Usage

Usage

Once you've opened a gald3r project (see Install), Throne presents its panels through the top navigation. This page describes what each area does. Availability varies by tier — some admin and coordination surfaces are Team/Org-gated and degrade to an honest offline state without a backend.

Project data (the read-only dashboard)

Throne reads your project's .gald3r/ state directly from disk via the Rust reader layer and presents live views. Reading never writes.

Panel	Reads	Shows
Project	PROJECT.md	Mission excerpt, identity (name / id / vault location)
Tasks	TASKS.md + tasks/	Task counts and rows by status ([📋] [🔄] [🔍] [✅] [❌])
Kanban	tasks	Board view of tasks by status column
Bugs	BUGS.md + bugs/	Bug counts and rows
Features / Releases / Plans / Subsystems / Constraints / Idea Board	the matching .gald3r/ files	Each domain's records
Activity / Topology	contributions, linking topology	Recent activity and the workspace link graph

The Tasks and Bugs sections also carry an Active Agent / Swarm Activity column that reads claim metadata from task/bug frontmatter — phase badge (SPEC / CODE / REVIEW), owner short-id, relative claim time, STALE detection, and swarm grouping — with a "show active only" filter.

Vault viewer

The Vault panel is a reader for your project's knowledge vault (.gald3r/vault/ metadata and notes). It surfaces vault content within the desktop app so you can browse research, memory, and ingested material without leaving Throne. Like every reader, it is read-through only — the vault on disk always belongs to you.

Install / update / repair a project

From the Workspace page, Setup / Update Project opens a folder picker with a New project vs Update existing chooser:

• New routes into the spawn wizard whose interview step is a PROJECT.md-basics web form (name / mission / tech stack / goals), then scaffolds .gald3r/ through the shared pipeline.
• Update runs a live version-check and a guarded update modal.

Every write goes through a dry-run preview → confirm gate → apply → rollback pipeline. The apply executor runs in Rust (no Python / engine-CLI dependency): it writes new files from the bundled template snapshot and merges framework files while an absolute user-data denylist (tasks, bugs, features, releases, vault, PROJECT.md, TASKS.md, …) guarantees your own content is never overwritten. A backup ZIP is taken and integrity-verified before anything is written, and a failed apply rolls back automatically.

AI: routing, local models, credentials

• AI Routing (Settings → AI Routing) chooses how Throne picks a provider per call, with four modes: Local first (default), Cloud first, Local only (offline), and Manual / ask. You set a default chat target, a drag-to-reorder fallback chain, a streaming preference, and per-provider cost labels. A Test provider action reports whether a candidate would route (capability + credential + availability), without issuing a real chat call. Preferences persist locally to ai_routing.json.
• Local models — Throne bundles adapters for three local server types and never auto-downloads or auto-starts a model; it connects to services you run:
  • Ollama (default http://localhost:11434, no API key) — a Check Endpoint action probes /api/version and lists installed models via /api/tags.
  • LM Studio (default http://localhost:1234, no API key) — readiness via /v1/models; no Authorization header is sent.
  • Custom local OpenAI-compatible (any base URL; optional key for keyed gateways such as vLLM, llama.cpp server, llamafile).
• Credentials (Settings → Credentials) stores provider API keys in the OS keychain (Windows Credential Manager / macOS Keychain / Linux Secret Service) — never in plaintext on disk. The on-disk metadata carries only a masked preview and non-secret fields. When the keychain is unavailable, credential operations fail closed with no plaintext fallback.
• AI Assistant / Chat History / Context Window / API Usage surface the assistant chat, its history, the live context window, and per-provider usage.

Mission Control (Team/Org)

Mission Control is a live god-view of the active org: Developers, Swarms, Claimed Territory, and an Escalation Inbox on one tenant-scoped data spine, plus a collision heatmap, velocity/drift analytics, a merge-conflict decision deck, and coordinator presence. Related coordination panels: Merge Conflicts, Active Coordinators, Agents, and Maestro. These are cloud-backed and require a reachable world_tree backend; without one the pages degrade to an honest offline state rather than failing silently.

Social & collaboration

• Social — the in-app social / chat surface for talking with teammates and coordinators.
• Collaborate / Stream / Sessions — collaboration surfaces, a stream mode, and a session browser.

Admin & registration (Company / Team / Org)

The Company area hosts cloud-backed identity and team management:

• Company Identity — register and manage your company/org identity.
• Team and Team Admin — team roster, roles, and administration.
• Organizations — org-level administration.

Deeper operational admin panels include Database Admin, Feature Flags, Crash Analytics, Update Health, Systems Health, Monitors, and API Usage. Like Mission Control, these read/write through world_tree and show a "No backend configured" banner when no backend is set.

The integrated IDE

Throne includes a host-side IDE for editing project files in place:

• File Explorer with a writable, syntax-highlighted code editor (Prism grammars bundled; debounced auto-save; on-disk-change conflict prompt) and LSP diagnostics for TypeScript / JavaScript.
• Terminal — a real host pseudo-terminal (ConPTY/PowerShell on Windows, openpty on unix) rooted at the workspace.
• Review — an IDE review surface.
• Installed Platforms, Agents Manager, Skills Manager, Commands Manager, Extensions Manager, Tools & MCPs, Personality & Skill Packs — manage the gald3r components installed in the project, including an in-modal editor for personality-pack rule files and skill-pack files.

Personalization: themes & avatars

• Theme (Settings) — pick light / dark / gald3r, an accessible WCAG-AA token system persisted locally.
• Themes / Theme Builder — browse theme packs and author your own with the theme editor.
• Personality Add-ons — install/uninstall personality packs (reference-counted, with required / recommended / optional mega-fan-skill associations) and author Custom Personality Packs (CRUD with reversible delete).
• Avatar / Avatar Studio / Animation — a 2D avatar player covering idle / listening / thinking / speaking / error states with a bundled placeholder rig, plus an overlay view for future OBS-style output. (Visuals are placeholder in v0.x.)

Automation & settings

• Scheduler / Schedules — the Chrono Scheduler for fire-and-forget scheduled gald3r command sequences (name, project, ordered commands, cron/preset trigger, run history). (In shipped thin-client builds the scheduler engine is server-side.)
• Client Profiles / Connections / Profile Binding / Cascade Inspector — client configuration and settings-cascade inspection.
• Data & Privacy (Settings → Profile) — one-click telemetry opt-out.
• Onboarding, Search, Recent Projects, Report an Issue, Security, Browser, Heartbeat — supporting surfaces for setup, search, quick project switching, issue reporting, and health.

Workspace-Control boundaries

When you open a Workspace-Control owner, each declared member repository stays an independent git root with its own branches, remotes, and history. Throne surfaces members for navigation; it never reaches into a member to perform writes the owning task did not declare in its workspace policy.

---

Gald3r Throne v0.2.0 — panels available depend on tier and backend configuration.