daily local development and full behavior validation
client-only JD.Writer.Web only (AiClient:Mode=local) Standalone local drafting with no API
docker Multi-target images + compose profiles Reproducible deployment and containerized testing
github-pages Static docs/studio app + DocFX Shareable in-browser demo with no server

Runtime Configuration

JD.Writer.Web supports:

  • AiClient:Mode
    • auto: try API first, then local fallback
    • remote: prefer API, fallback locally on failures
    • local: never call API; local heuristics only
  • ApiServiceBaseUrl: explicit URL for web-to-API calls

Examples:

# full client/server mode
export AiClient__Mode=remote
export ApiServiceBaseUrl=http://localhost:5081

# standalone client-only mode
export ASPNETCORE_ENVIRONMENT=ClientOnly
export AiClient__Mode=local

Local Packaging Scripts

PowerShell:

./scripts/release/publish-release-models.ps1

Bash:

./scripts/release/publish-release-models.sh

Output is created under artifacts/release/:

  • client-server/web
  • client-server/api
  • client-server/apphost
  • client-only/web
  • docker/{Dockerfile,compose.yaml,.dockerignore}

Docker Usage

Client/server profile:

docker compose --profile client-server up --build
  • Web: http://localhost:18080
  • API: http://localhost:18081

Client-only profile:

docker compose --profile client-only up --build web-standalone
  • Standalone Web: http://localhost:18082

If Ollama runs on host, default API endpoint is http://host.docker.internal:11434. Override with AI_OLLAMA_ENDPOINT when needed.

GitHub Release Outputs

.github/workflows/release.yml publishes:

  • client/server tar archives
  • client-only standalone tar archive
  • Docker bundle tar archive
  • GHCR images:
    • ghcr.io/<owner>/<repo>/jd-writer-api:<version>
    • ghcr.io/<owner>/<repo>/jd-writer-web:<version>
    • ghcr.io/<owner>/<repo>/jd-writer-web-standalone:<version>

.github/workflows/docker.yml validates Docker targets on pull requests.

GitHub Pages Studio Lite

DocFX publishes Studio Lite from docs/studio/index.html.

  • Repo path: docs/studio/
  • Published path: https://jerrettdavis.github.io/JD.Writer/studio/
  • Storage: browser localStorage (jdwriter.pages.studio.v1)
  • AI behavior: deterministic local heuristics only
Edit this page