IntelligenceX

Setup Web UI

Edit on GitHub

Run the local IntelligenceX setup web UI, review the flow, and understand the localhost-only onboarding experience.

Setup Web UI (Preview)

Run the local setup UI:

intelligencex setup web

This starts a local web server and opens the wizard in your browser (http://127.0.0.1 only).

Screenshots

Quick flow

1) Start the web UI
2) Choose onboarding path (new setup / fix auth / cleanup / maintenance)
3) Run auto-detect preflight
4) Authenticate with GitHub (device flow or app install)
5) Select repos
6) Plan + Apply

flowchart LR
  classDef step fill:#BAE6FD,stroke:#0369A1,color:#082F49,stroke-width:2px;
  classDef auth fill:#FDE68A,stroke:#B45309,color:#451A03,stroke-width:2px;
  classDef apply fill:#A7F3D0,stroke:#047857,color:#052E2B,stroke-width:2px;

  A["Path"] --> B["Auto-Detect"]
  B --> C["GitHub Auth"]
  C --> D["Repos"]
  D --> E["Configure/Auth"]
  E --> F["Plan"]
  F --> G["Apply + Verify"]

  class A,B,D,E,F step;
  class C auth;
  class G apply;

Path requirements (GitHub/repo/AI auth) and Bot contract checks are defined in Web Onboarding Flow.

Operations available:

  • Setup / update workflow + config
  • Update provider secret only
  • Cleanup (remove workflow/config)
  • Maintenance (inspect and choose operation)
  • Optional GitHub App manifest flow (create app + installation token)
  • Load existing config from a repo (manage existing setup)
  • Load workflow preview for the managed workflow
  • Save/load browser presets for provider, model, review knobs, and config overrides
  • Export/import presets as JSON files
  • Import prompts before overwriting existing presets

Advanced options:

  • Provider toggle (openai | claude | copilot)
  • Provider-aware model field in the Configure step, with quick picks for common OpenAI and Claude models
  • Review step summarizes why the selected provider/model pair is a good fit before apply
  • Browser presets now retain the named provider/model profile alongside the raw provider/model values
  • Static analysis controls when generating preset config (analysisEnabled, analysisGateEnabled, analysisRunStrict, packs, export path)
  • OpenAI account routing supports primary-only setup (rotation/failover can be configured without account ids)
  • OpenAI auth bundle input for INTELLIGENCEX_AUTH_B64
  • Claude API key / key-path input for ANTHROPIC_API_KEY

For YAML vs JSON ownership and precedence, see Workflow vs JSON.

GitHub App flow (optional)

If you want to avoid personal access tokens, you can use the GitHub App manifest flow:

  1. Enter App name + App owner (org login).
  2. Click “Create App (manifest)”. A browser window opens to create the app.
  3. Install the app in the org/user and return to the wizard.
  4. Click “List installations”, select the installation, then click “Use installation token”.
  5. The GitHub token field is populated with the installation token; proceed to load repos.

Current limitations

  • OpenAI login runs locally in the web UI ("Sign in with ChatGPT") and returns an auth bundle (authB64) you can upload as INTELLIGENCEX_AUTH_B64.
  • Claude setup in the web UI uses an API key or key file path; there is no browser-login equivalent.
  • Update-secret in the web UI requires provider-specific credentials:
    • OpenAI: authB64 or authB64Path
    • Claude: anthropicApiKey or anthropicApiKeyPath
  • The UI supports multi-repo setup (plan/apply), repo inspection, and setup recommendations.
  • GitHub App installation tokens can only list repos the app is installed on.
  • Buttons are disabled until required inputs are provided (token, repo selection, and any required provider secret input).
  • Inline hints describe what is missing before plan/apply can run.
  • Browser presets intentionally do not store GitHub tokens, OpenAI auth bundles, or Claude API keys.
  • Switching providers resets incompatible default models so stale GPT/Claude model values are not carried forward accidentally, while still letting you type a custom model id.

Tips

  • Use the "Load workflow preview" button before applying changes.
  • If you want zero secret handling in the UI, enable "Skip provider secret" and paste secrets manually in GitHub.
  • Save a browser preset after choosing your provider/model pair if you frequently onboard repos with the same setup shape.
  • Start with auto-detect to get a recommended path before selecting repositories.
  • If you automate setup with Bot tools, verify contractVersion + contractFingerprint match before apply.

Security notes

  • The server listens on 127.0.0.1 only over HTTP (no HTTPS binding).
  • GitHub tokens are sent to the local CLI server and never leave your machine.
  • Use this on trusted machines only.