Installation¶
Get Daystrom running in under 5 minutes. Choose between npm (recommended) or Docker.
Prerequisites¶
Before you begin, make sure you have:
- Node.js 20+ (LTS recommended) — check with
node -v - Prisma AIRS access — scan API key + management API OAuth2 credentials
- LLM provider credentials — at least one supported provider (see options)
Install from npm¶
Verify the installation:
Set Up Credentials¶
Daystrom needs credentials for both the LLM provider and AIRS APIs. The fastest way is an .env file or shell exports.
All five credential values are required for a functional run. The LLM key depends on your chosen provider — see the provider table for alternatives to ANTHROPIC_API_KEY.
Docker¶
No Node.js required — just Docker. The multi-arch image supports both amd64 (Intel) and arm64 (Apple Silicon, Graviton).
First, create a .env file with your credentials:
ANTHROPIC_API_KEY=sk-ant-...
PANW_AI_SEC_API_KEY=your-scan-api-key
PANW_MGMT_CLIENT_ID=your-client-id
PANW_MGMT_CLIENT_SECRET=your-client-secret
PANW_MGMT_TSG_ID=your-tsg-id
Then run Daystrom:
docker run --rm --env-file .env \
-v ~/.daystrom:/root/.daystrom \
ghcr.io/cdot65/daystrom generate \
--profile my-security-profile \
--topic "Block phishing attempts" \
--intent block
The -v mount persists run state and learnings between containers.
Shell alias
Add to your .bashrc / .zshrc for convenience:
alias daystrom='docker run --rm --env-file .env -v ~/.daystrom:/root/.daystrom ghcr.io/cdot65/daystrom'
daystrom generate, daystrom list, etc.
Where Data Lives¶
Daystrom stores everything under ~/.daystrom/:
| Path | What's in it |
|---|---|
~/.daystrom/config.json |
Your persistent settings |
~/.daystrom/runs/ |
Saved run states (one JSON per run) |
~/.daystrom/memory/ |
Cross-run learnings (one JSON per topic category) |
On Windows, ~ resolves to %USERPROFILE% (typically C:\Users\<username>).
Install from Source¶
For development or contributing:
Requires pnpm >= 8 (corepack enable to install).
Verify everything works: