AI Calls, Texts, Dating and Social Media for The Sims 4

Your sims have things to say. Hand them a phone.

Download on CurseForge
Download on GitHub
Download on Nexus Mods

Your sims' phone, finally interesting.

Every call and text reads your save state before a word gets written. Traits, career, aspirations, family role, who lives with them. Live weather where each sim is standing. Recent milestones, parties you both attended, the in-person chat you had yesterday, group texts you were both in, freeform notes you've written about the sim. All of it lands in the prompt.

So your dad opens with "hey kiddo, how's the hospital?" because he knows you're on the doctor career. Your ex mentions the funeral you both went to last week. Your best friend brings up the storm you're stuck in. And when you tell someone you need space, the mod picks that up — the next reply from them is a boundary conversation, not another cheerful check-in.

A flirty teen texts in lowercase with dramatic slang. An elder writes long, warm paragraphs. A grandparent dotes; an enemy sneers. A parent never opens with "hey man"; a Goofball cracks jokes mid-call. The voice fits.

What it does

Phone Calls

Random sims call you in character — your kid's friend, your boss, your ex. Each voice shaped by who they are.

Text Messages

Check-ins, group drama, gossip. Hit Reply and they keep the thread going — full back-and-forth, in character.

Llamadate

Dating app for your sim. Opt in for inbound matches, or browse Hinge-style profiles and message someone yourself.

Llamagram

Social media for your sims. Post to friends or the world, and real sims from your save comment, like, and follow. Friends post too, and they bring it up when they call.

Interacts with your game

Writes back where it matters. Emotional texts apply moodlets and nudge friendship or romance up and down. New sims land in your relationships panel when you engage.

One mod, every voice.

Same code, totally different writing — the AI adapts to each sim's age, traits, and how they feel about you. Here are real texts from the same save:

Text message from a close friend in casual, candid tone
From a close friend
Casual, candid, easy banter.
Call referencing a real in-game event the player's sim just did
Knows what just happened
References real life events — quit jobs, promotions, breakups.
Call referencing an upcoming funeral both sims are attending
Knows what's coming up
References upcoming funerals, weddings, holidays — naturally, never invented.
Text message from an elder partner — warm, affectionate, full sentences
From a partner
Warm, affectionate, age-appropriate tone.
Text message from an enemy in cold, hostile tone
From an enemy
Cold, dismissive, no warmth.
Phone call from an infant babbling goo goo ga ga
Unpredictable moments
Yes, the baby called. Yes, it was babbling.

Llamagram: social media for your sims (v3.8)

Your sims are online now. Post to your friends or to the whole world, and the sims in your save react: family, friends, exes, and strangers who've never met you. Sometimes a few people comment, sometimes nobody does. Your friends post about their own lives too, and the people you text and call have seen it all.

The Llamafone app menu listing Llamadate, Llamagram Notifications, Llamagram Post, and Send Text
Two new tiles in the app
Llamagram Post to share something, Llamagram Notifications to see what everyone said.
A text box titled New Llamagram post as Francesca, with Next and Cancel buttons
You write the post
Whatever your sim wants to say, in your words. Every household member posts from their own account.
Audience picker asking who can see Francesca's post, with Friends and Public options
Friends or the whole world
Friends-only stays with the people your sim knows. Public reaches strangers and fans too.
Notification: Llamagram, 2 new comments on your post, from Vivian Woodruff and Lucas Leroy, with a running count of likes and comments
Comments trickle in
A few at a time, in batches, with likes and comments climbing as the post spreads. Mom checks in. Your friend relates.
Llamagram Notifications for Francesca: follower count, 3-star celebrity, her latest post's views, likes, and comments, and a list of new comments
Followers, views, likes, fame
Posts that land well grow your following, and once in a long while one blows up. Big accounts draw fans. With Get Famous, posts that truly go viral raise your fame.
Francesca's friends-only post with its stats and two comments from friends
Open any post
See how far it went and every comment. Reply to a comment and they answer back.
Llamagram Notifications for Aksel with 26 followers: comments from his parents, family, and friends on his post about being a dad
Not famous? Your people still show up
Aksel has 26 followers, and his family and friends all chimed in on his dad post.
Vanessa Hansen posted on Aksel's feed about a quiet winter night on the ranch, with Comment as Aksel and Scroll past buttons
Your friends post too
About their own lives: the ranch, the new job, the baby. Comment, or scroll past.
Call from grandmother Gabriela Hansen saying she saw the post about fatherhood and asking how the baby is
"I saw your post"
Posts carry into calls and texts. Gabriela saw Aksel's post, and she's calling about it.
Text from Ingrid Negrete to Ben: three more kids came in the last few hours, there are seven babies in the apartment and 1389 simoleons, so stop posting garbage online and figure out what you're doing
She saw your posts too
What happens when you test birth announcements a few too many times. Seven babies, 1,389 simoleons, and a father who won't stop posting.

Messages move relationships, world context, sim bios, service NPCs (v3.6)

Flirty and warm texts now nudge the game's own friendship and romance bars. Plus player-authored context that shapes every prompt, hired service NPCs that write in the right register, and a fifth AI provider option.

