Changelog

All notable changes to Ssscript are documented here. Newer releases are listed first.

Unreleased

Changes in development and not yet released.

[0.5.8] — 2026-04-09

[0.5.7] — 2026-04-07

Changes in development and not yet released.

Added

Multi entry starter — New catalog id multi-entry with template slug starter-multi-entry: production builds emit per-page JS bundles (multi-page deploy profile). Bundled offline in the desktop app next to Single entry.

Changed

Starters — Single entry (basic) and Multi entry (multi-entry) are the shipped choices. Legacy catalog ids (animation, multi page, page transition) still resolve to the basic zip. Older project.json starter ids continue to work.

[0.5.6] — 2026-04-07

[0.5.5] — 2026-04-06

Added

Component catalog package — Added packages/component-catalog with manifest build tooling and starter catalog items.
Catalog UI integration — Added catalog preview/paste surfaces in desktop app and web routes for component showcase pages.

Changed

Catalog interaction polish — Refined catalog rendering and paste-block handling across app/web surfaces.
Default copy behavior — Disabled default auto-copy behavior in the app chat flow.

[0.5.4] — 2026-04-06

Added

Build URL diagnostics script — Added apps/app/scripts/print-build-urls.mjs and root shortcuts to inspect effective build targets before packaging.

Changed

Release workflow checks — scripts/release.sh now includes clearer target verification steps in the release flow.
Tauri build env handling — Updated desktop/Tauri-side environment wiring used by signed build commands.

Fixed

Staging build confidence — Reduced ambiguity around which backend URL a signed build will use.

[0.5.3] — 2026-04-06

Changed

Desktop URL targeting — Updated Tauri/runtime URL resolution logic to better distinguish local, staging, and production API targets.
Web API base handling — Tightened web API base selection behavior in app-side URL utilities.

Fixed

Build-time/deploy-time mismatch risk — Reduced the chance of desktop builds silently targeting an unintended web backend.

[0.5.2] — 2026-04-06

Changed

Toolchain metadata updates — Workspace lockfile and turbo configuration were refreshed for release consistency.
Version alignment — App and release metadata files were bumped and regenerated for 0.5.2.

Notes

This release is primarily maintenance-focused with no major user-facing feature additions.

[0.5.1] — 2026-04-06

Changed

Release metadata refresh — Version numbers, release index files, and generated changelog references were aligned for 0.5.1.

Notes

This release does not include functional product code changes beyond release bookkeeping.

[0.5.0] — 2026-04-06

Added

@repo/agent-prompts package — Centralized runtime prompt assets (system/workflows/webflow variants) with bundled and filesystem loading support.
Runtime prompt endpoint — Added GET /api/ai/prompts in apps/web for server-delivered prompt text and metadata.
Turn strategy + tests — Introduced structured turn strategy logic and coverage in packages/agent-runtime.
Model catalog expansion — Added more provider/model entries and updated app/runtime model-selection plumbing.
Prompt workflow set — Added workflow prompt files (build-layout, chat-light, designer-canvas, general-build, publish-site, repo-module, webflow-data).

Changed

Runtime prompt loading — Runtime can prefer remote prompts and fall back to bundled prompt text with versioned loading behavior.
MCP/runtime loop internals — Refined MCP manager and loop execution behavior for multi-tool turns and safer prompt orchestration.
Desktop API routing defaults — Strengthened web API base resolution and env handling (staging vs production defaults).
Telemetry and usage plumbing — Updated app/web telemetry and usage-meter internals for more consistent reporting.

Fixed

AI panel reasoning formatting — Improved streamed reasoning/text formatting and copy behavior in the editor chat panel.
Entitlements summary stability — Fixed an app-level entitlement summary issue that could produce stale/incorrect display state.

[0.4.0] — 2026-03-31

[0.3.0] — 2026-03-30

Developer / repo layout

Agent prompts — Canonical runtime text lives in packages/agent-runtime/src/prompts/ (system-prompt.md, webflow-builder-prompt.md). The @repo/system package was removed; release scripts do not embed these paths.

[0.2.0] — 2026-03-30

[0.1.2] — 2026-03-29

[0.1.1] — 2026-03-29

Documentation (this site)

Roadmap — New page with two lists: Work in progress (shared KB, production env, etc.) and Future (integrations, telemetry ops, polish, longer-horizon items).
Bug reports & AI diagnostics — New reference page describing crash recovery vs the AI panel Report this error action, and what diagnostic context is included.
AI edits — Expanded with “When something goes wrong” (soft vs hard errors, link to bug-report reference).
Troubleshooting — New subsection for reporting AI errors with a link to the reference page.

Added

AI panel — Report this error — On real AI failures (not soft “no changes” / warning-style messages), you can submit a sanitized bug report while signed in. The payload includes an aiContext bundle: prompt and recent prompts, selected model and agent, view state, activity stream tail, thinking snippet, terminal tail, workspace and open file paths, runtime address, connection/MCP metadata, pending diff file paths, and conversation slot id; plus existing breadcrumbs. The server stores an extended text record (e.g. for support review).
Model catalog expansion — Added support for 10 AI models across OpenAI, Anthropic, xAI, and Google through the shared AI proxy provider.
Model picker UI — Added in-app model selection in the AI panel (Auto + fast + smart model groups), persisted per conversation slot.
Per-model pricing map — Added model-specific pricing in shared data utilities for fallback spend estimation and UI usage views.
Past conversations (agent) — History drawer can open an archived thread; if the original carousel slot is still open, the app switches to that tab instead of duplicating.

Changed

Usage metering — AI usage tracking now records prompt/completion/reasoning tokens and prefers gateway-reported cost when available.
Runtime project config hydration — Project agent.json setup now auto-upgrades provider/model catalog entries and normalizes legacy model IDs.
AI request flow reliability — Improved runtime prompt/message fallback handling and richer activity/status streaming during runs.
Runtime second-message / transport — Event stream (/event SSE) is allowed to finish closing after each prompt (bounded wait) so the next session HTTP call does not fail with error sending request. Activity status chunks preserve leading spaces for correct streamed text merging; connectivity handling recognizes the same error pattern.

Fixed

Streamed assistant text spacing — Merged status tokens no longer routinely produce glitches like missing spaces between words or broken contractions when displaying live assistant output.
Diff accept UI (pending message) — Accepting a diff without auto-apply no longer leaves the pending diff message stuck due to a batched state update fix.

[1.0.0] — 2026-03-01

Initial release.

Added

Projects — Create and open projects from a dedicated projects folder.
Editor — File tree, code editor (CodeMirror), terminal (Bun), and dependency management.
AI edits — Natural-language edits via the agent runtime; accept or reject proposed diffs.
Publish & embed — Build in-app, publish to Cloudflare, copy Webflow-ready embed snippet.
Account — Sign in (Clerk); publish and AI can be gated to authenticated users.
Keyboard shortcuts — Shortcuts for general, code view, AI panel, and modals.

Format is based on Keep a Changelog.