Sasha Studio Release Notes: v1.0.624 to v1.0.904

Release Period: December 2025 - January 2026
Version Range: 1.0.624 → 1.0.904
Total Commits: 819
Actual Code Changes: 1,833 files changed, 736,446 insertions, 27,893 deletions

TL;DR - Business Summary

What's New in Plain English:

This release transforms Sasha into a comprehensive analytics and quality management platform:

Advanced Reporting Dashboard - New reports for session quality, hook analytics, permission audits, error analysis, and user activity. See exactly how your team uses Sasha with drill-down capabilities and user filtering.
Prompt Library with Starring - Star your most-used prompts for quick access. A dedicated "Starred" section surfaces your favorites, and clicking any prompt inserts it directly into your chat input.
CLAUDE.md Version Tracking - Every change to your system prompt is now versioned with full diff comparison. See exactly what changed, when, and by whom—with quality metrics tied to each version.
Quality Feedback System - Thumbs up/down feedback on AI responses feeds into quality analytics. Track response quality over time and identify prompts that need improvement.
Streamlined Knowledge Navigation - Documentation now embeds directly (no iframe), with nested folder support and improved breadcrumb navigation. Access your knowledge base faster.

Business Value:

Quality Assurance: Track AI response quality with metrics tied to prompt versions
Team Visibility: See who's using Sasha, how often, and with what results
Faster Workflows: Starred prompts and click-to-insert save time on repetitive tasks
Audit Trail: Full version history of system prompts with diff comparison
Better UX: Compact tables, expandable rows, and icon-based controls reduce clutter
A/B Testing Foundation: Version tracking and quality metrics lay groundwork for testing prompt variants

Executive Summary

This release focuses on analytics, quality management, and user experience refinement. The headline additions are a comprehensive reporting dashboard with five new report types and a prompt starring system for quick access to frequently-used prompts.

The CLAUDE.md version tracking system now captures every prompt change with quality metrics, enabling teams to correlate prompt modifications with response quality over time. The feedback system lets users rate responses, feeding data into quality analytics visible in the reporting dashboard.

UI improvements emphasize information density and discoverability. Tables throughout the reporting page now use compact, expandable rows—click anywhere to reveal details. The sidebar sort controls switched from text labels to icons, and the knowledge panel embeds documentation directly instead of through an iframe.

Infrastructure work includes session management improvements, hook audit tracking for understanding Claude CLI interactions, and significant styling refinements across the application.

Major Features & Improvements

Analytics & Reporting

Hook Audit Reporting - Track Claude CLI hook executions with user attribution and filtering
Session Quality Report - Analyze quality metrics across sessions with drill-down capability
Permission Audit Report - Review permission grants and denials across the system
Error Analysis Report - Surface and categorize errors for debugging and improvement
Resource Usage Report - Monitor token consumption and resource utilization
User Activity Report - See per-user engagement metrics and activity patterns
Redesigned Tables - "Sessions needing attention" and "CLAUDE.md versions" now use compact, click-to-expand rows with relative timestamps
Timeline Features - Zoom controls and change markers on activity timeline
Fixed Usage Trend Chart - Resolved ResponsiveContainer sizing issue that prevented chart rendering

Prompt Management

Prompt Starring - Star prompts from any category for quick access
Starred Prompts Section - Dedicated section surfaces your favorite prompts
Click-to-Insert - Clicking a prompt inserts it directly into the chat input
Persistent Stars - Starred prompts survive across sessions

CLAUDE.md Version Management

Version Tracking - Every CLAUDE.md change is versioned and timestamped
Quick Diff - Compare any version against current or previous versions
Timeline Markers - Discrete version markers on the CLI activity timeline
Quality Backfill - Endpoint and UI trigger to backfill quality metrics for existing data
Settings Integration - Version comparison available directly in Settings

Quality & Feedback System

Response Telemetry - Track quality signals for AI responses
Thumbs Feedback - Users can rate responses with thumbs up/down
Feedback Toast - Visual confirmation when feedback is submitted
Per-User Metrics - Feedback counts aggregated by user in reporting
Lean Events - Lightweight feedback event capture for chat interactions

Knowledge & Documentation

Embedded SPA - Documentation embeds directly without iframe by default
Nested Folders - Knowledge docs grouped by folder in sidebar
Improved Navigation - Tighter spacing, better breadcrumbs, pinned controls
Direct Downloads - Download files directly from embedded knowledge view
React Knowledge View - New default React-based knowledge rendering

Sidebar & UI Improvements

Icon Sort Controls - Recent/Alphabetical/Starred/Archived now use icons instead of text
Visual Selection State - Filled icons indicate active sort/filter
Chat Counts - Project list shows conversation counts
Visible Drag Handles - Clear affordance for draggable items
Removed Drag-and-Drop - Simplified sidebar by removing drag reordering
Updated PWA Icons - New purple "S" logo across all icon sizes

Chat & Markdown

GFM Support - GitHub Flavored Markdown with emoji rendering
Markdown Tables - Tables render as tables instead of code blocks
Softened Feedback Icons - More subtle visual treatment for feedback UI

UX Improvements

Layout & Spacing

Tightened Layouts - Reduced padding and margins throughout for better information density
Aligned Headers - Consistent header alignment across settings, modals, and panels
Compact Cards - Prompt library and MCP cards now more compact with better spacing
Knowledge Navigation - Tighter line spacing, compressed padding, breadcrumbs above content

Interactive Components

New Project Modal - Replaced inline form with modal dialog including folder search and scroll
Expandable Panels - Settings tabs now use expandable info panels (e.g., Shared Drives, Brain Update)
Hover Action Groups - Prompt cards show actions on hover instead of always visible
Pagination - Hook audit reports now paginated for better performance

