co-operative bylaw generator

a tool that guides users through creating bylaws for co-operatives using ai-powered generation with ollama models.

features

• multi-step form for collecting co-operative information
• 12 ollama models to choose from (70B, 32B, and 8B parameter sizes)
• jurisdiction-specific bylaw generation (canadian provinces and territories)
• pdf export functionality
• real-time job processing with status updates
• works within vercel free tier (5-minute function timeout)

setup

1. install dependencies

npm install

2. set up upstash redis

for async job processing:

1. go to upstash.com and create a free account
2. create a new redis database (free tier: 10,000 commands/day)
3. copy the REST URL and token

3. configure environment variables

copy the example file:

cp .env.example .env

required environment variables:

# authentication
APP_PASSWORD=your-secure-password

# upstash redis (for job queue)
UPSTASH_REDIS_REST_URL=https://your-redis-instance.upstash.io
UPSTASH_REDIS_REST_TOKEN=your_upstash_token

# ollama api
OLLAMA_URL=https://roo.ai.hypha.coop/api/generate
OLLAMA_USERNAME=your-username
OLLAMA_PASSWORD=your-password

# frontend (for cors)
FRONTEND_URL=http://localhost:3000
REACT_APP_API_URL=

development

option a: traditional development (express + react)

# start both proxy server and react app
npm run dev

or run separately:

# terminal 1: proxy server
npm run proxy

# terminal 2: react app
npm start

app available at http://localhost:3000

option b: vercel local development

# install vercel cli
npm i -g vercel

# run locally
vercel dev

simulates production vercel environment locally.

deployment

vercel (recommended)

1. push to github (make repo public for auto-deploy)

2. connect repository in vercel dashboard

3. set environment variables in vercel:

   • UPSTASH_REDIS_REST_URL
   • UPSTASH_REDIS_REST_TOKEN
   • APP_PASSWORD
   • OLLAMA_URL
   • OLLAMA_USERNAME
   • OLLAMA_PASSWORD
   • FRONTEND_URL (your production url)
   • REACT_APP_API_URL (empty string)

4. deploy:

git push origin main

or manually:

vercel --prod

advantages of vercel

• ✅ automatic https
• ✅ serverless scaling
• ✅ built-in ci/cd
• ✅ 5-minute function timeout (hobby plan)
• ✅ free tier available

available models

large models (70B+) - best quality

• qwen 3.5 35b (23 GB)
• gpt oss 120b (65 GB)
• mistral large (73 GB)
• cogito 70b (42 GB)
• deepseek r1 70b (42 GB)
• llama 3.3 70b (42 GB)
• hermes 3 70b (39 GB)

medium models (balanced)

• qwen 3.5 9b (6 GB) - default
• qwen 3.5 27b (17 GB)
• qwen 3 32b (20 GB)
• gpt oss 20b (13 GB)
• qwen 3 coder (18 GB)

small models (8B) - fastest

• qwen 3.5 2b (2.7 GB)
• qwen 3 8b (5.2 GB)
• hermes 3 8b (4.7 GB)
• gemma 3 (3.3 GB)

architecture

job processing flow

1. submit job (/api/jobs/submit): accepts form data, starts processing
2. poll status (/api/jobs/status): frontend polls for progress
3. get results: results delivered when complete

benefits

• ✅ no timeout issues (processes within 5-minute limit)
• ✅ real-time progress updates
• ✅ fault tolerant (jobs persist in redis)
• ✅ works on vercel free tier

api endpoints

authentication

• POST /api/auth/login - authenticate with APP_PASSWORD
• POST /api/auth/logout - end session
• GET /api/auth/status - check auth status

job processing

• POST /api/jobs/submit - submit bylaw generation job
• GET /api/jobs/status?jobId=xxx - check job status and get results

security

• password-protected access
• secure session cookies
• cors restricted to frontend domain
• encrypted credentials support (optional)

building for production

npm run build

builds optimized production bundle to build/ folder.

license

see LICENSE file for details.