alchemist/backlog.md

Alchemist Backlog

Current and future work for Alchemist, organized around the actual repo state rather than historical priorities.

Alchemist should remain an automation-first media optimization tool, not drift into a general-purpose media workbench.

---

Implemented / In Progress

These items now exist in the repo and should be treated as current product surface that still needs hardening, documentation, or iteration.

Conversion / Remux Workflow

• Dedicated Convert page for single-file upload-driven conversion
• Probe-driven UI with container, video, audio, subtitle, and remux-only controls
• FFmpeg command preview
• Temporary upload/output lifecycle under ~/.config/alchemist/temp
• Reuse of the existing queue and worker system
• Status polling and download flow
• Treat this as an experimental utility, not a second core product track

Notification Platform Expansion

• Provider-specific notification target model backed by config_json
• Discord webhook, Discord bot, Gotify, generic webhook, Telegram, and email targets
• Richer event taxonomy:
  • encode.queued
  • encode.started
  • encode.completed
  • encode.failed
  • scan.completed
  • engine.idle
  • daily.summary
• Per-target event filtering
• Daily summary scheduling via daily_summary_time_local

API Token Authentication + API Docs

• Named static API tokens with read_only and full_access classes
• Hash-only token storage, plaintext shown once at creation
• Token management endpoints and Settings UI
• Hand-maintained OpenAPI contract plus human API docs

Distribution Foundation

• In-repo distribution metadata sources for:
  • Homebrew
  • AUR
  • Windows update-check metadata
• Release workflow renders package metadata from release assets/checksums
• Windows in-app update check against GitHub Releases

Expanded Library Intelligence

• Duplicate groups remain
• Storage-focused recommendation categories added:
  • remux-only opportunities
  • wasteful audio layouts
  • commentary/descriptive-track cleanup candidates
• Direct actions now exist for queueing remux recommendations and opening duplicate candidates in the shared job-detail flow

Engine Lifecycle + Planner Docs

• Runtime drain/restart controls exist in the product surface
• Backend and Playwright lifecycle coverage now exists for the current behavior
• Planner and engine lifecycle docs are in-repo and should now be kept in sync with shipped semantics rather than treated as missing work

Jobs UI Refactor / In Flight

• JobManager has been decomposed into focused jobs subcomponents and controller hooks
• SSE ownership is now centered in a dedicated hook and job-detail controller flow
• Treat the current jobs UI surface as shipping product that still needs stabilization and regression coverage, not as a future refactor candidate

---

Active Priorities

0.3.1 RC Stability Follow-Through

• Keep the current in-flight backend/frontend/test delta focused on reliability, upgrade safety, and release hardening
• Expand regression coverage for resume/restart/cancel flows, job-detail refresh semantics, settings projection, and intelligence actions
• Keep release docs, changelog entries, and support wording aligned with what the RC actually ships

Per-File Encode History Follow-Through

• Attempt history now exists in job detail, but it is still job-scoped rather than grouped by canonical file identity
• Next hardening pass should make retries, reruns, and settings-driven requeues legible across a file’s full history
• Include outcome, encode stats, and failure reason where available without regressing the existing job-detail flow

AMD AV1 Validation

• Validate Linux VAAPI and Windows AMF AV1 paths on real hardware
• Confirm encoder selection, fallback behavior, and defaults
• Keep support claims conservative until validation is real
• Deferred from the current 0.3.1-rc.5 automated-stability pass; do not broaden support claims before this work is complete

---

Later

Documentation

• Architecture diagrams
• Contributor walkthrough improvements
• Video tutorials for common workflows

Code Quality

• Increase coverage for edge cases
• Add property-based tests for codec parameter generation
• Add fuzzing for FFprobe parsing

Planning / Simulation Mode

• Promote this only after the current Active Priorities are done
• Single-config dry run first
• No comparison matrix or scenario planner until the first simple flow proves useful

Audio Normalization

• Add opt-in EBU R128 loudness normalization during transcode
• Surface loudness metrics in job detail
• Keep copy-mode bypass behavior explicit
• Keep this secondary unless it clearly supports the automation-first mission

Auto-Priority Rules

• Add explainable enqueue-time priority automation
• Manual priority overrides must still win
• Matched rules must be visible in the UI to keep queue behavior trustworthy

UI Improvements

• Tighten settings and detail-panel consistency
• Improve dense forms, empty states, and action affordances
• Keep this narrowly scoped to automation-supporting UX problems

Keyboard Shortcuts

• Add a concrete shortcut set for common jobs/logs/conversion actions
• Avoid a vague “shortcut layer everywhere” rollout
• First likely cut if scope pressure appears

Features from DESIGN_PHILOSOPHY.md

• Add batch job templates

Distribution Follow-Ons

• Flatpak / Snap packaging
• Additional installer polish beyond the current Windows update-check flow
• Only promote these if they become strategically important

---

Out of Scope

• Custom FFmpeg flags / raw flag injection
• Distributed encoding across multiple machines
• Features that turn Alchemist into a general-purpose media workbench
• Fuzzy media-management intelligence that drifts away from storage quality and encode operations