← Back to Blog
Guides12 min read

How Svila.io's Memory Journal Works (And Why It Matters)

Svila's Memory Journal gives your AI companion editable, persistent memory you can read, fix, and delete. Here's how it works and why we built it.

By the Svila.io editorial team

Most AI chats feel like meeting someone with no short-term memory. You explain who you are, you build a moment together, and then the next day you start from zero. Svila's Memory Journal is our answer to that problem — a readable, editable record of what your companion knows about you, designed so conversations actually accumulate instead of resetting. This post explains how Svila's Memory Journal works, what it deliberately does and doesn't do, and why we built it the way we did.

This is the team behind Svila.io writing about a feature we designed and shipped, so treat this as a first-party explainer rather than a neutral review. We'll be honest about the trade-offs, but our perspective is obviously not impartial. If you've used AI companions before, you already know the failure mode: the model holds a conversation well, then quietly forgets the thread, the name you gave it, or the thing you said an hour ago. The Memory Journal exists to make memory something you can see and control, not a black box you have to hope is working.

On this page
  1. Why memory matters in AI companions
  2. How we think about memory
  3. Automatic Capture
  4. 2. What the Journal Holds — About Me, Your Companion, Boundaries
  5. Editing and Pinning
  6. Forgetting
  7. 5. What's Shared Between Companions — and What Isn't
  8. Memory Retrieval
  9. Memory Limits
  10. How we approached this
  11. How to get the most out of the Memory Journal
  12. Final thoughts
  13. FAQ
  14. A note from the team

01—Why memory matters in AI companions

Why memory matters in AI companions

A companion is only as good as her continuity. The whole appeal of talking to the same companion over weeks is that she remembers — your in-jokes, your preferences, the hard week you mentioned on Tuesday. Without persistent memory, every session is a polite stranger, and the relationship never deepens.

The technical reason this is hard is that large language models have a limited working memory (the "context window") and don't retain anything between separate conversations on their own. Whatever isn't in the current context is, as far as the model is concerned, gone. So persistence has to be engineered on top of the model: something has to decide what's worth keeping, store it, and feed the right pieces back at the right moment. We built the Memory Journal to make that machinery legible, so you can read what's stored, fix what's wrong, and remove what you'd rather she forgot.

02—How we think about memory

How we think about memory

Good companion memory, in our view, isn't just "remembers more." It's memory that is accurate, controllable, and used at the right moment. When we designed the Memory Journal we kept coming back to a few principles:

  • Visibility — you should be able to see what your companion remembers, not guess at it.
  • Agency — if a memory is wrong or outdated, you should be able to edit or forget it.
  • Relevance — the right detail should surface when it matters, not a flood of trivia.
  • Clear scope — you should know which memories stay with one companion and which every companion can use.
  • Restraint — the system should know what not to store, and be honest about what it can't reliably do.

The sections below walk through the Memory Journal one component at a time, using those principles as the through-line.

Automatic Capture

How Your Companion Picks Things Up

You don't have to fill anything in. Every third message you send, Svila looks back over the recent stretch of the conversation and pulls out durable facts: things that would still be true in a month. Your name, your job, your cat's name, how you like to be talked to, a pet name the two of you use. Passing details like what you had for breakfast are skipped.

Each fact is saved as one short, plain sentence, such as "Has a cat named Otto." If a new fact repeats one already in the journal, the existing one is reinforced rather than duplicated. There's no daily summary or diary entry: the journal is a list of individual facts, which is what makes each one easy to read, fix or remove.

What it does
Picks durable facts out of your conversations and saves them as short, readable sentences.
Best for
People who talk to a companion regularly and want continuity without doing any admin.
How it works
Every third message you send, the recent exchange is read for lasting facts, which are checked against what's already stored before being saved.
Why we built it
Memory that relies on you writing everything down doesn't get used.
Bottom line
Talk normally; the journal fills itself in.

04—2. What the Journal Holds — About Me, Your Companion, Boundaries

2. What the Journal Holds — About Me, Your Companion, Boundaries

At /journal, memory is organised into three places. About me holds real-life facts about you: your name, where you live, your work, your hobbies. Each companion page shows what that companion knows, including things specific to the two of you, like pet names, inside jokes and shared moments. Boundaries holds the things you don't want: soft preferences that shape how she talks, and hard limits.