Llamafone Settings menu showing new v3.6 rows: Save notes, Sim bios, Messages affect relationships toggle, and a relationship-impact cap slider
New knobs
Save notes, sim bios, and the relationship-impact toggle all live under Settings alongside the existing controls.
Sim bio input dialog where the player types character notes for a specific sim
Bios you write yourself
Backstory, motivations, secrets, quirks — anything you want the AI to know about who this sim really is. Every call, text, or reply involving them reads from it.
Save-level world context dialog where the player writes free-form notes that shape every AI prompt in this save
World context, save-wide
One field, injected as the top-priority instruction in every prompt. Set the era, the challenge rules, whatever binds the whole save.
Text from a butler NPC written in professional, deferential register referencing household dinner arrangements
Your butler sounds like your butler
Hired service NPCs stay in role — formal for butlers, task-focused for maids, kid-first for nannies, plant-talk for gardeners.
Sim relationship panel showing the romance bar growing after flirty text exchanges
Romance and friendship build where they should
Flirty back-and-forth fills the romance bar; warm banter fills friendship. The game's own relationship bars respond to how the conversations actually go.
A flirty text exchange between two sims building romantic tension through the messages
The flirting is real
Now your sim builds relationships while their phone call gives them butterflies.

Standalone app + Llamadate (v3.5)

Llamafone got its own tile on the phone. And inside: Llamadate. Opt in, browse profiles, message someone, or wait for a match to slide into your DMs.

Llamafone's llama icon as a standalone app tile on The Sims 4 phone home screen, next to other apps
Its own home-screen tile
The llama's on the phone home screen. No more digging through Social.
The Llamafone app screen showing its four tiles: Call, Send Text, Llamadate, and Settings
Everything under one llama
Call, text, date, tweak settings — all in one app on your sim's phone.
Sim picker titled Llamadate showing portraits of unmet sims to potentially message
Swipe through the town
A rotating feed of sims your sim hasn't met yet.
Sim's dating profile card with a short bio at the top and three prompt-and-answer pairs below in Hinge style
Their opening pitch
Bio and three prompts, in the sim's voice.
Llamadate settings for a specific sim: opt-in toggle, gender filter, age filter, and a bio field
Set it up per sim
Opt-in, gender, age, and a bio you write yourself. Each played sim gets their own.
Dahlia Loden's Llamadate profile: a bookworm working in publishing, looking for something real
The bookworm
Publishing lifer looking for the long game.
Felix Hoffman's Llamadate profile: a food-obsessed, outdoorsy, child-free sim
The foodie
Chases the perfect meal like it's a sport.
A reply from a Llamadate match responding to the player's intro
They actually read your bio
Reactions weigh what you wrote against their vibe. Sometimes they're into it. Sometimes they're not.
An inbound cold-outreach text where the sender was introduced through a mutual friend
Friend of a friend
When you share a mutual, the intro comes through them.
Inbound cold-outreach text from a total stranger who saw the player's profile on Llamadate
Or a total stranger
No mutual? Llamadate itself is the intro — with a plausible IRL detail to make it land.

Per-relationship contact preferences (v3.4)

Alice broke up with Chad. She wants him to stop calling her every hour. She doesn't want him gone from the game — just from her phone. Bob (her brother) is still friends with Chad and should keep getting his texts. This is what per-relationship preferences are for: mute, "asked for space," or favorite — scoped to a specific household member's view of a specific contact.

Llamafone Settings panel with a new Manage Contacts row
Reached from Settings
Phone > Social > Llamafone Settings > Manage contacts. Scoped to whichever household sim opened Settings.
Sim picker listing contacts with markers for those that have preferences set
Contact list with markers
Contacts with an existing state or note surface first, tagged in brackets ([paused], [muted], note).
Per-contact action picker with mute, paused, priority, clear, and note options
Per-contact actions
Mute, Ask for space, Favorite, Clear, Edit note, or Wipe. The current active state is marked with ←.
Text input dialog for a freeform note about a contact
Freeform notes
Any context you want the AI to know — "kid's teacher," "asked for space after breakup." Surfaced in every prompt for this pair.
A text message from a contact respecting the player's request for space
Here's what changes
Once paused is set, the AI knows about the boundary. A decent contact respects it; a love-heavy ex might ignore it; a mean one might guilt-trip.

The paused ("asked for space") state is the interesting one. Instead of blocking the contact, it drops their auto-event rate to 20% and tells the AI you asked for space — so their response is driven by who they are.

Group texts (v3.4)

Send one message to 2–4 sims at once. Each recipient replies in their own voice, staggered like real texts arriving over a few minutes. Every reply sees what the others just said — so responses stay distinct instead of a chorus of "same." The dialog title and the reply prompt both list the full group roster, so you always know who you're talking to.

Multi-select sim picker allowing 1 to 4 recipients
Multi-select picker
Same "Send Text" entry as before — now picking 1 sim gives a normal text, picking 2–4 starts a group thread. Checkboxes replace the radio buttons when the picker's in group mode.
Incoming group text with the full roster shown in the dialog title
Roster visible on every message
Title reads Group text with Sarah, Bob, Alice, and Kate. The message body is prefixed with the sender's name so you always see who just spoke.
A different group member's reply arriving in the same thread
Staggered replies from everyone
Each recipient's reply lands after its own randomized delay (same delay range as 1:1 texts). Serial, not parallel — so participant B's prompt includes what participant A just said.
Reply dialog explicitly listing the whole group as the destination
Reply fans out to the whole group
The Reply button opens a dialog that spells out your reply goes to the whole group: Sarah, Bob, Alice, and Kate. Each active participant may respond to the next round.

