Using ollama with any model produces the same results for auto-complete, quiet failures. Meanwhile the Llama-Coder extension has no trouble and the CodeGPT Chat works fine. Seems like a CodeGPT-specific problem I haven't figured out how to resolve.
Create a comment and CMD+SHIFT+I or simply wait for an auto-complete to fail
Expected Behavior:
Auto-completion of any kind
Actual Behaviour:
Silent failures
Environment:
MacOS Sonoma 14.4.1
CPU: M2 Pro
RAM: 32GB
VSCode: 1.89.1 (Universal)
Commit: dc96b837cf6bb4af9cd736aa3af08cf8279f7685
Date: 2024-05-07T05:14:24.611Z
Electron: 28.2.8
ElectronBuildId: 27744544
Chromium: 120.0.6099.291
Node.js: 18.18.2 - I noticed a recommendation for Node 20, but this is what VSCode ships with for Mac
V8: 12.0.267.19-electron.0
OS: Darwin arm64 23.4.0
Using ollama with any model produces the same results for auto-complete, quiet failures. Meanwhile the Llama-Coder extension has no trouble and the CodeGPT Chat works fine. Seems like a CodeGPT-specific problem I haven't figured out how to resolve.
Steps to Reproduce:
Expected Behavior:
Actual Behaviour:
Environment: MacOS Sonoma 14.4.1 CPU: M2 Pro RAM: 32GB
VSCode: 1.89.1 (Universal) Commit: dc96b837cf6bb4af9cd736aa3af08cf8279f7685 Date: 2024-05-07T05:14:24.611Z Electron: 28.2.8 ElectronBuildId: 27744544 Chromium: 120.0.6099.291 Node.js: 18.18.2 - I noticed a recommendation for Node 20, but this is what VSCode ships with for Mac V8: 12.0.267.19-electron.0 OS: Darwin arm64 23.4.0
CodeGPT: 3.3.25 ollama: 0.1.32 Models: codellama, deepseek-coder
Other Logs: From Extension Host after a CMD-SHIFT-I
From VSCode Output > CodeGPT Autocomplete, same repeated message with any key strokes :
[{timestamp}] completion provider = Ollama
From VS Code DevTools Console, when switching context a Python tab: