← Back to Help

Chat and INSIGHT

What it does

An assistant that answers on top of precise, pre-computed data: planetary positions, aspects, orbs, houses, and the lunar calendar are calculated by deterministic algorithms from real astronomical ephemerides (data accuracy). The model never computes the sky itself — it interprets the ready numbers under clear rules.

Features

Saved profiles in chat

If you have several saved profiles, call the profile exactly as it is saved — chat builds the chart for that person without switching your active profile:

Show Maria's natal chart
Show transits for Client A
Name the Sun, Moon, and Ascendant for profile Roman

• Without a name, chat uses the active profile.
• If the assistant can't find a profile, check the exact name: ask "Show my profiles" or open Settings → Profiles.

To compare two saved profiles, use the /synastry page — both slots pick from your saved profiles.

Managing profiles in chat

Chat can also manage your profiles with plain phrases:

Create profile Maria: 1991-03-12, 08:30, Moscow
Update Maria's birth time to 09:15
Rename profile Masha to Maria
Make profile Maria active
Delete profile Test

• To create a profile, date, time (if known), and place are enough — known cities resolve without manual coordinates.
• If data is incomplete, the profile is stored as a draft and not activated — the assistant tells you what's missing.
• Your own profile cannot be deleted — it's your account's data.
• After each action the assistant confirms the result; if something failed, it says so plainly. You can always verify with "Show my profiles".

Profile groups in chat

Organize profiles into groups (family, friends, clients...) right from chat:

Show my groups
Show profiles in group Friends
Move profile Maria to group Clients
Create group Archive
Rename group Clients to Consulting
Delete group Archive

• A group is just a label: moving a profile never changes its birth data.
• Only an empty group can be deleted — move or delete its profiles first.
• The owner group and your own profile are protected.

If you read charts for several people — family, friends, clients — this works as a roster: keep the pool organized in groups ("Clients", "Family"), build charts by name mid-conversation, switch the active person with one phrase — without leaving the chat. For synastry of any saved pair, use /synastry.

Personas — right in chat

The response style switches without opening settings too — just ask:

Set the Jung persona
Add Marcus Aurelius as well
What personas are there?
Remove the personas

• One or two personas at a time; "synthesis" personas (a tradition's collective voice) go alone.
• Picking a persona applies its style preset right away (tone, humor, depth…) — exactly like choosing it on the Style tab.
• Takes effect from the next message; "Remove the personas" returns persona-free mode — your manual style settings are kept.
• Works on the site and in the Telegram bot.
• How to choose a lens, pair two personas and tune the style: the Personas and lenses guide.

Model and effort — right in chat

The chat model and its reasoning effort switch by phrase too:

What models can I use?
Set the model to GPT 5.4 Nano
Set effort to high
Turn on style polish

• The assistant offers and accepts only models available on your plan — it will tell you the exact names.
• Effort options depend on the selected model (each has its own set; "optimal" auto-picks per request).
• Style polish (two-pass styling) works in the web chat while a persona or manual style is active — the assistant reports honestly whether it took effect.
• Several settings can change in one phrase: "Turn on style polish and set effort to high" — they apply together; if any value is invalid, nothing is applied and the assistant lists every problem.
• General settings too: "Set the Koch house system and answer in English". The house system, the lunar day by moonrise and names in answers apply from the next message; the answer language (any language the site offers — see Settings) applies after a page reload on the site and from the next message in Telegram (in a chat linked to your account; an unlinked chat does not switch its language from a message yet). The assistant tells you exactly when each setting takes effect.
• Everything else applies from the next message; works on the site and in the Telegram bot. The one exception is the answer language on the site — it applies after a page reload (see above).

Style — right in chat

Every property of the Style tab switches by phrase too:

Make the tone direct and the humor sharper
Address me informally, feminine grammar
Add a goal: learn to read transits
Exclude the topic: politics
Reset the style to the persona's
Turn off the micro-questions
What style settings are applied?