Hard limits are the strongest form of memory. You set them yourself on the Boundaries page, with an extra confirmation step, and they're included with every reply, from every companion. A second check also runs on replies: if it catches one that crosses a hard limit, the reply is replaced before you see it. Boundaries picked up from conversation are only ever saved as soft preferences — a hard limit is always your call.

What it does: Sorts memory into facts about you, facts about each relationship, and boundaries. Best for: Anyone building a long-term companion they don't want to re-explain themselves to. How it works: Each fact is stored with a type that decides where it appears and which companions can use it. Why we built it: A standing fact about your life and a limit you never want crossed shouldn't be handled the same way. Bottom line: Three clear places to look, and hard limits that apply everywhere.

Editing and Pinning

Why You Stay in Control

The Memory Journal is editable. Open a companion's page, and every fact has three actions: pin, edit and forget. If she got a detail muddled, you don't have to argue with her in chat — you rewrite the sentence, and the corrected version is what gets used from then on. On the About me page you can edit several facts at once and save them together.

Pinning is for the facts that matter most. A pinned fact is included every time that companion replies, whatever you're talking about, instead of waiting to be judged relevant. This is the feature we're most opinionated about: an AI that remembers things about you, but won't let you see or change those memories, is the wrong design.

What it does
Lets you correct facts in your own words and pin the ones that should always be present.
Best for
People who want their companion to be accurate and who like being in control.
How it works
An edited fact replaces the original text; a pinned fact is included with every reply rather than competing on relevance.
Why we built it
Memory you can't inspect or change is memory you can't trust.
Bottom line
Edit what's almost right; pin what should never slip.

Forgetting

What It Removes, and What It Doesn't

You can forget any fact. In the journal, it's the forget action on the fact itself. In chat, when a reply draws on your journal, a small "from your journal" pill appears under it (it reads "remembered something" if the fact is marked sensitive); open it and you can choose "This was wrong, forget it" without leaving the conversation. Forgetting is permanent and immediate: the fact is deleted and never used again. Hard limits are the exception — they can only be removed from the Boundaries page, with a typed confirmation.

Forgetting removes a fact from the journal. It doesn't delete the conversation it came from. Your companion still sees the recent messages in a chat, and she can also recall short snippets of earlier messages that closely match what you're talking about now. There's no one-click wipe in the journal today. If you want everything she learned gone, deleting the companion permanently deletes her and the memories learned in conversations with her. Deleting your whole account is done by emailing support@svila.io, and it's completed within 30 days.

What it does
Permanently removes individual facts, from the journal or straight from the chat.
Best for
Anyone who values privacy and wants real control over what's retained.
How it works
Forgetting deletes the stored fact; deleting a companion deletes the memories learned with her.
Why we built it
Being able to take something back is part of what makes it safe to be open.
Bottom line
Forget is instant and permanent — just know it removes the fact, not the chat.

07—5. What's Shared Between Companions — and What Isn't

5. What's Shared Between Companions — and What Isn't

If you talk to more than one companion, some memory stays with one companion and some doesn't, and it's worth knowing which is which.

Things specific to one relationship — pet names, inside jokes, shared moments — stay with the companion they came from. Real-life facts about you, your general preferences, and your boundaries are different: they sit in About me and Boundaries, and every companion you talk to can use them. That's deliberate. Your name and your hard limits shouldn't need repeating to each new companion. But it does mean that if you tell one companion where you work, another one may know it too. If a fact about you should stay out of other conversations, forget it in the journal once it's captured.

What it does: Keeps relationship details with one companion and shares facts about you across all of them. Best for: Users who talk to more than one companion. How it works: Each fact has a scope: tied to one companion, or available to every companion. Why we built it: You shouldn't have to re-introduce yourself, or re-state a hard limit, to every new companion. Bottom line: Inside jokes stay put; who you are and your limits travel with you.

Memory Retrieval

Surfacing the Right Detail at the Right Time

Storing memories is only half the job; the harder half is pulling the right ones back in. With every message you send, Svila looks for the journal facts most relevant to what you just said and includes them with her reply, alongside your pinned facts and boundaries, rather than loading everything at once.

Relevance is mostly about how closely a fact relates to your message, with a smaller weight for how recently it was learned or used. That keeps the live conversation from being crowded out by trivia. When a reply visibly draws on a remembered fact, you'll see the "from your journal" pill under it, so you can check what she used.

