SafeGraph AI

SafeGraph AI is a local AI coding assistant project built around an AWS Bedrock-powered VS Code extension. The repository also keeps earlier Bedrock chatbot prototypes for web and Streamlit testing.

Current Status

• Main product: safegraph-ai-vscode/
• Current extension version: 0.8.1
• Bedrock region default: ap-southeast-1
• Authentication: Amazon Bedrock API key from VS Code SecretStorage, environment variables, or local .env
• Packaged builds: .vsix files are stored under safegraph-ai-vscode/

Repository Layout

.
├── safegraph-ai-vscode/       # VS Code extension source and packaged VSIX builds
├── chatbot-web/               # Flask web chatbot prototype using Bedrock API keys
├── src/                       # Streamlit app and credential setup scripts
├── docs/                      # Setup, credential, and usage notes
├── config/                    # Python requirements and env examples
├── examples/                  # Bedrock usage examples
└── README.md                  # This file

VS Code Extension

The extension provides:

• Sidebar chat view inside VS Code
• Bedrock chat integration
• Local repository context/RAG
• Inline edit command with Cmd+K on macOS or Ctrl+K on Windows/Linux
• Accept/reject inline edits with Cmd/Ctrl+Enter and Esc
• Safe command auto-run setting for agent workflows

Install From VSIX

Use the latest package in safegraph-ai-vscode/, for example:

code --install-extension safegraph-ai-vscode/safegraph-ai-0.8.1.vsix --force

Or in VS Code:

Extensions: Install from VSIX...

Develop The Extension

cd safegraph-ai-vscode
npm install
npm run build

Open safegraph-ai-vscode/ in VS Code and press F5 to launch an Extension Development Host.

Package A New VSIX

cd safegraph-ai-vscode
npm run package

Bedrock API Key Configuration

The extension checks for a Bedrock API key in this order:

1. Environment variables: AWS_BEARER_TOKEN_BEDROCK, API_KEY, BEDROCK_API_KEY, AWS_BEDROCK_API_KEY
2. Workspace .env files
3. VS Code SecretStorage via Safegraph AI: Set Bedrock API Key

Recommended local .env format:

API_KEY="bedrock-api-key-..."
ARN="arn:aws:bedrock:ap-southeast-1:ACCOUNT_ID:application-inference-profile/PROFILE_ID"

Do not commit real .env files or real API keys.

Web Chatbot Prototype

The Flask prototype lives in chatbot-web/.

cd chatbot-web
pip install -r requirements.txt
python app.py

Default URL:

http://localhost:5001

It reads credentials from the repo root .env and chatbot-web/.env.

Streamlit Prototype

The older Streamlit app is still available under src/.

pip install -r config/requirements.txt
streamlit run src/app.py

Credential helper:

python src/setup_credentials.py

Security Notes

• Real API keys must stay out of Git.
• .env files are ignored and should remain local.
• Prefer VS Code SecretStorage for the extension.
• Rotate/revoke any key that was shared, logged, or committed.
• After rewriting history, avoid merging old remote branches back into clean branches.

Useful Docs

• SafeGraph VS Code install guide
• Renew API key guide
• AWS credential guide
• Project docs index

License

See LICENSE.