Delivery outcomes

Delivery follows the dependencies between usable capabilities. GitHub Issues, milestones and the Project track implementation progress; this page explains how the product outcomes build on each other.

Outcome	Result	Depends on
Platform and dependency contracts	Native CLI/runtime and desktop communication, pinned Cueson and graph libraries, hidden process launch, encrypted secrets and offline help on Windows, macOS and Linux	Shared application contracts
Runtime and configurable storage	Workspace ownership, durable jobs, migrations, model registry and complete filesystem/S3 and SQLite/PostgreSQL behavior	Platform contracts
Unified media library	Audio/video import, original identity, raw metadata capture, recording dates, per-item batch inputs and relocation	Runtime and storage
Correlated local subtitles	Supplied subtitle tracks, mapped audio derivatives, local recognition, source-timed speaker intervals and Cueson artifacts	Media library
Configurable processing	Readable presets, task capabilities, local/hosted routing and per-stage overrides	Subtitle and job contracts
Speaker continuity and terminology	Voices, names, aliases, reversible attribution, library-wide audio selection and transcription context	Processing and source-timed intervals
Desktop library and help	GUI packages that include the CLI, common core operations, playback, progress and release-matched offline docs	Runtime, media and speaker contracts
Evidence and querying	Chunked source assertions, ordered LadybugDB/ArcadeDB projection and source-linked word/speaker/time queries	Subtitle and speaker contracts
Visual exploration	Calendar/media timeline, saved force-directed graph views and keyboard/table alternatives	Desktop and graph queries
Optional query assistance	Configured model suggestions and elected validated read-only execution	Processing and query contracts
Durable distribution	Portable backups, expanded acquisition, signed packages and synchronized release/docs/schema publication	Installable packages and recovery contracts
Optional speaker models	Immutable corpus snapshots, elected training, versioned speaker-associated artifacts and CLI discovery/retrieval	Runtime, processing and speaker contracts

flowchart TB Platform[Platform and dependency contracts] --> Runtime[Runtime and configurable storage] Runtime --> Library[Unified media library] Library --> Subtitles[Correlated local subtitles] Subtitles --> Processing[Configurable processing] Processing --> Speakers[Speaker continuity and terminology] Speakers --> Desktop[Desktop library and help] Speakers --> Queries[Evidence and graph queries] Speakers --> Models[Optional speaker models and retrieval] Desktop --> Visual[Timeline and saved graph views] Queries --> Visual Visual --> Assistance[Optional query assistance] Desktop --> Distribution[Durable distribution] Queries --> Distribution

Diagram source

flowchart TB
  Platform[Platform and dependency contracts] --> Runtime[Runtime and configurable storage]
  Runtime --> Library[Unified media library]
  Library --> Subtitles[Correlated local subtitles]
  Subtitles --> Processing[Configurable processing]
  Processing --> Speakers[Speaker continuity and terminology]
  Speakers --> Desktop[Desktop library and help]
  Speakers --> Queries[Evidence and graph queries]
  Speakers --> Models[Optional speaker models and retrieval]
  Desktop --> Visual[Timeline and saved graph views]
  Queries --> Visual
  Visual --> Assistance[Optional query assistance]
  Desktop --> Distribution[Durable distribution]
  Queries --> Distribution

CLI operations establish the application contract before their desktop controls. Installing the GUI includes the CLI. Core feature parity is the goal; matching release documentation identifies any advanced or experimental CLI capabilities awaiting GUI controls.

Optional speaker training can proceed once speaker, corpus and adapter contracts exist, alongside graph and desktop work. It does not require completion of every visual or distribution feature. A platform experiment can change a framework, dependency pin or packaging method while preserving the three required operating systems and supported backend behavior.

Direct library access and querying do not require an AI account or a model download. Training and query assistance are elective. Users' observations after releases become focused corrective work through the same issue-driven workflow.