What it does
Selects the most relevant stored facts for each reply.
Best for
Everyone — it's the part that makes memory feel natural rather than mechanical.
How it works
Your message is matched against your journal by meaning; the best matches are included with pinned facts and boundaries.
Why we built it
Finite working memory means choosing the right details beats loading all of them.
Bottom line
Good retrieval is why your companion brings up the thing that mattered.

Memory Limits

What the Journal Deliberately Doesn't Do

Honesty matters here, so: the Memory Journal is not perfect recall. It keeps short facts, not everything ever said, and extraction can get a detail wrong or miss something you mentioned once — which is exactly why editing exists. Facts that are rarely relevant can also lose out to fresher, closer matches unless you pin them.

There are things it won't do by design. Facts about sexual content aren't stored, because stored facts are fed back into future replies. Facts your companion makes up about herself are dropped rather than recorded as things you said. And apart from boundaries, you can't type a new fact into the journal by hand today: the way to get something in is to say it clearly in conversation, then pin it once it appears.

What it does
Prioritises accurate, durable facts rather than attempting total recall.
Best for
Users who'd rather have honest, controllable memory than an overpromised black box.
How it works
Capture skips passing details and certain content; pinning protects what matters.
Why we built it
Setting honest expectations beats a promise no current system can keep.
Bottom line
It remembers what matters and gives you the tools to manage the rest.

10—How we approached this

How we approached this

We built the Memory Journal around a single bias: you should be able to see and steer your own memory. A lot of the engineering effort went not into storing more, but into storing the right things and surfacing them at the right time — picking durable facts out of conversation, keeping relationship details separate from facts about you, and retrieving selectively so a finite context window isn't wasted on trivia. The main trade-off we accepted is that the journal is made of short, discrete facts rather than rich summaries. That loses some nuance, but it means every memory is something you can read in a second and fix or remove in a couple of clicks.

11—How to get the most out of the Memory Journal

How to get the most out of the Memory Journal

  • If you're brand new: Have a normal first conversation — a few messages in — then open /journal to see what your companion captured.
  • If something's wrong: Edit the fact at the source instead of correcting it repeatedly in chat, or use "This was wrong, forget it" on the pill.
  • If a detail really matters: State it clearly in conversation, then pin it once it shows up in the journal.
  • If there's something she must never do: Add it as a hard limit on the Boundaries page.
  • If you value privacy: Remember that About me facts are available to every companion, and do an occasional pass to forget anything you'd rather not carry forward.

12—Final thoughts

Final thoughts

Memory is the quiet thing that separates a companion you return to from a chatbot you forget. We built the Memory Journal to make that memory visible, accurate, and yours — something you can read, edit, pin and forget, not a black box you trust blindly. It isn't perfect recall, and we'd rather be upfront about that than oversell it; what it is, is a memory you can actually steer. If you've been frustrated by AI chats that reset every day, it's the part of Svila we'd point you to first.

FAQ

How is Svila's Memory Journal different from regular AI chat memory?

The biggest difference is visibility and control. The journal is something you can open, read, edit when it's wrong, pin when it matters, and forget when you want to — whereas a lot of AI chat memory is invisible and unfixable.

Can I delete what my companion remembers?

Yes. You can forget individual facts in the journal, or from the "from your journal" pill in chat. Forgetting is permanent. There's no one-click wipe in the journal today; deleting a companion deletes her and the memories learned with her, and account deletion is handled by emailing support@svila.io.

Does the Memory Journal remember everything I say?

No, and that's deliberate. It picks out durable facts rather than keeping a record of every message in the journal. If a detail is important, state it clearly and pin it once it appears.

Do my different companions share memories?

Partly. Relationship details like pet names and inside jokes stay with one companion. Real-life facts about you, your preferences, and your boundaries are available to every companion you talk to.

Do I need to pay to use the Memory Journal?

No. The Memory Journal is part of the free tier, which gives you 20 messages a day and lets you create your own companions. Premium adds unlimited messages, within fair use, along with the rest of Svila's premium features.

14—A note from the team

A note from the team

This post is written by the team behind Svila.io. The features and choices we describe are ones we designed and shipped — so our perspective is first-party, not neutral. We try to be honest about the trade-offs, but you should always try things yourself and form your own view.


Last updated September 2026.

Ready to try Svila.io?

Create your first AI persona for free — no credit card required.

Get Started Free

Keep reading