A briefing call at group creation (using the default model) synthesizes each participant's voice and their web of relationships, then gets cached for every subsequent reply — so per-turn cost stays sane even over long threads. Persisted per-save in GroupTexts.json so quitting mid-thread doesn't lose it.

Now in your sim's phone (v2.0)

No cheat console required for everyday use. Call, text, and tweak settings from the in-game phone the way your sim would.

Phone Social menu showing Settings, Send Text, and other Llamafone options alongside base game items
Phone > Social
Call Someone, Send Text, and Settings live next to the base game's phone options.
A sim with the Call Someone interaction queued in their action bar
Queues like any action
Call Someone slots into your sim's action queue alongside their other plans.
A sim standing outside actually using their phone — the brief phone-tapping animation plays as the interaction runs
Plays the phone animation
Your sim actually pulls out their phone and uses it while the picker opens.
Sim picker dialog titled Who to call with portraits of contacts
Pick a recipient
A real sim picker scoped to your contacts — household + relationships above the friendship threshold.
Text dialog with the contact's portrait and a multi-line text area for composing the message
Type what you want to say
Multi-line text area with the recipient's portrait. the AI crafts the message in their voice from what you wrote.
Llamafone Settings grouped into folders: AI settings, Incoming calls and texts, Texts and replies, Llamagram, Manage contacts, and Dating
Settings, in-game
Everything in folders: AI settings, incoming calls and texts, texts and replies, Llamagram, contacts, and dating. Changes save instantly and write back to llamafone.cfg, keeping your comments.
Llamafone AI settings in game: Provider, Model, Thinking, Max tokens, Language, and Test connection
AI settings, in-game
Switch provider and model, set how hard the AI thinks, and test the connection without leaving the game. Only the API key lives in llamafone.cfg.

It knows what's actually going on.

Every call and text is filtered through what's true in your save right now — not generic AI fill.

Geography aware

If a sim lives in your world, they'll suggest grabbing coffee. If they live in San Myshuno and you're in Willow Creek, they keep it long-distance — no "come over" from across the map.

Weather aware

Live read of the weather where each sim actually is — light rain, blizzard, heatwave, thunderstorm. Dramatic weather comes up in conversation; routine weather stays background. Off-world callers know THEIR climate too, so Sulani friends won't pretend it's snowing during a globally-Winter season.

Life events show up

Sim quit a job, got promoted, had a baby, got married, aged up? The mod catches it automatically and your friends bring it up — "heard you quit, you good?" — within days of it happening.

Knows who's at work

Text your sim's dad while he's on shift at the hospital and he might write back "sorry, at work, catch you later" — not generic small talk. Reads live career state — currently at work, or should be but isn't — plus school schedules for teens and kids, so people who should be busy sound busy.

Service NPCs stay in role

The butler, maid, babysitter, gardener, and repair tech write in the right register. Butlers stay professional and deferential; maids talk supplies and schedules; babysitters check in on the kids; gardeners talk plants and season; repair techs quote parts and timing. The mod reads the household's live service records so a hired NPC calling about tomorrow's schedule lands like they should, not like a random friend.

Parties, weddings, holidays

Sims 4 drops calendar events from the UI the moment they end. Llamafone snapshots them. Text a guest the morning after a house party — "great party last night." Text your ex the day of a funeral you both attended. Love Day and Winterfest come up on the day.

Recent hangouts remembered

When your sim actually chats, flirts, argues, or kisses another sim in-game, the mod logs it. Next time you text or call, the AI knows you saw each other in person — so the opener is "great catching up earlier," not generic small talk. Uses the same Sims 4 relationship bits (Just Chatted, Just Kissed, Social Context) the game itself tracks.

Family roles override traits

Your Outgoing Adventurous dad doesn't text you like a frat bro — he texts like your dad. Parents sound parental, siblings tease, spouses get intimate, grandparents ramble warmly. Traits are interests, not voice.

Friendship tone is law

Friends are warm. Acquaintances are polite. Enemies are openly hostile — no "rooting for you 💕" from someone who hates your guts. The current relationship beats anything from old chat history.

Your notes, your context

Write a freeform note on any contact — "kid's teacher," "asked for space after breakup," "boss's boss," "childhood best friend." Surfaced in every prompt for that pair, weighted heavily, so the AI stays in character on nuance the game itself can't label. Per household member — Alice's notes stay Alice's.

Boundaries carry through

Mute a sim or ask them for space in the Contacts UI, separately for each household member. Every subsequent AI prompt for that pair leads with the boundary, not with warmth. A love-heavy ex ignores it and keeps calling; a decent friend apologizes; a mean one guilt-trips. The drama plays out based on who they are.

Group thread memory

When Alice later texts Sarah 1:1, the AI knows they were both just in a group text with Bob and Kate. The last few turns from that group thread land in the prompt — so 1:1 conversations reference shared context like real friend groups do. Runs both directions, capped to recent threads only.

