Introduction
Jangka is a self-hosted workspace for writing BRD, PRD and TSD documents with LLM assistance.
Generic AI chat turns one prompt into a long, polished document full of invented assumptions. Jangka works the other way around: the assistant interviews you, shows which template sections are still missing information, and drafts each section into a standardized template. Sections it has no information for stay empty instead of being made up.
How it works
- Interview. Pick a template and answer the assistant's questions about what the template needs.
- See the gaps. Every section shows its state: ○ Empty, ◐ Partial or ● Complete.
- Draft and review. Sections are drafted one at a time and streamed live. A rewrite arrives as a diff you accept or discard, so nothing silently overwrites your edits.
Documents live in collections with group-based permissions, link to each other for traceability (BRD → PRD → TSD), and export to DOCX and PDF.
What you need
Jangka installs like Metabase: one Docker image plus Postgres. One instance serves one organization.
| Requirement | Notes |
|---|---|
| Docker | With Compose v2 |
| Postgres 18 | Bundled in the compose file, or bring your own |
| An LLM API key | Anthropic, any OpenAI-compatible endpoint, or Gemini |
The image is published as iqbalr3000/jangka:
docker pull iqbalr3000/jangka:1.0.0-rc.1Where to go next
- Quick start: Jangka running on your machine in a few minutes.
- First-run setup: create the organization and the first admin.
- Install with Docker Compose: what the compose file sets up and how to run it on a server.
- Configuration reference: every environment variable.
- Backups and Upgrading: keeping an instance safe and current.