BrightVision

Website: brightvision.digitaldefiance.org

A local-LLM-first desktop IDE (Tauri + React) for AI-assisted coding — spec-driven tasks, superproject git, built-in editor, and a headless agent you control from the UI (never the terminal TUI).

Built in partnership with the Cecli team — coding agent from dwash96/cecli. BrightVision adds bright_vision_core (Vision HTTP/SSE on :8741) so React talks to sessions over the API. Agent: submodule cecli/ → Digital-Defiance/cecli.

Chat — Thinking/Answer streaming, fenced code, and rendered Mermaid diagrams

Architecture

flowchart LR
  UI["React shell<br/>src/ + Tauri"]
  API["Vision API<br/>bright_vision_core :8741"]
  Agent["Cecli<br/>cecli/ submodule"]
  LLM["Ollama / cloud<br/>via LiteLLM"]

  UI -->|"HTTP + SSE"| API
  API --> Agent
  Agent --> LLM
  Agent -->|"events"| API
  API -->|"SSE"| UI

Loading

Layer	Role
Head	Chat, Tasks, Terminal, Git, Editor, Settings — left rail, not a VS Code clone
Vision API	Sessions, todos, git superproject, SSE → src/ipc/events.ts
Cecli	Models, coders, tools, slash commands, agents, MCP
Local LLM	Rust panel starts Ollama; turns run in Python core

What BrightVision does

Pillar	Highlights
Chat	Streaming Thinking/Answer, Mermaid + highlighted fences, proposed edits, confirm/queue/stop, /clear, suggested-files tray, model router, agent chips
Tasks	EARS/spec workflow — todos, layered specs, generate/refine, Implement with active task
Git	Status, diffs, commit graph, stage/commit/undo (desktop)
Editor	CM6 tabs, resizable file explorer, git badges, open-from-chat, optional language packs
Local LLM	Ollama panel, hopper preload, ping, resource overlay (CPU/RAM/GPU in rail)
Timing	Live Response/Think bar, per-model ETA, Settings history (TPS, resources, CSV)
Agents	/agent, /invoke-agent, /spawn-agent, /reap-agent + sub-agent registry

Full catalog: docs/FEATURES.md · backlog: docs/ROADMAP.md

Screenshots

Tasks — spec-driven work	Editor — tabs + explorer
Git — graph & working tree	Activity bar + confirm
Settings — timing history & resource overlay

Quick start

macOS (Homebrew)

brew tap digital-defiance/tap
brew trust --cask digital-defiance/tap/brightvision
brew install --cask brightvision

Homebrew 4.6+ requires trusting the cask once before install (unsigned tap cask). If brew install brightvision fails with a trust error, run the brew trust line above and retry.

From source

git clone https://github.com/Digital-Defiance/BrightVision.git
cd BrightVision
git submodule update --init cecli
yarn install
source activate.sh
yarn tauri dev

1. Install Ollama and copy local-llm.env.example → local-llm.env (DATA_MODEL; optional FAST_MODEL / HEAVY_MODEL / MODEL_ROUTER — docs/LOCAL_LLM.md)
2. Terminal → Start (Vision API on :8741)
3. Chat when the session is live

Tech stack

• Shell: Tauri v2 (Rust) + React 18 + TypeScript + Vite + MUI v6
• Engine: Cecli (cecli/) + Vision API (bright_vision_core/ in this repo)
• Tests: yarn test:fast · yarn test:local · TESTING.md

Configuration

Settings — model, workspace, fonts, timing stats, suggested files, model hopper, editor languages, agents (subagent_paths in cecli config).

Agents (cecli) — define sub-agents as *.md under paths in .cecli.conf.yml; use chat Agents chips or /invoke-agent reviewer …. See Settings → Agents & sub-agents.

Documentation

Doc	Topic
FEATURES.md	Product feature catalog
ROADMAP.md	Status & fix order
UPSTREAM_CECLI.md	Cecli submodule + Vision API layout
LOCAL_LLM.md	Ollama & local panel
SPEC_DRIVEN_DEV.md	Tasks workflow
IPC.md	HTTP API & SSE
DEVELOPMENT.md	Dev setup
TROUBLESHOOTING.md	Stuck sessions, :8741
TESTING.md	Test matrix

License

MIT — see LICENSE. Copyright (c) 2026 Digital Defiance, Jessica Mulein

Contributing

Read ROADMAP.md before substantive work. Open issues with repro steps (workspace path, expected vs actual).