Social media posts

Post on Llamagram and your friends saw it. Next time one of them texts or calls, they can bring it up — "saw your post, are you okay?" — and they remember what the two of you said in the comments. Works both ways: you can ask about their latest post too. Only the last few in-game days count, and only posts they could actually see, so a friends-only post never reaches a stranger.

What time it is

Every prompt now carries the current in-game clock — 7:15 PM (evening), 9:03 AM (morning), etc. No more "dinner ready by six" from your butler when it's already 7. Times of day sound like real times of day, and future-tense references (tonight, tomorrow, next week) land in the right window.

Remembers how they met

Match with someone on Llamadate today, text them three in-game weeks later, and the AI still knows the origin — that's how they'll bring it up. Mutual-friend intros remember the mutual's name too. Stored on the pair alongside contact preferences, ready for future asymmetric events (who kissed who, who proposed to who).

Commands at a glance

Open the cheat console with Ctrl+Shift+C, type a command, press Enter.

Phone

  • llama.call — incoming call
  • llama.text — incoming text
  • llama.sendtext Bella Goth hey!
  • llama.sendcall Bella Goth I have news
  • llama.reply <message>

Llamagram

  • llama.post hello world — post to friends
  • llama.post public hello world
  • llama.npcpost — a friend posts now
  • llama.inbox — notifications
  • llama.feed — followers & recent posts

System

  • llama.status — setup & all commands
  • llama.auto_events on — timed calls & texts
  • llama.chat <message> — freeform
  • llama.journal — recent history
  • llama.reload — after editing config

Common questions

Why use AI in The Sims 4?

For storytelling. The base game's phone calls and texts are limited dropdown options — useful for gameplay, repetitive for narrative. This mod replaces them with text written specifically for your household by the AI provider you configure (Claude, OpenAI, Gemini, or a local model through Ollama or LM Studio).

If you're a storytelling-focused player who's tired of the same scripted lines and wants your sims' relationships to feel alive, this is for you.

Does this affect my saved game?

The mod never edits your save files. It reads game data (who lives in the household, their traits, careers, and relationships) to write its calls and texts. A few things it does change in-game, which the game then keeps when you save: messages nudge friendship and romance (turn it off under Llamafone Settings), a charged message can give a sim a moodlet, and with Get Famous a Llamagram post that truly goes viral can raise fame.

Llamafone's own data (journal, sim bios, save notes, Llamagram posts, and so on) lives in its own folder per save: Documents/Electronic Arts/The Sims 4/saves/Llamafone/Slot_…. Those are plain JSON files you can open and edit. To uninstall, delete Llamafone.ts4script, Llamafone.package, and llamafone.cfg from your Mods folder, and the saves/Llamafone folder if you don't want to keep the history.

What does the AI know about my sims?

Just what's needed for the current command — names, ages, traits, careers, aspirations, household members, and relationship history. Nothing personal about you, no real names beyond what your sims are named, no save file contents.

Your prompts go from your computer directly to whichever AI provider you configured — there's no middleman server. Each provider has its own data-retention policy: Anthropic, OpenAI, Google, and OpenRouter all state that paid API traffic is not used to train models. If you really don't want any data to leave your machine, use Ollama — it runs entirely locally.

Is my API key safe?

Yes. The key lives only in llamafone.cfg on your computer. It's never uploaded anywhere except to make the API call to whichever provider you chose. The mod doesn't have a server, doesn't phone home, and isn't open to anyone else.

That said, treat the key like a password — don't paste it in screenshots, Discord, or screen-shares. If you ever suspect it leaked, revoke it on your provider's dashboard and make a new one. (Or just switch to Ollama or LM Studio — no key, no exposure.)

Can I disable specific features?

Yes. Everything is opt-in. Auto-events (random calls/texts/stories while you play) can be toggled with llama.auto_events off in the cheat console. You can also edit llamafone.cfg to disable specific auto-event types, change interval/chance, or restrict which sims can call you.

Auto-events are off until you turn them on. Birth announcements are on by default; turn them off under Llamafone Settings (or birth_announcements = false in the cfg). Llamagram can be turned off, kept with its pop-ups silenced, or have friends post less (or never) under Llamafone Settings > Llamagram.

Does it work on Mac?

Yes — drop the .ts4script into ~/Documents/Electronic Arts/The Sims 4/Mods/ and it works identically to Windows. Same commands, same behavior.

How much does it cost?

You pay your AI provider directly for what the mod actually uses — there's no subscription to me or anyone else. Cost depends on which provider you pick:

Prices and model names change. These are rough estimates from October 2026; check your provider's pricing page for current rates.

ProviderModelPer call
Anthropic Claudeclaude-haiku-5-5 (recommended)~$0.001
Anthropic Claudeclaude-opus-5-5~$0.05
OpenAIgpt-4o-mini~$0.003
OpenAIgpt-4o~$0.03
Google Geminia current Flash modelfree (rate limited)
Ollama / LM Studio (local)any modelfree (uses your GPU)

A typical session of ~30 calls costs a few cents on Claude Haiku 5.5 and under a dime on GPT-4o-mini. Heavier sessions on the premium models (like Claude Opus) can run a dollar or two. Gemini's free tier covers most casual play. Ollama and LM Studio are fully free if you've got the hardware.

