Consult is a web application that combines AI with human oversight to process public consultation responses at scale to inform public policy. Once consultation responses are uploaded to the app, the AI identifies themes across the responses using the themefinder package. Users review and finalise these themes — selecting, editing, or creating new ones — before AI assigns the finalised themes to individual responses. The results are presented in a dashboard for users to analyse and draw insights from.
The repository is split into a Django REST backend (backend/), an Astro and Svelte frontend (frontend/), AI processing pipelines that run on AWS Batch (pipeline-sign-off/, pipeline-mapping/), Lambda functions that sync pipeline results to the database (lambda/), and Terraform infrastructure (terraform/).
Important
Incubation Project: This project is an incubation project; as such, we don't recommend using this for critical use cases yet. We are currently in a research stage, trialling the tool for case studies across the Civil Service. If you are a civil servant and wish to take part in our research stage, please contact us at i-dot-ai-enquiries@cabinetoffice.gov.uk.
Installation instructions assume using a Mac with Homebrew.
- Docker Desktop
- uv (instructions)
- nvm (instructions)
- GraphViz (
brew install graphviz), used for generating database diagrams - pre-commit (
brew install pre-commit) - Postegres(optional) (
brew install postgresql) if you are gettingpsycopg2error duringmake setup
Make sure that you have coreutils installed:
brew install coreutils
Also note that you will need to add a personal access token on github with read:packages access and then use this to log in on the command line before running the end-to-end tests:
echo $PASSWORD | docker login ghcr.io -u <username> --password-stdin
git clone git@github.com:i-dot-ai/consult.git
cd consult
make install
make setup
The make install command installs the correct Python and Node versions and all dependencies. And make setup creates .env files from templates and sets up the database with dummy data and an admin user (email@example.com / admin).
make serve
This starts the backend (API server + RQ workers) at http://localhost:8000 and the frontend (Astro dev server) at http://localhost:3000.
You can also run them separately with make backend and make frontend.
To generate new migrations after changing models:
make migrations
To apply migrations:
make migrate
Running make migrate also regenerates the entity-relationship diagram at docs/erd.png (requires graphviz). The current schema:
Run backend tests:
make test-backend
Run frontend tests:
make test-frontend
Run end-to-end tests:
docker compose up -d postgres # postgres must be running already
make test-end-to-end
If you are getting error while running e2e that the frontend is failing to start during the docker spin up its likely because of the timeout module that is missing and you will need to run
brew install coreutilsThe scripts/ directory contains CLI tools for preparing a consultation's
data for the ThemeFinder pipeline:
# Generate an opinionated Q.U. workbook template with live in-sheet validation:
make build-consultation-template
# Validate a Q.U. workbook against response data, build the ThemeFinder
# input layout, and upload it to S3:
make setup-consultation name=my_consultationSee scripts/README.md for the full pipeline
walkthrough and scripts/setup_consultation_checks.md
for the list of validation rules.
This project includes VSCode configuration files to ensure consistent development experience:
.vscode/settings.json- Workspace settings for formatting, linting, and language support.vscode/extensions.json- Recommended extensions for the project
When you open the project in VSCode, you'll be prompted to install recommended extensions. These include:
- Python - Python language support with uv integration
- Ruff - Python linter and formatter
- ESLint - JavaScript/TypeScript linter
- Prettier - JavaScript/TypeScript code formatter
- Astro - Astro framework support
- Svelte - Svelte framework support
- Tailwind CSS IntelliSense - Tailwind CSS tooling
The workspace settings are configured to:
- Format code on save (using appropriate formatter per language)
- Auto-fix ESLint issues on save
- Enable TypeScript support in Svelte files
You can override these settings in your User Settings if you prefer different personal configurations. See the VSCode settings documentation for more information on the settings hierarchy.