• Tone, empathy, humor, psychology/philosophy depth, interpretive focus, motivation, spiritualness, formality, address — any subset in one phrase; an invalid value refuses the whole request and lists the valid ones.
• With a persona active, a changed property becomes your override on top of the preset (italicized on the Style tab); "reset the style to the persona's" restores the preset.
• Up to 5 goals and 10 excluded topics (exactly what the assistant reads); removals are always allowed.
• The status is honest: it shows both the stored value and what actually reaches the answer (e.g. "sharp" humor is not applied while a persona is active).
• Micro-questions are a web-chat feature; on Telegram the assistant says so honestly.
• With "sync to cloud" disabled, style changes from chat are not written — the assistant warns you.
• Everything applies from the next message; works on the site and in the Telegram bot.

Insight & vision — right in chat

The Insight/Discuss model (the Insight buttons on pages) and the vision model (photo analysis in the Telegram bot) — and their efforts — can be switched by phrase too:

Which Insight models are available to me?
Set the Insight model to Gemini 3 Flash
Insight effort high
Switch the vision model to GPT 5.2
What settings are applied right now?

• These are separate from the chat model — the same settings as Settings → Models.
• The vision setting applies to photo analysis in the Telegram bot linked to your account.
• In the web chat these settings need a signed-in account (anonymously they would have no effect — the assistant refuses honestly).
• Ask for any setting's status in one go: "what settings are applied right now?" — the answer honestly distinguishes stored vs actually effective.

Telegram

The Telegram bot understands the same profile commands — use explicit action verbs:

Create profile Maria: 1991-03-12, 08:30, Moscow
Show profiles in group Clients
Move profile Maria to group Clients
Set the Nietzsche persona
Remove the personas

If you asked for an action plus a list and the list isn't visible, ask for the list as a separate message. Bot linking lives in Accounts.

How to ask the assistant well

The question sets the precision of the answer. Build it from four blocks — not all are required, but the more you give, the sharper the reply:

1. Object — who or what: "for me", "for profile Maria", "the current sky", "group Clients".
2. Data layer — what to look at: natal chart, transits, Moon and nakshatras, Human Design, a Tarot / rune / I-Ching spread.
3. Horizon — for when: "now", "today", "the next 48 hours", "this week".
4. Format — how to deliver it: "facts first, then interpretation", "three main themes", "no technical overload", "like a professional astrologer".

Vague → better

The "facts first" trick. If an answer feels too generic or "poetic", ask: "First list the data — aspects, orbs, the Moon, and key transits — then give the interpretation and a practical takeaway". The assistant works on top of exact numbers and will show them if you ask.

Strong combined questions

• Everything about yourself at once: "Show my natal chart, current transits to it, and forming aspects for the next week".
• The current sky, precisely: "Show forming aspects for the next 24 hours with an orb up to 2°", "Which transits affect me most today?".
• By profile name: "Build the natal chart for profile Maria and highlight the most important current transits".
• The Moon across traditions: "Break down today's Moon across all systems: Western, Jyotish, Chinese and Arabic mansions — where do they agree and differ?".
• Oracle + sky synthesis: "Synthesize this hexagram with my current transits", "Connect the drawn rune with the current lunar situation", "Overlay the spread onto my transits and pull out 3 main themes".
• Depth on request: "Give a detailed, neutral reading — like a professional astrologer".

Assistant settings

• Models (chat / Insight / vision) — Settings → Models. The model only affects interpretation style and depth — the calculations are always the same.
• Response style (personas, tone, address, humor, depth) — Settings → Style; a persona can also be switched right in chat: "Set the Jung persona" / "Remove the personas".

Deepen one reading without drawing again

Continue the displayed result: name the question, spread, positions, cards and orientations. For runes, include the alphabet; for I-Ching, include original/related hexagrams and changing lines.

For example: “Continue this same reading. Do not draw new cards. Explain the relationship between the second and third positions, then offer one reflection question.” You can change persona or request another interpretation while retaining the symbols.

Regenerating an answer does not guarantee the same draw: the request can run again. If a long chat no longer includes the source result in context, supply it explicitly. Chat history, divination history and the collection are different stores. Finish editing before PNG; export can start another generation. Example journey.

See also

• Settings
• Natal chart
• Synastry
• Artifacts