How xiaollull/xiria-aicanvas works
XiriaCanvas AI is a local image-generation application with a React workspace and a Python inference backend. The principal workflow is configuring a generation in the workspace, running it through a native engine, and viewing results in the gallery. The README also documents model management, image-to-image work, postprocessing, LoRA management, and an AI assistant as product capabilities. Several backend and UI connections are inferred from product documentation rather than sampled call sites, so those links are dashed and unverified.
Main entry points: App.jsx Workspace UI · inference_server.py Inference server · anima_pipeline.py Anima engine · flux_pipeline.py FLUX.1 engine
Components 21
Generation Workspace
- Workspace UI
src/App.jsx - Engine settings
src/engine-settings.js - Image-to-image
src/image-to-image.js
Inference Runtime
- Inference server
backend/inference_server.py - Anima engine
backend/anima_pipeline.py - FLUX.1 engine
backend/flux_pipeline.py - FLUX.2 engine
backend/flux2_pipeline.py - Krea 2 engine
backend/krea2_pipeline.py - Quantized weights
backend/comfy_quant.py - Prompt encoding
backend/prompt_encoding.py
Models and LoRAs
- Model downloader
src/ModelDownloader.jsx - Model locations
backend/model_paths.py - LoRA manager
src/LoraManagerPage.jsx - LoRA cards
src/lora-cards.js
Images and Tools
- Image gallery
backend/gallery.py - Resample chain
src/resample-chain.js - Background removal
backend/background_removal.py
Assistant
- Assistant page
src/AiAssistantPage.jsx - Assistant client
src/assistant-client.js - Assistant providers
src/ai-assistant-providers.js
Other components
- Local user
Connections 22
- Local user → Workspace UI: uses
README.md - Workspace UI → Engine settings: configures
- Workspace UI → Image-to-image: configures
src/App.jsx - Workspace UI → Inference server: submits generation
- Inference server → Anima engine: dispatches to
- Inference server → FLUX.1 engine: dispatches to
- Inference server → FLUX.2 engine: dispatches to
- Inference server → Krea 2 engine: dispatches to
- Anima engine → Quantized weights: loads weights
backend/anima_pipeline.py - Anima engine → Prompt encoding: encodes prompts
backend/anima_pipeline.py - FLUX.1 engine → Quantized weights: loads weights
backend/flux_pipeline.py - FLUX.1 engine → Prompt encoding: encodes prompts
backend/flux_pipeline.py - Local user → Model downloader: manages models
README.md - Model downloader → Model locations: uses locations
- Local user → LoRA manager: manages LoRAs
README.md - LoRA manager → LoRA cards: uses cards
src/LoraManagerPage.jsx - Local user → Image gallery: views results
README.md - Workspace UI → Resample chain: configures
src/App.jsx - Local user → Background removal: uses postprocessing
README.md - Local user → Assistant page: opens assistant
- Assistant page → Assistant client: uses
- Assistant client → Assistant providers: selects provider
src/assistant-client.js