Llamagram adds a little. Each of your posts is one larger call (it writes all the comments at once), and each reply you post to a comment is one more. With auto-events on, friends posting on their own is one call per post, about one in five auto-events at the default setting. Set social_npc_post_weight = 0 to turn friends' posts off.

Does this slow down the game?

No. All API calls happen on background threads, so the game keeps running while text is generated. You'll see a notification popup when the response arrives, usually within a few seconds.

What if I'm offline?

The mod needs an internet connection to call the AI provider. If you're offline, commands silently fail without interrupting your game — no crashes, no popups. Your sims still play normally; you just won't get AI-generated content during offline sessions.

Is this against the Sims 4 Terms of Service?

No. The Sims 4 supports script mods officially through its Mods folder and Custom Content settings. This mod uses the documented script-mod system and doesn't modify any game files. EA's stance on script mods is that they're allowed; they just won't support issues that occur while mods are installed.

Get started

  1. 1

    Download the mod

    Three places to grab it, same files everywhere — CurseForge (recommended for most players), Nexus Mods, or the GitHub releases page. You need three files: Llamafone.ts4script (the script), Llamafone.package (the phone UI tunings), and llamafone.cfg (the settings file).

  2. 2

    Pick an AI provider and get a key

    Llamafone works with six different AI providers — pick whichever fits your budget, preferences, or paranoia. You only need credentials for the one you pick. Ollama and LM Studio are free because they run entirely on your computer.

    Model names change. AI companies retire old models and release new ones every few months, and prices change too. The names below were current as of October 2026; check your provider's model list for the latest. A "model not found" error usually means a model was retired. If you get empty, cut-off, or very slow replies, try a model that doesn't "think" (reason) before answering, like gpt-4o-mini or, locally, Llama 3.2.

    Replies seem confused? Try a different model. On Claude, turn up Thinking to xhigh (it's already the default on Haiku 5.5).

    🧡 Anthropic Claude

    What the mod was built and tuned against.

    1. Go to console.anthropic.com and sign up
    2. Verify email/phone if asked
    3. Add a credit card under Settings → Billing and load $5 minimum (thousands of in-game calls on claude-haiku-5-5)
    4. Go to API Keys, click Create Key, name it "Sims 4 mod"
    5. Copy the sk-ant-api03-… key immediately — you only see it once

    In llamafone.cfg: provider = claude. Recommended model: claude-haiku-5-5; for richer writing try claude-sonnet-5-5 or claude-opus-5-5.

    💚 OpenAI (GPT)

    Works with GPT models like gpt-4o-mini and gpt-4o.

    1. Go to platform.openai.com/api-keys and sign in (or sign up)
    2. Add billing under Settings → Billing; load credit ($5 minimum)
    3. On the API Keys page, click Create new secret key
    4. Copy the sk-… key — only shown once

    In llamafone.cfg: provider = openai + set model to gpt-4o-mini

    🔷 Google Gemini

    Has a generous free tier — you can play for a while without a card on file.

    1. Go to aistudio.google.com/apikey and sign in with a Google account
    2. Click Create API key (free tier requires no billing setup)
    3. Copy the AIza… key

    In llamafone.cfg: provider = gemini + model set to a current Flash model from Google's model list (Flash models are on the free tier).

    🔀 OpenRouter one key, many models

    Aggregator that proxies dozens of models (Claude, GPT, Llama, Mistral, DeepSeek, etc.) behind one API key. Useful if you want to A/B-test models without juggling separate accounts.

    1. Go to openrouter.ai and sign up
    2. Add credits under Settings → Credits ($5 minimum)
    3. Go to openrouter.ai/keys and click Create Key
    4. Copy the sk-or-… key

    In llamafone.cfg: provider = openrouter + model set to anthropic/claude-haiku-4.5 (browse the full catalog at openrouter.ai/models)

    🦙 Ollama (local — no API key) techy

    Runs on your own computer. No API key, no internet, no cost. More setup than the cloud options — you install Ollama, keep it running in the background, and pull a model yourself.

    1. Install Ollama. Make sure the tray icon stays running.
    2. Pull a model — e.g. ollama pull llama3.2:3b (small, fast) or a bigger one if your hardware handles it.
    3. In Ollama's settings, set Context length to 16k. Ollama's default is too small for Llamafone's prompts and it quietly cuts them short.
    4. Leave api_key blank.

    In llamafone.cfg: provider = ollama + model set to llama3.2:3b (must match ollama list). Run llama.testconnection in the cheat console to verify setup.

    🖥️ LM Studio (local — no API key) techy

    Also runs on your own computer, with an app for finding and loading models. No API key, no internet, no cost.

    1. Install LM Studio and download a model (a small one like Llama 3.2 3B Instruct is a good start).
    2. Load it with Context Length 16384 (12288 at the very least) — Llamafone's prompts run around 8,000 tokens.
    3. Switch to Power User mode, open the Developer tab, and turn the server on.
    4. Leave api_key blank.

    In llamafone.cfg: provider = lmstudio + model set to the model's name in LM Studio. Run llama.testconnection to verify setup — it lists the names LM Studio offers.

