Remediation Creator Next

Modern Streamlit app to generate, review, and publish Microsoft Intune Endpoint Analytics detection and remediation scripts.

Why this project

Creating Intune remediation scripts manually is slow and repetitive. This app helps you:

• draft detection and remediation PowerShell scripts with LLM support
• validate scripts before upload
• search an existing community script catalog and reuse good patterns
• publish directly to Intune via Microsoft Graph payload

Current UI flow

Tabs in the app:

1. Find Scripts
2. Generate
3. Review
4. Publish

UI screenshots

Find Scripts

Generate

Review

Publish

Find Scripts

• Searches JayRHa/EndpointAnalyticsRemediationScripts via GitHub API
• Shows scored matches (folder + file relevance)
• Select saves a project for review

Generate

• Collapsible Model & Generation Settings block:
  • Provider (Azure OpenAI or OpenAI)
  • Preset / custom model
  • Mode (Detection only or Detection and Remediation)
  • Temperature and max tokens
• Description and optional extra requirements
• Script generation

Review

• Shows selected community project preview (if selected in Find Scripts)
• Use selected scripts in editor to copy community scripts into review editors
• Built-in validation hints and export as .ps1

Publish

• Intune payload preview
• Graph auth connect/disconnect
• Upload to device health scripts endpoint

Key features

• Azure OpenAI and OpenAI provider support
• GPT-5 model support with automatic Responses API routing
• Automatic fallback for models that only allow default temperature
• Community script discovery with selection flow into review
• Script validation (exit code checks, risky command hints)
• Download support for detection.ps1, remediation.ps1, and payload.json

Tech stack

• streamlit==1.54.0
• azure-identity==1.25.2
• openai==2.21.0
• requests==2.32.5

Quick start (recommended)

Because many macOS Python installs are "externally managed" (PEP 668), use a virtual environment:

cd /path/to/Remediation-Creator
python3 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install -r requirements.txt

Create secrets file:

cp .streamlit/secrets.toml.example .streamlit/secrets.toml

Run app:

python -m streamlit run app.py

Open:

http://localhost:8501

Configuration

Model / deployment in UI is primary. secrets.toml values are fallback only.

Example .streamlit/secrets.toml:

AZURE_OPENAI_KEY = "..."
AZURE_OPENAI_ENDPOINT = "https://YOUR-ENDPOINT.cognitiveservices.azure.com"
AZURE_OPENAI_CHATGPT_DEPLOYMENT = "gpt-5.2-chat" # fallback
AZURE_OPENAI_API_VERSION = "2025-04-01-preview"

OPENAI_API_KEY = "" # optional
OPENAI_MODEL = "gpt-5.2-chat" # fallback

APP_REGISTRATION_ID = "..."
GRAPH_SCOPE = "https://graph.microsoft.com/.default"

Model notes

• For gpt-5* models, the app prefers the Responses API automatically.
• If a model rejects non-default temperature, the app retries without custom temperature.
• If UI model field is set, it overrides fallback values from TOML.

Community select workflow

1. Open Find Scripts
2. Search for topic (example: bitlocker)
3. Click Select on a result
4. Open Review
5. Click Use selected scripts in editor (optional)

Security and operations

• Never commit real secrets (.streamlit/secrets.toml is gitignored)
• Rotate keys if exposed
• Review all generated scripts before production use

Troubleshooting

No module named streamlit

Use venv and install requirements:

source .venv/bin/activate
python -m pip install -r requirements.txt

externally-managed-environment

Do not install globally with Homebrew Python. Use .venv.

Unsupported value: temperature

Handled automatically in current app versions; update to latest code if you still see it.

DeploymentNotFound

Your model/deployment name does not exist on that Azure resource. Check your deployment name in Azure and set it in UI Model / deployment.

Project structure

app.py
modules/
  community_search.py
  prompts.py
  utility.py
.streamlit/
  config.toml
  secrets.toml.example
requirements.txt
run.sh
run.ps1

License

Apache 2.0