How framesphere/frametrain-app works
FrameTrain is a local-first desktop application for preparing datasets, training and evaluating AI models, then managing and exporting the results. The main path is user interaction in the React app, through Tauri commands and Rust managers, to Python training or inference and back to persisted run/model state. Major parallel capabilities include the visual Synapse builder, dataset Studio, testing, model versioning, and hosting. Edges below are limited to relationships stated in the README or visible in the supplied excerpts and source index; several subsystem details are unsampled.
Main entry points: training_manager.rs Training manager · App.tsx React app · train_engine.py Python train engine · SynapseBuilder.tsx Synapse canvas
Components 25
Desktop app
- React app
src/App.tsx - Training UI
src/components/TrainingPanel.tsx - Training state
src/contexts/TrainingContext.tsx - Tauri backend
src-tauri/src/main.rs
Data and Studio
- Dataset management UI
src/components/DatasetFileManager.tsx - Dataset manager
src-tauri/src/dataset_manager.rs - Metadata database
src-tauri/src/database.rs - Dataset Studio UI
src/components/studio/StudioPanel.tsx - Studio services
src-tauri/src/studio_manager.rs
Train and evaluate
- Training manager
src-tauri/src/training_manager.rs - Python train engine
src-tauri/python/train_engine/train_engine.py - Training plugins
src-tauri/python/train_engine/plugins - Evaluation plugins
src-tauri/python/test_engine/plugins - Run analysis
src-tauri/src/analysis_manager.rs
Models and hosting
- Model manager UI
src/components/ModelManager.tsx - Model manager
src-tauri/src/model_manager.rs - Model versioning UI
src/components/VersionManager.tsx - Hosting UI
src/components/hosting/HostingPanel.tsx - Hosting services
src-tauri/src/hosting_api.rs
Synapse builder
- Synapse canvas
src/components/synapse/SynapseBuilder.tsx - Synapse service
src-tauri/python/synapse/service.py - Graph and code nodes
src-tauri/python/synapse/graph.py - Canvas loader
src-tauri/python/train_engine/plugins/canvas/loader.py
Other components
- User
- Ollama
Connections 28
- User → React app: uses
- React app → Training UI: shows
src/App.tsx - Training UI → Training manager: starts training
- Training manager → Python train engine: runs
src-tauri/src/training_manager.rs - Python train engine → Training plugins: loads plugin
src-tauri/python/train_engine/train_engine.py - Python train engine → Evaluation plugins: uses evaluation
- Training manager → Dataset manager: uses datasets
src-tauri/src/training_manager.rs - Training manager → Metadata database: records runs
src-tauri/src/training_manager.rs - Training manager → Run analysis: requests analysis
src-tauri/src/training_manager.rs - Training manager → Training state: updates progress
- User → Dataset management UI: manages data
- Dataset management UI → Dataset manager: invokes commands
- Dataset manager → Metadata database: stores metadata
src-tauri/src/dataset_manager.rs - User → Dataset Studio UI: creates datasets
- Dataset Studio UI → Studio services: invokes commands
- User → Model manager UI: imports models
- Model manager UI → Model manager: invokes commands
src/components/ModelManager.tsx - Model manager → Metadata database: checks model records
src-tauri/src/model_manager.rs - User → Model versioning UI: compares versions
- User → Hosting UI: runs models
- Hosting UI → Hosting services: requests hosting
- Hosting services → Ollama: integrates
- User → Synapse canvas: builds network
- Synapse canvas → Synapse service: invokes service
- Synapse service → Graph and code nodes: checks and exports
src-tauri/python/synapse/service.py - Synapse service → Canvas loader: loads canvas
src-tauri/python/synapse/service.py - Synapse canvas → Training manager: trains canvas model
- React app → Tauri backend: invokes commands
src/App.tsx