Codexon AI — VS Code Extension
Official IDE extension for the Ammqr NPU and Avicenna Compiler ecosystem.
Features
- Smart Compile — Right-click
model.safetensors or config.json → compile to .codexon
- CODEXON Model Viewer — Custom editor with metadata, 128-bank SRAM heatmap, quant drift chart, PMU trace
- Python Bridge — Centralized
ammqr_master_compiler.py / python -m ammqr integration
- Live NPU Telemetry — Status bar via
npud UDS (/tmp/npu_telemetry.sock) + JSON snapshot fallback
- NPU Chat Panel — Sidebar webview running
python -m ammqr chat
- Generate Runner — Create
run_model.py next to any .codexon
- Ammqr Explorer — Tree of HF models, compiled artifacts, and quant reports
Prerequisites
- Node.js 18+
- VS Code 1.80+
- Python 3 with Ammqr stack (
Compiler/, software/python/)
- Optional:
make -C software runtime-npud for live telemetry
Development
cd vs-extentions/codexon-ai
npm ci
npm run compile
Press F5 in VS Code (Run Extension) with this folder open, or add the repo root as workspace.
Settings
| Key |
Default |
Description |
codexon-ai.pythonPath |
python3 |
Python interpreter |
codexon-ai.compilerRoot |
${workspaceFolder}/Compiler |
Compiler tree |
codexon-ai.pythonPackageRoot |
${workspaceFolder}/software/python |
ammqr package |
codexon-ai.npudSocket |
/tmp/npu_telemetry.sock |
Telemetry UDS |
codexon-ai.telemetryPollMs |
2000 |
Status bar poll interval |
Build & Package
From repository root:
make vscode-ext-build
make vscode-ext-package
Install the generated dist/codexon-ai.vsix via Extensions → Install from VSIX.
Tests
cd vs-extentions/codexon-ai
npm run test:compile
Demo flow
- Open Yildirim workspace root in VS Code
- Right-click
Compiler/Gemma/model.safetensors → Ammqr: Compile to .codexon
- Open
Compiler/build/Gemma_compiled.codexon in the custom viewer
- Check status bar for NPU telemetry (start
software/build/npud if needed)
- Right-click
.codexon → Generate Python Runner
- Open Codexon AI activity bar → NPU Chat
| |
