How dreamsartificial648-sketch/adam-code-architecture works
ADAM is a desktop application for planning and running AI model training and image generation, with dataset workspaces, video/gameplay features, and remote control. Its main workflow starts in the desktop UI, turns a request into a plan, gates execution through job management, then dispatches registered tools and tracks results. The graph also shows dataset and generation workspaces plus the remote API. README content was empty, and several subsystems are represented by their tree paths because their implementation was not sampled; uncertain internal wiring is omitted.
Main entry points: registry.py Training tools · main_window.py Desktop UI · job_manager.py Job manager · app.py Application setup
Components 23
Desktop application
- Application setup
app.py - Desktop UI
main_window.py
Planning and jobs
- Request planner
planner.py - Tool registry
registry.py - Job manager
job_manager.py - Tool executor
executor.py - Job records
job_manager.py
Model workflows
- Image generation
generations.py - Training tools
registry.py - LoRA workflow
lora.py - Diffusion workflows
ddpm.py - Flow workflows
flow_matching.py - Video LoRA
video_lora.py - Oasis gameplay
oasis_player.py
Datasets and media
- Asset registry
assets.py - Dataset studio
studio.py - Video datasets
video_lora.py - Frame reviewer
cnn_reviewer.py
Remote access
- Remote API
remote_access.py - Remote dispatcher
remote_dispatcher.py
Other components
- User
- Browser or device
- Model backends
Connections 31
- User → Desktop UI: uses
- Application setup → Desktop UI: creates
- Application setup → Request planner: wires
- Application setup → Tool registry: creates
- Application setup → Job manager: creates
- Application setup → Tool executor: creates
- Desktop UI → Request planner: plans requests
- Desktop UI → Job manager: manages jobs
- Request planner → Tool registry: checks tools
- Job manager → Tool executor: runs steps
- Job manager → Job records: persists jobs
- Tool executor → Tool registry: resolves tools
- Tool executor → Training tools: dispatches tools
- Desktop UI → Image generation: opens workspace
- Image generation → Tool registry: selects providers
- Image generation → Asset registry: reads models
- Image generation → Job manager: submits plans
- Training tools → LoRA workflow: provides workflow
- Training tools → Diffusion workflows: provides workflow
- Training tools → Flow workflows: provides workflow
- Training tools → Video LoRA: provides workflow
- Training tools → Oasis gameplay: provides workflow
- Dataset studio → Asset registry: uses assets
- Desktop UI → Dataset studio: opens workspace
- Video LoRA → Video datasets: manages clips
- Frame reviewer → Video datasets: reviews frames
- Browser or device → Remote API: sends requests
- Remote API → Remote dispatcher: dispatches commands
- Remote API → Job manager: controls jobs
- Remote API → Request planner: requests plans
- Training tools → Model backends: uses backends