Skip to content

Latest commit

 

History

History
139 lines (104 loc) · 4.54 KB

File metadata and controls

139 lines (104 loc) · 4.54 KB

HED-BOT Deployment Guide

This document helps you choose the right deployment option for your use case.

Quick Decision Matrix

Use Case Deployment Option Documentation
Production server (e.g., hedtools.ucsd.edu) Production Docker deploy/README.md
Cloud hosting (Render, Railway, Fly.io) Production Docker deploy/README.md
Local development Claude Platform on AWS + Local Python README.md

All deployments use Anthropic Claude large language models (LLMs) served via the Claude Platform on Amazon Web Services (AWS), an Anthropic-operated Messages API billed through the AWS Marketplace (not Amazon Bedrock). See docs/deployment/claude-platform-aws.md for credential setup. Local Ollama-based inference is no longer supported.

Deployment Options

Option 1: Production Deployment (Recommended for Public Access)

Best for:

  • Production servers (like hedtools.ucsd.edu)
  • Cloud hosting platforms
  • Public-facing deployments

Features:

  • No graphics processing unit (GPU) required (uses the Claude Platform on AWS)
  • Optimized 806MB Docker image
  • API key authentication and audit logging
  • CI/CD with GitHub Actions
  • Auto-deployment with hourly updates
  • Cloudflare Worker proxy (optional caching)
  • OWASP Top 10 compliant security

Setup:

# See complete guide:
cat deploy/README.md

# Quick start:
python scripts/generate_api_key.py
./deploy/deploy.sh prod

Documentation:

Cost:

  • Server hosting: FREE (use existing server) or $5-25/month (cloud)
  • LLM API (Claude Platform on AWS): usage-based, billed through your AWS account
  • Total: hosting cost plus AWS Marketplace LLM usage

Option 2: Local Development

Best for:

  • Quick testing and development
  • Laptop/desktop development (no GPU needed)
  • Testing against the Claude Platform on AWS

Features:

  • No Docker required
  • Fast iteration (no container rebuilds)
  • Uses the Claude Platform on AWS (cloud LLM)

Setup: See main README.md and docs/deployment/claude-platform-aws.md for the required ANTHROPIC_API_KEY, ANTHROPIC_BASE_URL, and ANTHROPIC_WORKSPACE_ID variables.

Cost:

  • LLM API (Claude Platform on AWS): pay-per-use through your AWS account

Comparison

Feature Production Deploy Local Dev
GPU Required No No
LLM Provider Claude Platform on AWS Claude Platform on AWS
Docker Single container Optional
Public Access Yes Local only
Security Full (API keys, audit) Basic
Auto-Updates Yes (CI/CD) Manual
Setup Time 15-30 minutes 10-15 minutes
Best For Production Quick development

Migration Paths

From Local Dev -> Production

# 1. Deploy (credentials carry over from your .env)
./deploy/deploy.sh prod

# 2. Add security configuration
python scripts/generate_api_key.py
# Add API_KEYS to .env

From a Legacy OpenRouter or Ollama Setup

# 1. Update .env to use the Claude Platform on AWS
LLM_PROVIDER=anthropic
ANTHROPIC_API_KEY=your-key-here
ANTHROPIC_BASE_URL=https://aws-external-anthropic.us-east-2.api.aws
ANTHROPIC_WORKSPACE_ID=wrkspc_your_workspace_id

# 2. Remove obsolete variables
# OPENROUTER_API_KEY, LLM_PROVIDER_PREFERENCE, and the *_PROVIDER
# variables are no longer read; legacy LLM_PROVIDER values
# (openrouter, ollama) are coerced to anthropic with a warning.

# 3. Redeploy
./deploy/deploy.sh prod

Need Help?


Last Updated: August 18, 2026