Open source, AGPL-3.0

The Jyotish practice workspace.

Charts from the Swiss Ephemeris, Lal Kitab with page citations, an AI that must cite the chart, and reports in your own brand. Self-hosted, so your clients stay yours.

Works with

What it does

Built for the practice, not the horoscope.

Seven decisions that set Ankana apart from chart software and from chatbots.

  1. 01

    Charts and a client book

    Swiss Ephemeris charts, dashas and a transit timeline, kept beside consent records, consultations and notes. Charts are immutable: the same birth data always gives the same record.

  2. 02

    AI that cites the chart

    Nine connection types, one claim verifier. Every sentence carries structured claims that a deterministic check compares with the chart before you see them.

  3. 03

    Lal Kitab with the page number

    108 planet and house rules, varshphal, lagna as house 1 and the one-remedy-at-a-time rule, each traced to a page of the source edition.

  4. 04

    Your team, your clients

    Owner, astrologer and assistant roles, a client portal, and sign-in that is built in, with optional OIDC for Google or Keycloak.

  5. 05

    Remedies and reminders

    Caution-aware remedies tied to the planet and house, with reminders over email, WhatsApp and SMS. Hindi throughout.

  6. 06

    Reports in your brand

    Your logo, colours and wording on every PDF, with an advisory line you control. A Today feed shows what needs attention each day.

  7. 07

    Privacy by design

    Client fields are encrypted per practice with AES-256-GCM, with tooling for DPDP requests such as export and erasure.

How it works

The model never has the last word on a fact.

  1. 1

    Birth data

    Goes in with the client's consent recorded.

  2. 2

    Swiss Ephemeris

    Sidereal positions, houses, nakshatras and dashas.

  3. 3

    Chart artifact

    Stored immutable and versioned. Everything downstream cites it.

  4. 4

    Rule engines

    Lal Kitab and Parashari rules fire on the artifact. Each result carries its source.

  5. 5 · The claim verifier

    AI narrator, then a check that calls no model

    Every claim is compared with the chart. A claim that fails is rewritten once. If it still fails, the sentence is replaced by the verified chart fact and flagged as not verified.

  6. 6

    Branded PDF

    Rendered from the same artifact, with your name on it.

House 1Lagna

Lal Kitab

Lal Kitab, with the page number.

Follows the 1952 Goswami edition. Rules are paraphrased, never quoted, and every row carries its volume and page.

  • House 1The lagna is house 1 by default. An Aries option exists.
  • 108 rulesEvery planet in every house, each with a page citation, plus varshphal for the year chart.
  • ExcludedRemedies that harm people or animals, or involve liquor, piercing or medicine doses.
  • In reviewExpert review is in progress. Rows flagged needs_review are not yet checked against the scan.

Accuracy

Honest about what is proven.

Lahiri ayanamsa, mean node, whole-sign houses and Vimshottari dashas from the Moon by default. A suite of reference charts spans hemispheres, eras, daylight-saving cases and a Moon-nakshatra boundary.

What is still pending

Reference values are Swiss Ephemeris outputs marked provisional. Exports from JHora, the independent check, have not arrived. Until they do, the suite proves consistency, not agreement with a second program.

  • Planet longitudes5 arcsec
  • Moon and ascendant10 arcsec
  • Ayanamsa1 arcsec
  • Dasha datesExact
12 reference cases, passingJHora exports pending

Bring your own AI

Your model, your keys, your machine.

Whichever provider you pick, its answers pass through the same claim verifier. In the live evaluation, Claude Haiku passed 115 of 118 cases and a Sonnet sample passed 30 of 30. Read the results.

Claude Code

Built in. Uses your local Claude Code login, no API key needed.

Fully local

Ollama

Local or Ollama Cloud. Open models on your own hardware, nothing leaves the machine.

Gemini

Through a Gemini API key or your local Gemini CLI login.

Codex

Uses your local Codex CLI login.

Anthropic API

Claude models with an API key.

Any OpenAI-style API

OpenAI, OpenRouter, Groq, LiteLLM, LM Studio or vLLM.

Self-host

Your server. Your database.

Docker brings up Postgres 16, Redis and the API. A Bicep template deploys the same stack to Azure, covered in the deploy guide. You need Docker and, for the web app, Node.js 22 with pnpm 10.

Full guide in the docs
git clone https://github.com/mrigankad/Ankana.git
cd Ankana
docker compose up -d --build migrate api worker
pnpm install && pnpm --filter @astro/web-practitioner dev

Status

Where it stands today.

Shipped
  • Swiss Ephemeris charts
  • Lal Kitab and varshphal
  • AI assistant and claim verifier
  • Built-in auth and OIDC
  • Team roles and client portal
  • Reminders
  • Hindi
  • Branded PDF reports
  • Today feed
  • Transit timeline
  • Azure deploy
In progress
  • Planets depth
  • Ashtakavarga
  • Dasha drill-down
  • Sade sati
  • Consult notes

FAQ

Straight answers.

Is this medical, legal or financial advice?

No. Ankana is advisory only. Charts, timing and readings are interpretive guidance for practitioners, and reports carry an advisory line by default. Practitioners remain responsible for what they tell clients.

Where does my client data live?

On your server, in your Postgres database. Secrets, configuration and client data are not source and stay private. With a local model such as Ollama, nothing leaves your machine.

What does AGPL-3.0 mean for me?

You can run, study and change Ankana freely. If you run a modified version for other people over a network, you must publish your changes under the same licence. Using it for your own practice needs nothing more.