    Using a local model (Ollama or LM Studio)? Free and private, with a few things to know.
    • Set the context length to 16k (12k at the very least). A call or text sends the AI everything it needs to know about your sims — relationships, history, life events, the calendar — which comes to about 8,000 tokens, plus room for the reply. With less, replies fail or the model quietly loses part of what it was told.
    • Replies are slower: about 15 seconds with a decent graphics card, a minute or more without one. Cloud models answer in a few seconds.
    • Small models make more mistakes. The kind that run alongside the game sometimes mix up details, like when an event is or what to call a family member. The cloud models follow the mod's instructions much more closely.
    Treat your API key like a password. Anyone who has it can spend money on your account.
    • Never paste it in Discord, Reddit, GitHub, screenshots, or any chat — even private DMs aren't safe long-term.
    • If it ever leaks, revoke it on the provider's dashboard and make a new one.
  3. 3

    Drop the mod files in your Mods folder

    Move both downloaded files into:

    Documents/Electronic Arts/The Sims 4/Mods/

    If you have Sims 4 mods already, you've been here before — same folder.

  4. 4

    Set your provider + API key in the config

    Open llamafone.cfg with any plain text editor (TextEdit, Notepad, VS Code — anything that saves plain text). Two lines to update near the top:

    provider = claude          # or: openai, gemini, openrouter, ollama, lmstudio
    api_key  = YOUR_API_KEY_HERE

    Replace YOUR_API_KEY_HERE with the key you copied. Examples per provider:

    provider = claude
    api_key  = sk-ant-api03-Xy7…
    
    # or
    
    provider = openai
    api_key  = sk-proj-…
    
    # or
    
    provider = gemini
    api_key  = AIza…
    
    # or
    
    provider      = openrouter
    api_key       = sk-or-…
    model = anthropic/claude-haiku-4.5   # vendor/model form -- see openrouter.ai/models
    
    # or (no key needed)
    
    provider      = ollama
    api_key       =
    model = llama3.2:3b     # must match the model you pulled -- see `ollama list`
    
    # or (no key needed)
    
    provider      = lmstudio
    api_key       =
    model = llama-3.2-3b-instruct   # the model's name in LM Studio

    Save the file. If you're on Ollama, also confirm ollama_endpoint = http://localhost:11434 matches wherever your server runs. Ollama users can run llama.testconnection in the cheat console after loading a save — it verifies Ollama is reachable, the model is downloaded, and end-to-end generation works. LM Studio users: the server runs at http://localhost:1234 by default; if LM Studio shows a different address, set lmstudio_endpoint to it. llama.testconnection checks LM Studio the same way.

  5. 5

    Enable script mods & launch the game

    In The Sims 4, open Game Options → Other and tick both:

    • Enable Custom Content and Mods
    • Script Mods Allowed

    Restart the game. When the main menu loads, you'll see a notification popup confirming the mod is active.

    Inside any household, open the cheat console with Ctrl+Shift+C and type llama.status to see all the commands and confirm everything's wired up.

    Something not working? See Common errors below.

Common errors

Got an error? Find the message you saw below. Most are a quick settings fix. First step for anything: type llama.status and llama.testconnection in the cheat console.

Setup

"Not configured" / "needs an API key" — but I set up a different provider

Llamafone isn't reading the settings you edited, so it's using its defaults. Type llama.status: it shows the provider, the llamafone.cfg file, and the Mods folder Llamafone is actually using.

  • "Config file: NOT FOUND" — the cfg isn't in your Mods folder. If your Documents folder is in OneDrive, update to v3.8.1 or newer, which finds it there.
  • Provider shows the wrong one — check the provider = line itself. Setting the Ollama address or models isn't enough; it has to say provider = ollama (or lmstudio, openai, and so on).
  • After editing the cfg, type llama.reload in the cheat console (or restart the game).
"curl not found"

Llamafone needs Windows 10 (version 1803 or newer) or Windows 11. It reaches the AI through curl, a small tool built into those versions of Windows. Windows 8 and older don't include it.

"Invalid API key" / "authentication" / "credit balance is too low"

The key in api_key is wrong, or your account with the provider has no credit. Copy the key again from your provider's dashboard (keys are only shown once, so make a new one if you need to), and check that credit is loaded under the provider's billing settings. Then llama.reload.

Models

"Model not found" / "no longer available" / "no longer offers the model" / "is not found for API version"

The model (Model in AI settings, or model in the cfg) has been retired by the provider. AI companies retire old models every few months. Pick a current one from your provider's model list and put it in both lines, then llama.reload. For Claude, claude-haiku-5-5 is a good choice; for Gemini, gemini-3.5-flash-lite. Google retires Gemini 2.5 on October 20, 2026, so if you use a gemini-2.5 model, switch before then.

"API error: Request contains an invalid argument" (Gemini)

Fixed in v3.8.3 — update to the latest version. Gemini 3 models (like the free gemini-3.5-flash-lite) then work. Gemini 2.5 is being retired on October 20, 2026, so move off it before then.

"Empty response" / replies are cut off or missing

Usually a model that "thinks" (reasons) before answering, and spends its whole reply on thinking.

