AI Complete Guide

AI Guide — Chart Readings, Teaching, and Predictions

OpenJyotish integrates with local LLMs (Ollama / LM Studio) for Vedic astrology readings, interactive teaching, and predictions. All computation happens locally — your chart data never leaves your machine.

Setup

1. Install an AI Server

LM Studio (recommended):

• Download from https://lmstudio.ai
• Load any model (Gemma, Llama, Mistral, Qwen...)
• Server mode → Start Server

Ollama:

ollama serve
ollama pull llama3.2    # or mistral, qwen2.5:14b, gemma3:12b

OpenJyotish auto-detects both. No config needed.

2. Test Connection

In the GUI: AI and Learn → Settings tab:

1. Select provider (LM Studio / Ollama)
2. Enter model name (auto-detected on first check)
3. Click Check Provider
4. Green = connected. All AI features unlock.

3. Build Vector Database (Recommended)

The vector DB enables textbook-backed readings. Without it, the AI works but lacks textbook authority.

In Settings tab:

1. Click Build Vector DB (after provider is verified)
2. Wait 5-10 minutes (shows live progress per textbook)
3. Green "READY" message when complete

This chunks 16 Vedic textbooks (1.96M chars) and indexes them for semantic search.

AI Features

OpenJyotish has three AI modes. All require a chart to be computed first.

Mode 1: Chart Interpretation (jhora ai)

Generates a full Vedic reading of your chart. The AI receives:

• All 9 planets with signs, houses, nakshatras, dignities
• Shadbala, Bhava Bala, Vimsopaka Bala strengths
• All detected yogas with planet lists
• Current dasa periods with upcoming transitions
• Current transits with SAV scores
• Ashtakavarga bindus per sign
• Upagrahas, special lagnas, chara karakas
• KP sub-lords, marana karaka, vaiseshikamsas
• Textbook citations via FTS5 or vector search

CLI:

# Full reading
jhora ai "1973-03-13 13:55 +0100 45.41 11.88"

# Professional style (more detail)
jhora ai --style professional "birthdata"

# Career focus
jhora ai --topic career "birthdata"

# Relationship focus
jhora ai --topic relationship "birthdata"

# Smaller context for weaker GPUs
jhora ai --context 2048 "birthdata"

Topics: general, relationship, career, health, spirituality, children, finance, mundane

GUI: AI and Learn → AI Chat tab. Select topic, click "Interpret Chart".

Example output:

***Vedic Astrological Analysis (Parasara School)***

**Lagna and Core Personality**
Your Ascendant (Lagna) is in Libra (Li), ruled by Venus (Ve). This places
a foundational emphasis on balance, harmony, and relationship dynamics.
The strongest planetary strength (Shadbala) belongs to Jupiter (530 virupas),
indicating that wisdom and moral guidance are central themes...

**Career and Professional Life**
The 10th House (Cn) relates to career. Multiple Dhana Yogas, including
Mercury (L9 in H7), indicate professional success linked to partnerships.
The Raja Yoga: Kendra lord Venus and kona lord Mars in conjunction suggests
high achievement through creative action...

Mode 2: AI Q&A (jhora ai --mode ask)

Ask specific questions about your chart. The AI answers with references to your computed data.

# Any question about your chart
jhora ai --mode ask -q "When will I get married?" "birthdata"
jhora ai --mode ask -q "What career suits me?" "birthdata"
jhora ai --mode ask -q "Explain my Saturn placement" "birthdata"
jhora ai --mode ask -q "What do my yogas mean for wealth?" "birthdata"

GUI: AI Chat tab → type in the question box → click "Ask"

Mode 3: Remedial Measures (jhora ai --mode remedies)

Suggest Vedic remedies (upayas) based on chart weaknesses.

jhora ai --mode remedies "birthdata"

Output includes:

• Afflicted planets and their effects
• Recommended gemstones (with metal, weight, finger, day)
• Mantras (with count and timing)
• Rituals or charitable acts (daan)

AI Teacher (jhora teach)

Interactive Vedic astrology instructor. Ask any question about astrology theory, chart interpretation, or prediction techniques.

# General questions (no chart needed)
jhora teach "What is Vimsopaka Bala?"
jhora teach "How do I predict career from a chart?"
jhora teach "Explain the difference between Shadbala and Bhava Bala"

# With your chart — contextual learning
jhora teach "Explain my 7th house" --chart "birthdata"
jhora teach "What does Jupiter in my chart indicate?" --chart "birthdata"
jhora teach "How do I predict marriage timing?" --chart "birthdata"

GUI: AI and Learn → AI Teacher tab. Type question, click "Ask Guru".

The Teacher searches all 16 textbooks for relevant passages and explains concepts step by step with references.

JSON API for AI Agents

For programmatic use, get all computed data as structured JSON:

jhora analyze "birthdata"
# → 10KB JSON, 16 sections, ready for AI tool-calling

import json, subprocess
data = json.loads(subprocess.run(
    ["jhora", "analyze", birthdata], capture_output=True, text=True
).stdout)

# Access specific data
current_md = next(md for md in data["dasa"]["mahadashas"] if md["current"])
strongest_planet = max(data["shadbala"].items(), key=lambda x: x[1]["total_virupas"])

See CLI Reference for all 27 commands.

Tips for Better Readings

1. Use a good model — 7B+ models (Llama 3.2, Mistral, Gemma) produce professional readings. 3B models work but may be less detailed.

2. Build the vector DB — readings gain textbook authority. Without it, the AI relies only on its training data.

3. Choose the right topic — --topic career focuses the reading on 10H, Sun, Saturn, rajayogas. --topic relationship focuses on 7H, Venus.

4. Use --style professional for the most detail. --style concise for quick overview.

5. Ask follow-up questions — after a reading, use --mode ask to drill deeper.

6. Check the data — all AI claims reference computed data. If something seems off, verify with jhora chart "birthdata".

7. LM Studio preferred — embedding generation is faster and more reliable than Ollama for the vector DB build.

Troubleshooting

Problem	Solution
"AI not connected"	Check server is running. Settings → Check Provider
Empty reading	Model may be too small. Try 7B+ model.
Garbled text	Clear __pycache__ directories and restart
"Embedding failed"	Load nomic-embed-text model in your AI server
Build hangs	Use batch_size=5, throttle=1s. Check LM Studio logs
PC freezes during build	The build runs in background. Wait for progress updates.