ELEVATE
AI-powered code feedback for Python, running entirely on your machine.
ELEVATE analyzes your Python code as you write it and streams structured feedback — issues, improvements, and complexity notes — directly into a side panel. Everything runs locally via Ollama: no API keys, no data sent to the cloud.
How it works
Open a Python file and ELEVATE gets to work automatically. A native C++ parser reads your code's block structure, a prompt builder turns it into a focused LLM query, and Ollama streams the response back in real time. When your cursor is inside a function or class, analysis scopes to that block rather than the whole file.
The feedback panel shows:
- Summary — a one-line description of what the code does
- Issues — line-numbered errors, warnings, and notes
- Complexity — an assessment of the code's overall complexity
- Improvements — suggested refactors with reasoning
Requirements
- Ollama installed and running locally
- A Python file open in the editor
- VS Code 1.109.0 or later
Quick Start
1. Install Ollama
macOS
brew install ollama
Or download the installer from ollama.com/download.
Linux
curl -fsSL https://ollama.com/install.sh | sh
Windows
Download the installer from ollama.com/download and run it.
2. Start Ollama and pull the default model
ollama serve
On macOS with the Ollama desktop app, it starts automatically in the menu bar — ollama serve is not needed.
Then pull the default model (~2 GB download):
ollama pull llama3.2:3b
Verify everything is working:
ollama list
3. Install ELEVATE
Install from the VS Code Marketplace, then open any .py file. Analysis starts automatically.
4. Trigger analysis manually
Open the Command Palette (Ctrl+Shift+P / Cmd+Shift+P) and run:
ELEVATE: Run Prompt (Ollama)
The feedback panel opens beside your editor and updates on every edit.
Settings
Configure ELEVATE in Settings → Extensions → ELEVATE or in settings.json:
| Setting |
Default |
Description |
elevate.defaultModel |
llama3.2:3b |
Ollama model used for analysis |
elevate.verbosity |
balanced |
concise · balanced · verbose — how much detail to include in feedback |
elevate.teachingStyle |
direct |
direct · socratic · step-by-step — how feedback is phrased |
elevate.customRules |
(empty) |
Extra instructions appended to every analysis (one rule per line) |
elevate.ollamaUrl |
http://localhost:11434 |
Ollama base URL if you're running it on a different port or host |
elevate.concurrency |
1 |
Max concurrent analysis jobs (1–4) |
elevate.editListener.enabled |
true |
Trigger analysis automatically on edit |
elevate.editListener.debounceMs |
350 |
Milliseconds to wait after the last keystroke before analyzing |
elevate.editListener.maxWaitMs |
2500 |
Maximum milliseconds to wait before analyzing during continuous typing |
elevate.cursorTracking.enabled |
true |
Scope analysis to the block under the cursor |
Verbosity
- concise — the 2–3 most important issues, brief explanations
- balanced — a full issues list with moderate detail (default)
- verbose — thorough explanations, all issues, full reasoning
Teaching style
- direct — states issues plainly
- socratic — phrases issues as guiding questions so you discover the problem yourself
- step-by-step — walks through reasoning one numbered step at a time
Custom rules
Add any instructions you want applied to every analysis. Examples:
Always mention time complexity for loops.
Focus on readability over cleverness.
Assume the reader is a beginner.
Commands
| Command |
Description |
ELEVATE: Run Prompt (Ollama) |
Manually trigger analysis on the active file |
ELEVATE: Cancel Job |
Cancel the in-progress analysis |
ELEVATE: Open Response Panel |
Bring the feedback panel back into view |
ELEVATE: Backend Status |
Show the extension's internal job status |
Troubleshooting
Panel doesn't appear after opening a Python file
- Check that Ollama is running: open a terminal and run
ollama list
- Confirm the model is installed:
ollama pull llama3.2:3b
- Open the VS Code Output panel and select ELEVATE from the dropdown for logs
"Ollama not responding" or connection errors
- Make sure
ollama serve is running (or the Ollama desktop app is open on macOS)
- If you changed the port, update
elevate.ollamaUrl in settings
Using a different model
Pull the model first, then update the setting:
ollama pull qwen2.5-coder:7b
Set elevate.defaultModel to qwen2.5-coder:7b in settings. Larger models produce better feedback but are slower.
Analysis seems slow
- Smaller models (
llama3.2:3b, qwen2.5-coder:1.5b) respond much faster
- Lower
elevate.verbosity to concise to reduce output length
- Raise
elevate.editListener.debounceMs to reduce how often analysis fires while typing
Contributing
See CONTRIBUTING.md for build instructions, C++ native component setup, and the test suite.
