CtxrBack to Ctxr

Product updates

Changelog

Everything new, improved, and fixed in Ctxr—one clear record of how the context layer is evolving.

v0.5.0Latest release

Public product surface and embedded agent

The public site now documents the complete Ctxr workflow, exposes product updates, and provides a bounded AI support surface before authentication.

Added

Ctxr Agent

Embedded a server-backed assistant in the navigation overlay for questions about Ctxr, project memory design, build planning, and coding-agent workflows.

Security

Bounded AI request pipeline

Chat input is schema-validated, length-limited, and capped to recent turns before reaching the OpenAI Responses API. API credentials remain server-only.

Added

Product changelog

Published a versioned release history with dedicated metadata, stable /changelog routing, and direct access from the hero and footer.

Changed

Authenticated URL namespace

Moved all workspace routes under /app, added a protected /app matcher, and retained non-permanent redirects for legacy links.

Added

Landing system

Shipped dedicated hero, product overview, feature matrix, token comparison, pricing, FAQ, final CTA, footer, and responsive navigation components.

Added

Motion and rendering

Added GSAP entrance and scroll-trigger sequences, split-text masks, Lenis scrolling, light-ray rendering, a seamless platform marquee, and a two-stage branded preloader.

Changed

Typography and identity

Standardized Inter across the application, retained local Azeret Mono for landing headlines, and deployed the Ctxr mark, wordmark, favicon, metadata, and social descriptions.

Fixed

Navigation overlay stability

Removed the glass and 3D composition that could collapse or cull menu content. Navigation and Agent faces now share the measured content height.

v0.4.0

GitHub repository integration

Projects can inspect repository structure directly and recover from the most common authentication and branch-configuration failures.

Added

Repository file browser

Added directory traversal, breadcrumb navigation, file metadata, bounded source previews, and direct GitHub links inside project detail pages.

Added

Project repository configuration

Project creation accepts a GitHub URL, default branch, framework, and language, with curated selections and explicit custom-value input.

Security

Repository URL and path validation

GitHub requests validate host, owner, repository, and path input before constructing API requests. Traversal and absolute-path input are rejected.

Fixed

Expired OAuth token fallback

A 401 response now triggers one anonymous retry, allowing public repositories to remain readable when a stored GitHub token has expired.

Fixed

Default branch resolution

A missing configured ref now triggers repository metadata discovery and a retry against GitHub's reported default branch.

Changed

GitHub API compatibility

Requests now send an explicit GitHub API version and use authenticated access first to preserve private-repository support and higher rate limits.

Changed

Repository error states

Separated invalid authorization, rate limiting, missing paths, unsupported content, oversized files, and connectivity failures into actionable messages.

v0.3.0

Repository-grounded AI workflows

Ctxr now converts repository evidence and durable project memory into structured context, architecture maps, and executable implementation plans.

Added

AI Memory Suggestions

Extracts architecture, conventions, constraints, dependencies, deployment requirements, and security decisions from selected high-signal repository files.

Added

AI Project Guide

Produces a typed implementation brief containing a summary, ordered approach, relevant files, risks, verification checks, and a portable agent prompt.

Added

Build Plans

Persists generated guides with draft, ready, in-progress, and completed lifecycle states, plus per-check completion tracking.

Added

Codebase Map

Models repository entry points, modules, responsibilities, dependencies, execution flows, data models, and concise navigation guidance for coding agents.

Added

Markdown export

Build prompts and plans can be copied or exported as Markdown for Codex, Cursor, Claude Code, and external application builders.

Changed

Evidence selection

AI workflows prioritize bounded configuration, schema, authentication, routing, dependency, and documentation files instead of sending an unrestricted repository snapshot.

Security

Ownership checks

Generation, save, update, and map operations verify the authenticated user owns the target project before reading repository data or writing results.

Fixed

Structured output validation

OpenAI responses are parsed against Zod schemas before persistence, preventing incomplete guide, suggestion, and map payloads from entering application state.

v0.2.0

Project memory workspace

The authenticated workspace now supports the complete project and memory lifecycle with searchable context, operational telemetry, and scalable indexes.

Added

Dashboard overview

Added project, memory, build-plan, and codebase-map totals with activity charts, category distribution, coverage indicators, readiness signals, and quick actions.

Added

Project management

Projects can be created from a modal, listed in a paginated table, opened as dedicated workspaces, and deleted with dependent data handled by database cascades.

Added

Durable memories

Memories support Markdown content, thirteen categories, four importance levels, pinning, project association, detail views, and modal creation from project pages.

Added

Context search

Added in-place search from the dashboard and navigation shell, including title, content, and project-name results without leaving the active page.

Added

Activity history

Project and memory mutations create timestamped activity records displayed in bounded, scrollable project timelines and dashboard summaries.

Changed

Paginated indexes

Project lists are limited to twelve rows per page and memory lists to fourteen, with stable pagination controls and empty states.

Changed

Project memory layout

Project detail pages render six memories initially in a three-column grid, with progressive expansion and consistent card dimensions.

Added

Interaction feedback

Creation, updates, AI saves, copy actions, and failures report through consistent toast notifications and pending button states.

Fixed

Modal composition

Normalized modal backdrops, conditional custom fields, focus behavior, content overflow, and close transitions across project, memory, search, and AI workflows.

v0.1.0

Application foundation

Initial production architecture for authenticated, user-owned project context backed by PostgreSQL and GitHub identity.

Added

GitHub authentication

Implemented Auth.js with the Drizzle adapter, database sessions, GitHub identity mapping, and repository authorization scopes.

Security

Protected workspace

Authenticated middleware guards the application namespace, while server-side data access resolves the current user before returning owned resources.

Added

PostgreSQL persistence

Added Neon serverless connectivity and Drizzle schemas for users, OAuth accounts, sessions, projects, memories, activities, build plans, and codebase maps.

Added

Relational integrity

Configured foreign keys, cascading project cleanup, unique ownership constraints, and indexes for user, project, category, pinned state, and update-time access patterns.

Added

Seed workflow

Added a repeatable TypeScript seed entry point with representative projects, memories, and activity data for local product evaluation.

Added

Application shell

Shipped a responsive sidebar, authenticated profile state, workspace navigation, loading UI, error boundary, empty states, and not-found handling.

Changed

Developer validation

Established TypeScript no-emit checks, ESLint validation, Drizzle migration scripts, and environment templates for GitHub, database, and OpenAI configuration.