  • Update to v3.8.2 or newer — it gives thinking models room and fixes this for Claude Haiku 5.5 and LM Studio.
  • Try a model that doesn't think, like gpt-4o-mini or, locally, Llama 3.2.
  • Or raise Max tokens in Llamafone Settings > AI settings (for example to 2048).

Local AI (Ollama and LM Studio)

"Can't reach Ollama" / "Can't reach LM Studio"
  • Ollama: make sure Ollama is running (the llama icon in your system tray). Open it from the Start menu if it isn't.
  • LM Studio: the app being open isn't enough — the server has to be on. Switch to Power User mode (bottom of the window), open the Developer tab, and turn the server on. It should say Running at http://localhost:1234.
"The conversation is longer than the model's context length" / "Context size has been exceeded"

A call or text sends the AI about 8,000 tokens. Reload the model with a larger context length: 16384 (12288 at the very least). In LM Studio, set it when loading the model. In Ollama, set Context length to 16k in its settings.

"Request timed out" / it says "Checking messages…" and nothing happens

Your PC is taking too long to write the reply. Without a supported graphics card, a local model can take minutes. Update to v3.8.2 or newer (it waits up to 5 minutes and tells you if it gives up), and try a smaller model, like a 3B one (llama3.2:3b).

Replies are confused or ignore what's going on (Ollama)

Ollama's default context is too small, and it quietly cuts off the start of Llamafone's prompt, which holds most of the instructions. In Ollama's settings, set Context length to 16k.

"No model is loaded" / the model in llamafone.cfg isn't installed

Run llama.testconnection: it lists the models your Ollama or LM Studio actually has. Set Model to one of those names (Llamafone Settings > AI settings, or model in the cfg, then llama.reload), or install the model you want (ollama pull <name>, or load it in LM Studio).

In game

I can't type in the message box

The game has lost keyboard focus, usually after alt-tabbing to another window. If the cheat console won't take typing either, restart the game. Try to avoid alt-tabbing while a message box is open.

Garbled punctuation like ’ or —

Fixed in v3.8.2 — update to the latest version.

Still stuck? Report it and include the output of llama.status and your Llamafone_Log.txt (in your Documents folder).

What's new

What's changed in each release, newest first.

  1. October 27, 2026

    Every language, Save As, and smoother setup (v3.8.5)

    Fixes from your reports. Drop-in over v3.8.4.

    • Buttons show in every language. On games set to a language other than English, the LlamaDate, Llamagram Post, and Llamagram Notifications buttons were blank. They now show their names.
    • Save As keeps your Llamafone data. A new save made with Save As starts with everything from the save you were playing: messages, memories, Sim bios, Save Notes, and Llamagram. Save As over a save that had its own Llamafone data? That data is moved to a backup folder, never deleted.
    • Settings opens before you're set up. New Ollama and LM Studio players can open Llamafone Settings and pick their provider in game. Before, Settings wouldn't open until the cfg was set up.
    • Sims who've met know each other. A sim who came to your dinner party, saw you recently, or has texted you before no longer asks if you remember them, even when your friendship is still low.
    • Events remember who hosted and where. Sims know you hosted the dinner party at your home, or that happy hour was at the bar.
    • Fewer repeats. In a long text conversation, sims don't keep opening with "good to hear from you."
    • Cleaner settings menus, without the instructions under each header.
    • Clearer help for empty replies, and Gemini keys are no longer part of the web address Llamafone calls, so they can't show up in a log.
  2. October 12, 2026

    Settings in game and in folders, real job titles, and smarter replies (v3.8.4)

    Change almost everything without leaving the game, and sims that know more about each other. Drop-in over v3.8.3.

    Llamafone Settings grouped into folders: AI settings, Incoming calls and texts, Texts and replies, Llamagram, Manage contacts, and Dating
    Llamafone Settings, now in folders
    Llamafone AI settings in game: Provider, Model, Thinking, Max tokens, Language, and Test connection
    Llamafone Settings > AI settings
    • AI settings in game. Pick a provider from a list, switch models, set the thinking level, max tokens, and language, and test the connection, all from Llamafone Settings. Only the API key stays in llamafone.cfg.
    • Settings in folders. Llamafone Settings is grouped into AI settings, Incoming calls & texts, Texts & replies, and Llamagram, so it's shorter and easier to find things. New in there: incoming calls and texts on or off, and birth announcements.
    • Language in game. Change the language your sims write in under AI settings. Dating profiles now follow it too (they were always in English).
    • Max tokens is a cutoff, not a target: it doesn't make sims write more. New installs start at 1024; yours keeps its current value. Raise it if messages get cut off or come back empty, which happens most with local models and OpenRouter.
    • One model setting. New llamafone.cfg files have a single model line instead of default_model and fast_model, and it's used for everything. Your existing cfg keeps working as is: if it has no model line, Llamafone uses your fast_model (then default_model), the same model your texts and calls already used.
    • Fewer mix-ups on Claude Haiku 5.5. It now thinks harder before replying, so it keeps track of who's who (like who just moved in). Adjustable under Thinking.
    • Real job titles and trait names. Sims say "Adept Attorney," not "level 5 in law," and traits read the way the game shows them, like "Heart on Your Sleeve" and "Family-Oriented."
    • Replies know your household. Replying in a conversation now includes your sim's family and news, so a best friend knows about the new baby.
    • Visiting a baby counts. Relatives who came over after the birth no longer say "can't wait to meet him," even after the baby grows into an infant.
    • Calls and texts only go to sims at home with you, not to someone at work.
    • No more auto-muting. Llamafone no longer mutes or pauses a sim based on words in a message (it misread jokes like "don't text me, get back to work! lol"). Mute and Ask for space are still in Manage contacts, and anyone already muted stays muted until you change it there.
    • Squashed bugs. No more pets getting phone calls, cleaner relationship and career details in messages, and settings that didn't save when tapped now do.
  3. October 10, 2026

