classifai-platform

Complex classifications made easy. Open-source, AI-assisted, and user-friendly. Built for ISCO, ISIC, COICOP, and beyond.

classifai-platform is the front door for the classifAI ecosystem: documentation, local development orchestration, and integration guidance for running classifAI with the optional taxomind AI service.

AI is optional. classifAI runs as a full labeling platform without taxomind.

Components

• classifAI: labeling UI + database + workflows
• taxomind: optional AI classification service (HTTP API)

Links

• Website: https://rowsquared.com/classifai/
• classifAI repo: https://github.com/rowsquared/classifAI
• taxomind repo: https://github.com/rowsquared/taxomind

Why classifAI

Faster & Cheaper

Get instant preliminary results from AI, then let your team verify and correct the suggestions. Time and cost drop as the model learns from corrections.

Improved Quality

Review and correct AI suggestions to continuously improve quality. Use review and double-annotation workflows to create gold standards.

Simple Process Management

Track progress, assign work, and manage workflows in one place. Custom filters, flexible assignments, and built-in commenting keep teams aligned.

Solid AI Integration

AI predictions can be wrong. classifAI (via taxomind) keeps humans in control, so teams review every prediction and teach the model over time.

Architecture summary

+-----------+        +-------------------+        +------------------------+
| Web users | <----> | classifAI UI/API  | -----> | taxomind API (optional) |
+-----------+        +-------------------+        +------------------------+
                           |
                           v
                      +---------+
                      | Postgres|
                      +---------+

• classifAI is the system of record (users, projects, annotations, audits).
• taxomind provides optional AI label suggestions over an HTTP API.
• classifAI authenticates to taxomind with a Bearer token.

Quickstart (Docker Compose)

1. Clone and enter the repo:

   git clone https://github.com/rowsquared/classifai-platform.git
   cd classifai-platform

2. Create your local env file:

   cp examples/.env.example .env

3. Start classifAI (default, no AI):

   docker compose -f dev/docker-compose.yml up -d

4. Open classifAI:
   • http://localhost:3000

Run without AI

• Use the default compose command (no profile).
• Leave taxomind stopped; classifAI still works in human-only mode.
• If you see AI connection errors, remove or blank AI_LABELING_API_URL and AI_LABELING_API_KEY in .env.

Enable AI suggestions

1. Ensure the integration values are set in .env:
   • AI_LABELING_API_URL=http://taxomind:8000
   • AI_LABELING_API_KEY=change-me
2. Start with the AI profile:

   docker compose -f dev/docker-compose.yml --profile ai up -d

Manual setup (advanced)

• Run classifAI from its repository (UI + API + DB).
• Run taxomind from its repository if you want AI suggestions.
• Point classifAI to taxomind with AI_LABELING_API_URL and AI_LABELING_API_KEY.

Configuration overview

• .env at the repo root drives dev/docker-compose.yml.
• Common classifAI settings: NEXTAUTH_SECRET, DEFAULT_ADMIN_EMAIL, DEFAULT_ADMIN_PASSWORD.
• AI integration settings: AI_LABELING_API_URL, AI_LABELING_API_KEY.
• For full configuration options, see the classifAI and taxomind docs.

Documentation

• docs/architecture.md - components, ports, and integration boundaries
• docs/local-dev.md - local dev workflow and troubleshooting
• docs/deployment.md - production patterns and scaling
• docs/integration.md - classifAI to taxomind configuration

Support

Open an issue in the relevant repository:

• UI/workflows: classifAI repo
• AI service: taxomind repo
• Stack/docs: this repository (classifai-platform)

License

MIT. See LICENSE.