Stock Analyst Agent

UOMP flagship example Agent (v1.0). Upgraded to a self-serve HTTP API deployable to Digital Ocean — anyone can connect, authorize, and analyze. Analysis dimensions: P&L, sector allocation, concentration, volatility, Sharpe, VaR, Beta, Alpha, correlation matrix, RSI/MACD indicators, rebalance suggestions, scenario analysis.

Full code and sample data are in the reference implementation under examples/stock-analyst.

Prerequisites

Node.js >= 20
pnpm 9 (corepack enable or npm install -g pnpm@9)
Clone uomp-core

1. Build and initialize

pnpm install
pnpm build
pnpm cli init

2. Start the Auth + Guard service

In Terminal 1, start the local service:

pnpm --filter @uomp/server start

Keep this terminal running.

3. Import private data

In Terminal 2, import the risk profile:

pnpm cli import ./examples/stock-analyst/sample-risk.json

Import the holdings CSV (high sensitivity):

pnpm cli import ./examples/stock-analyst/sample-holdings.csv \
  --tag portfolio:holdings \
  --sensitivity high

4. Discover and connect the Agent

pnpm cli discover ./examples/stock-analyst
pnpm cli connect ./examples/stock-analyst

5. Authorize the Agent

Interactive authorization (recommended):

pnpm cli authorize ./examples/stock-analyst

After selecting tags, the CLI prints:

export UOM_TOKEN="<token>"
export UOMP_BASE_URL="http://127.0.0.1:9374"

Scripted authorization:

pnpm cli authorize ./examples/stock-analyst \
  --scope portfolio:holdings profile:risk \
  --output /tmp/uomp.env

6. Run the Agent

source /tmp/uomp.env
node ./examples/stock-analyst/index.js

The Agent reads authorized data, fetches market quotes, and writes a Markdown report to output/.

7. Review sessions and audit logs

pnpm cli sessions -a
pnpm cli audit --limit 20

8. Revoke authorization

pnpm cli revoke <session-id>

9. Self-serve mode (HTTP API)

v1.0 adds an HTTP server mode: the Agent runs as a standalone service with a Web UI and REST API. Users paste their Token in a browser — no CLI needed.

node examples/stock-analyst/server.js
# Visit http://localhost:3080/ and paste your token
# Or use the API:
curl -X POST http://localhost:3080/analyze \
  -H 'Content-Type: application/json' \
  -d '{"token":"...","gateway_url":"..."}'

Deploy to Digital Ocean to make it a public service. See the full README.

10. Remote mode: access via Gateway

If the Agent runs on a different machine or container, use apps/gateway to expose the local Memory Guard as an HTTPS + mTLS entry point. The Agent connects via mTLS, runs aggregate queries first, reads authorized data, and generates reports — then uploads the encrypted Payload and submits a deletion proof.

One-command end-to-end test (covers local + Gateway mode):

./scripts/test-stock-analyst.sh

Analysis dimensions: P&L, sector allocation with target deviation, HHI concentration, annualized volatility, Beta, stop-loss/take-profit signals, drawdown alerts, portfolio volatility, bilingual reports.

See apps/gateway/README.md for full configuration.

Full README → Implementation Design