@userreason/wizard 0.2.0 → 0.3.0

package/README.md

@@ -36,6 +36,22 @@ terminal. The [first survey guide](https://docs.userreason.com/get-started/first
 walks through creating and publishing a survey. Installing the SDK does not
 publish a survey.
 
+In development mode (`--dev`), an interactive installation continues into a
+guided first-survey flow. Choose an improvement, workflow-friction or satisfaction
+goal and immediately review a ready-to-edit question. You can save that starter
+without AI or skip the step. Optional AI personalization runs while you review
+the starter and never replaces your edits automatically.
+
+Local source preparation starts after sign-in and stays in memory. It does not
+upload code, read Git history or make a model call. Personalization requires a
+separate review and approval of the exact source packet. Changed files are
+rechecked before upload and before using a source-based suggestion.
+
+Saving creates an unpublished draft. Choose a verified completion event or
+configure audience and timing in the editor; the latter leaves the draft unable
+to match anyone. No survey is published automatically. This development flow
+does not change the production wizard or require an SDK release.
+
 For a different application directory:
 
 ```sh
@@ -137,3 +153,90 @@ locally and is not uploaded to an AI service.
 
 See [troubleshooting](https://docs.userreason.com/get-started/troubleshooting/)
 for help with installation, sign-in, permissions and browser verification.
+
+## Optional installation measurement
+
+The wizard sends bounded step outcomes and durations to the UserReason API to
+help improve setup. It records when initialization, agent guidance, the agent
+skill and CLI access are offered, selected, skipped, already present, applied,
+cancelled or fail. Choosing an option and successfully applying it are separate
+outcomes. The optional choices share one prompt, so its decision time is measured
+once; file and package operations have their own durations. Flow and content
+versions identify the shipped wizard experience. No experiment assignment or
+Statsig integration is included.
+
+These events include framework, package manager, execution mode, rerun status,
+version and known workspace/site scope. Source code, guidance content, file paths,
+project names, email addresses, credentials and raw errors are excluded. Before
+sign-in the wizard uses a random local installation ID. Authenticated events are
+associated with the verified account; queued anonymous events remain anonymous.
+Setup checks do not count as proof that the application has run the SDK.
+
+Disable measurement with `--no-telemetry`, `UR_TELEMETRY_DISABLED=1`,
+`DO_NOT_TRACK=1`, or the shared `userreason telemetry disable` setting. Dry runs
+send and store no analytics. The bounded queue stays outside the project,
+expires after seven days, and retries best-effort uploads with short deadlines.
+Measurement failures never fail an installation.
+
+### Browser-guided setup (explicit opt-in)
+
+Prepared compatible builds support:
+
+```sh
+npx -y @userreason/wizard@latest --browser-setup
+npx -y @userreason/wizard@latest --resume-setup SETUP_UUID
+```
+
+Run from the same project and use the same `--api-url` or `--api-environment`
+on resume. A custom API also needs its trusted `--browser-origin`. Ordinary
+startup keeps the existing terminal wizard. If a server has not enabled browser
+setup, a new connection falls back before creating a session or local effects.
+Malformed protocol data and already-started setups never fall back silently.
+Server rollout flags remain off by default; publication is a separate release.
+
+The browser owns workspace/site selection, optional choices, local-plan approval,
+source/provider consent, candidate selection, inactive draft saves, sample skip,
+finish and cancellation. The local runner executes only approved typed effects
+using the separate owner-only setup capability. It does not read or replace
+normal CLI credentials. Stopping the local process leaves the setup resumable;
+cancelling in the browser does not undo completed edits or delete created drafts.
+Terminal results list verified, failed and manual outcomes separately.
+
+Source analysis starts only when the server advertises availability. Before
+consent, the terminal displays the exact bounded excerpts, hashes and line ranges;
+the browser receives metadata only. No Git history is collected. Approving the
+source packet and providers is separate from approving installation. Immediately
+before upload, the runner revalidates full files and installed package versions.
+Changed source requires fresh browser review. Raw source stays in memory and is
+not written to setup receipts. A lost analysis response is polled rather than
+repeated; browser retry starts a fresh preparation and consent.
+
+The optional footer choice defaults to No. `planFooter` preserves the reachable
+host footer and adds a plain `Powered by UserReason` link when placement is
+unambiguous. Missing/ambiguous placement and customized skills are manual outcomes.
+Package managers use fixed allowlisted arguments, disabled scripts and exact
+versions. Lockfiles are declared generated outputs and verified by recorded hashes.
+Interrupted commands reconcile local receipts instead of blindly reinstalling.
+
+At the optional agent test, the terminal prints a setup-scoped command and the
+prompt “What do UserReason dummy respondents think?”. It reads only the static
+sample and records its receipt; it does not read production responses. Installing
+the agent skill and reading that sample never grant ongoing CLI access. Authorize
+continuing access separately with `userreason login --open` when wanted.
+
+### Optional first survey
+
+After a successful interactive installation, choose Review suggestions, Continue
+with my coding agent, or Finish setup. `--yes`, JSON and diff modes never approve
+source analysis. Suggestions reuse the CLI's bounded collector, explicit provider
+consent and per-candidate draft review; failures leave installation intact. An
+ongoing site-read/survey-write connection is required, not the installation grant.
+No automatic source edits or publication. Resume/cancel/delete with the CLI's
+`first-survey` group. See the CLI README and installed skill for privacy, retention,
+hash-guarded manual instrumentation and separate publication authorization.
+
+The production profile for wizard 0.3.0 selects `@userreason/js@0.5.0` and
+`@userreason/cli@0.2.0`. New installs use those exact versions; an installed
+stable patch in the same release line is reused. Earlier SDK 0.4 and CLI 0.1
+releases are upgraded when installation is selected. Browser setup still
+requires explicit `--browser-setup` and a server advertising the capability.

package/dist/agent-note.js

@@ -3,12 +3,12 @@ import {
   applyAgentNote,
   offerAgentNote,
   planAgentNote
-} from "./chunk-KV6GSBTQ.js";
-import "./chunk-SJVLAYGG.js";
-import "./chunk-6DCCEVPD.js";
-import "./chunk-YMSGDC4B.js";
-import "./chunk-OEGNGGZ2.js";
-import "./chunk-GGARVSPT.js";
+} from "./chunk-Z6ZY7X2K.js";
+import "./chunk-GOYBYAUH.js";
+import "./chunk-5IO5LY22.js";
+import "./chunk-CADA5WZ3.js";
+import "./chunk-NMHTFC5Q.js";
+import "./chunk-77SK7TKZ.js";
 import "./chunk-WL3NOIM6.js";
 export {
   applyAgentNote,

package/dist/agent-skill.js

@@ -3,11 +3,11 @@ import {
   applyAgentSkill,
   offerAgentSkill,
   planAgentSkill
-} from "./chunk-5Z5J6PY5.js";
-import "./chunk-SJVLAYGG.js";
-import "./chunk-YMSGDC4B.js";
-import "./chunk-OEGNGGZ2.js";
-import "./chunk-GGARVSPT.js";
+} from "./chunk-NZ65XDNL.js";
+import "./chunk-GOYBYAUH.js";
+import "./chunk-CADA5WZ3.js";
+import "./chunk-NMHTFC5Q.js";
+import "./chunk-77SK7TKZ.js";
 import "./chunk-WL3NOIM6.js";
 export {
   applyAgentSkill,

package/dist/analytics.js

@@ -0,0 +1,9 @@
+import { createRequire as createWizardRequire } from 'node:module'; import { fileURLToPath as wizardFileURLToPath } from 'node:url'; import { dirname as wizardDirname } from 'node:path'; const require = createWizardRequire(import.meta.url); const __filename = wizardFileURLToPath(import.meta.url); const __dirname = wizardDirname(__filename);
+import {
+  WizardAnalytics
+} from "./chunk-UN24JFV2.js";
+import "./chunk-77SK7TKZ.js";
+import "./chunk-WL3NOIM6.js";
+export {
+  WizardAnalytics
+};