jngka

BRD/PRD/TSD·self-hosted

Precise specs for product & engineering teams.

Jangka interviews you, shows what's still missing, and drafts BRDs, PRDs and TSDs into your templates. Self-hosted. Your LLM key. Your data.

docker composesh
$ curl -fsSL https://raw.githubusercontent.com/iqbalr3000/jangka/v1.0.0-rc.1/docker/docker-compose.yml -o docker-compose.yml$ echo "MASTER_KEY=$(openssl rand -base64 32)" > .env$ echo "JANGKA_VERSION=1.0.0-rc.1" >> .env$ docker compose up -d
Read the docsView on GitHub
Fig. 1: PRD outline with section status

The approach

A chat window writes the whole document. Then you hunt for what it invented.

Generic AI chat

one prompt→long document→invented assumptions

4.2 Performance

The system shall support 10,000 concurrent users with 99.99% uptime, in line with the existing platform SLA.

3 claims nobody gave it

Jangka

asks first→flags gaps→drafts what it knows

4.2 PerformanceEmpty

Not enough information yet.

Ask me about this

How it works

Interview, check the gaps, review the draft.

  1. 01

    Interview

    The assistant asks for what the template needs before drafting anything.

  2. 02

    See the gaps

    Every section shows its state, so you know what's missing before anyone reviews it.

  3. 03

    Draft and review

    Sections stream in. Rewrites arrive as a diff you accept or discard.

Features

Writing, review and operations in one instance.

Writing

  • Interview-first drafting
  • BRD, PRD and TSD templates, plus your own
  • Completeness check per section
  • Rewrites shown as a diff
  • Mermaid diagrams, edited as code

Collaboration

  • Collections with group permissions
  • Review with comments and approvals
  • BRD → PRD → TSD traceability with impact flags
  • Section history and audit log
  • Full-text search

Operations

  • DOCX and PDF export
  • Anthropic, OpenAI-compatible, Gemini
  • Password or SSO (Google, OIDC)
  • Secrets encrypted at rest, AES-256-GCM
  • No telemetry

Traceability

Change a business requirement. See every spec it touches.

Requirements link by ID from BRD to PRD to TSD. When an upstream requirement changes, the requirements that depend on it are flagged for review.

BRDCheckout v2Approved
BR-003Guests can buy without creating an account.Edited 2h ago
PRDGuest checkoutApproved
FR-012Checkout accepts an email address in place of a login.Needs review: upstream changed↑BR-003
TSDCart serviceApproved
TR-031Anonymous cart sessions keyed by a signed token.↑FR-012

Self-hosted

One Docker image. One Postgres. Nothing else.

Jangka runs inside your network. Requirements go to the LLM provider you configure and nowhere else.

LLM
Bring your own key: Anthropic, OpenAI-compatible or Gemini.
Secrets
Encrypted with your MASTER_KEY.
Telemetry
None.
Tenancy
One instance per organization.
License
MIT.

Quick start

Running in three commands.

Then open http://localhost:3000 and follow the setup wizard.

Full install guide →
terminalsh
# 1. Download the compose filecurl -fsSL https://raw.githubusercontent.com/iqbalr3000/jangka/v1.0.0-rc.1/docker/docker-compose.yml -o docker-compose.yml# 2. Generate the key that encrypts stored secretsecho "MASTER_KEY=$(openssl rand -base64 32)" > .env# 3. Pin the Jangka versionecho "JANGKA_VERSION=1.0.0-rc.1" >> .env# 4. Start Jangka and Postgresdocker compose up -d

FAQ

Questions

Is it free?

Yes. Jangka is MIT licensed and runs on your own infrastructure. The only running cost is what your LLM provider charges for tokens.

Which LLMs does it support?

Anthropic, any OpenAI-compatible API (including self-hosted models behind one), and Google Gemini. Your organization configures its own key.

Does my data leave my server?

Only as requests to the LLM provider you configure. Jangka sends no telemetry and there is no hosted service in between.

What do I need to run it?

Docker and Postgres 18. The compose file starts both.

Can I use my own templates?

Yes. BRD, PRD and TSD templates ship by default. Duplicate one, edit its sections and guidance, and publish it; the completeness check follows your structure.