Visual Refinements

Blue Panel Style - Brain Update header converted to consistent blue panel styling
Phosphor Icons - Home icon in breadcrumbs now uses Phosphor icon library
Date Range Sync - Button highlighting now syncs with actual selected dates
MCP Card Sizing - Cards properly sized and clamped in Tools settings

Reporting Tables

Compact Row Design - All major tables converted to single-line expandable rows
Click-to-Expand - Full details revealed on row click with chevron indicators
Relative Timestamps - "2d ago" format for quick scanning
Expandable UI Pattern - Consistent expand/collapse across all report components

Stability & Reliability

Reporting

Fixed usage trend chart not rendering (ResponsiveContainer sizing)
Fixed SQL syntax error in user activity report
Connected quality charts to correct database table
Fixed expand functionality in report tables

Knowledge Embedding

Fixed double embed load issues
Resolved docs-router injection duplicates
Fixed style injection into document head
Improved embed initialization and logging

Session Management

Fixed session-created timing for deferred sends
Better JSONL title preference in sidebar
Improved WebSocket user ID tracking for hooks

Hooks & Events

Fixed hook command filter quoting
Support for pretty-printed hook event ingestion
Improved certainty score calculation

UI Polish

Restored blue brand icon in headers
Aligned settings header layout
Fixed quick action height alignment
Softened tool pills styling

AWS Bedrock Integration

1M Context Support - Implemented patch approach for extended 1M token context window
Header Propagation - Fixed header propagation to Claude CLI for extended context
Config Fallback - Extended context now uses config when environment variables missing
Payload Fixes - Corrected extended context payload structure
Restore Verification - Fixed Bedrock restore verification flow
Discovery UX - Hardened Bedrock discovery user experience
Documentation - Added evidence and explanation for internal context management vs API limits

Developer Experience

Documentation

Added certainty score design documentation
Comprehensive feature docs for hook user tracking
Upstream repo reference for session comparison

Infrastructure

Canonicalized Claude projects root path
Removed debug file logging (console only)
Enhanced error handling and logging throughout

Upgrade Notes

Reporting Dashboard

New Tables: "Sessions needing attention" and "CLAUDE.md versions" use compact row design
Click to Expand: Click anywhere on a row to reveal full details
No Configuration: New reports appear automatically in the Reports section

Prompt Starring

Enable via UI: Star any prompt using the star icon
Persistence: Stars stored in browser localStorage

Knowledge Panel

Embed Mode: Docs now embed directly by default (iframe option still available)
Clear Cache: May need to clear browser cache if experiencing display issues

PWA Icons

Cache Refresh: Clear browser/PWA cache to see updated purple "S" icons

Changelog Summary (since v1.0.624)

Features

Hook audit reporting with user filtering
Session Quality, Permission Audit, Error Analysis, Resource Usage, User Activity reports
Prompt starring with persistent storage
Starred prompts section with click-to-insert
CLAUDE.md version tracking with diff comparison
Quality backfill endpoint and UI trigger
Response quality telemetry and feedback system
Timeline zoom controls and change markers
Embedded docs SPA without iframe
Nested knowledge folders in sidebar
Icon-based sidebar sort controls
Chat counts in project list
GFM and emoji support in chat markdown
Markdown table rendering

Fixes

Usage trend chart rendering
Report table expand functionality
SQL syntax in user activity report
Double embed load issues
Hook command filter quoting
Session-created timing
ResponsiveContainer sizing
Docs-router injection duplicates
Bedrock 1M context header propagation
Bedrock extended context payload and config fallback
Bedrock restore verification flow

UI/UX

Compact expandable table rows
Relative timestamps (2d ago, 5h ago)
Softened feedback icons
Tightened knowledge navigation spacing
Aligned settings and sidebar layouts
Updated PWA icons to purple "S"

Looking Ahead

A/B Variant Testing: The versioning and quality metrics infrastructure prepares the platform for A/B testing of CLAUDE.md system prompts and prompt library variants—measure which prompts produce better outcomes
Enhanced Analytics: More granular quality metrics and trend analysis
Team Collaboration: Shared prompt libraries across users
Advanced Filtering: Filter reports by date range, project, and custom criteria
Export Capabilities: Download reports as CSV/PDF for external analysis

Jargon Buster - Technical Terms Explained

AWS Bedrock

Amazon's managed AI service that hosts Claude and other models
Like renting Claude from Amazon instead of directly from Anthropic
Bedrock supports extended 1M token context windows for longer conversations

MCP (Model Context Protocol)

Standard protocol for AI tools to interact with external systems
Like USB for AI—a common interface for connecting to databases, APIs, etc.
Sasha uses MCP to integrate with AWS, Quickbase, Google Drive, and more

CLAUDE.md

The system prompt file that defines Sasha's behavior and knowledge
Like an instruction manual that tells the AI how to respond
Version tracking lets you see exactly what changed and when

ResponsiveContainer

A Recharts component that automatically sizes charts to fit their container
The fix ensures charts display correctly regardless of screen size

GFM (GitHub Flavored Markdown)

Extended markdown syntax used by GitHub
Adds tables, task lists, strikethrough, and emoji support
Makes chat messages more expressive and readable

PWA (Progressive Web App)

Technology that lets web apps behave like native apps
Enables "Add to Home Screen" with custom icons
The new purple "S" icons appear when Sasha is installed as a PWA

Thanks for upgrading. This release gives you unprecedented visibility into how your team uses Sasha and the quality of AI responses—plus faster workflows with prompt starring and improved navigation.