    Gemini 3, newer AI models, clearer errors, and Llamagram fixes (v3.8.3)

    Keeps Llamafone working with the newest AI models. Drop-in over v3.8.2.

    • Gemini 3 works. Gemini 3 models (like the free gemini-3.5-flash-lite) failed with "Request contains an invalid argument." Gemini 2.5 is being retired on October 20, so if you use Gemini, update before then.
    • Newer OpenAI models work. Llamafone uses OpenAI's current request settings and leaves room for models that think before answering.
    • New installs default to Claude Haiku 5.5. Haiku 4.5 is reaching end of support, but it still works and has no retirement date. Already set up? Your settings don't change.
    • New Year's Eve is New Year's Eve. Sims no longer talk about the countdown as if it already happened; it's at midnight.
    • Clearer errors. When your AI service is overloaded, too slow, or has retired your model, Llamafone says so in plain words and tells you what to do.
    • Llamagram comments feel real. They arrive sooner on big accounts, in fewer pop-ups, and the comment count stays in step with what you can read. If the AI is busy, Llamafone tries again and lets you know.
    • Realistic follower growth for big Llamagram accounts. A celebrity no longer gains tens of thousands of followers from an ordinary post.
    • Pregnant again? Sims now know about a new pregnancy that starts soon after a birth.
    • Up-to-date model examples in llamafone.cfg, with a reminder that model names change.
  4. October 9, 2026

    Trip memory, LM Studio, and squashed bugs (v3.8.2)

    Sims remember vacations, a new free local AI option, and a round of bug fixes. Drop-in over v3.8.1.

    New:

    • Trip memory. Llamafone now notices when your household goes on a vacation or getaway, and who came along. Sims on the trip talk about being away together while it's happening and remember it together afterward ("Apollo and Francesca went to Gibbi Point together… got back yesterday"). Everyone else hears about it the way news travels: parents within a few hours, close friends in a couple of days, acquaintances only if they're told. Trips stay in conversations for a week of game time after you get home. Type llama.trips to see what's been recorded.
    • LM Studio support. Set provider = lmstudio to run Llamafone on a model loaded in LM Studio: free, private, no API key. Load the model with a context length of 16k. llama.testconnection checks the server, lists your models, and tells you exactly what's wrong if something's off.

    Other bug fixes:

    • Claude Haiku 5.5 works. Longer texts failed with "Empty response from Claude" because Llamafone only read the first part of the reply.
    • Babies you've met aren't strangers. A grandparent who's been introduced to the new baby no longer gets a "you may not know this person" warning about them.
    • Best friends aren't described as exes. Old "No Longer Friends" tags from a past falling-out no longer count when two sims are close now.
    • Times read right. Something from last night is "last night," not "earlier today," and every part of a conversation agrees on when something happened.
    • Life events stick around. They're remembered for a week of game time, however long you're away from the game between sessions.
    • Events last as long as they should. Weddings, parties, and holidays stay "happening now" for as long as they actually last.
    • In-person memories last. "Saw each other in person" memories no longer clear early.
    • Llamagram comments survive a break. Comments still on the way when you quit always arrive, even after a long break.
    • Fair Llamagram feeds. Every household sim gets friends' posts, from casual friends too, without the same friend posting twice in a day.
    • Shared weather. A sim on the same lot or trip as you shares your weather.
    • No phantom thank-yous. Baby follow-ups don't thank you for congratulations you never sent.
    • Cleaner labels. Relationships read "Good Friend," not "Good," and Lovestruck's hidden relationship settings no longer show up as traits.
    • No real-world dates. Conversations only ever use in-game time.
    • Local model fixes. Ollama and LM Studio get up to 5 minutes per reply, models that "think" first (like Gemma 4) no longer come back empty, and dashes and apostrophes no longer come out garbled.
    • Clearer errors. LM Studio problems are explained in plain English with the fix, an empty reply shows an error instead of nothing, and failed requests are noted in Llamafone_Log.txt.

What's next

Where the mod is heading. Not a contract — priorities shift based on what you all report and ask for.

Found a bug? Tell me.

The mod is actively maintained and bug reports get read. Use the form below and you'll be filling out a GitHub issue with structured fields (what happened, log excerpt, screenshots) — no GitHub experience required, just sign in.

Report a bug

Or browse existing reports first to check if it's already known.

Love the mod? Support it on Patreon.

Become a patron

Prefer Bitcoin? Scan or copy the address:

Bitcoin donation QR code 351FwouuUN7QRNGy6iDXXFwr7rgt4qDnHx