Settings
Settings applies to the whole app. Open it from the top right of the window; the search box at the top finds any option by name. Options for a single workspace are in its Configuration.
Connections
Section titled “Connections”Where requests go. Your AI profiles and Kallamo’s own roles pick their models from these. A connection has a provider, a name, a base URL, the API key (or the cloud credentials for Vertex AI and AWS Bedrock) and a list of models. From a connection’s menu you can edit it, duplicate it, add a model or delete it. See Connect a model.
AI roles
Section titled “AI roles”Kallamo’s helpers that work behind the scenes. Your AI profiles still write every reply.
| Option | What it does |
|---|---|
| Default model (System AI) | The model every role uses unless it has its own. |
| Language for AI notes | The language of what the roles write for you to read: summaries, analyses. |
| Narrative State | Keeps the roleplay straight. System AI, own model, or off. |
| Tagger | Marks the Worldbuild entities each passage mentions. Propose unknown entities suggests new entries for names it cannot match. |
| Retrieval planner | Plans searches for profiles with smart search. Plan every message turns off the skip for short follow-ups. Native tool calling uses the provider’s own tools when it can. |
| Summarizer | Titles and summarises archived conversations. Archive summaries turns the summaries on. |
Generation
Section titled “Generation”| Option | What it does |
|---|---|
| Stream replies | Show the reply as it is written. Also avoids timeouts on slow local models. |
| Use the prompt cache | Send the unchanging part of every request first, so the provider can reuse it. |
| In the system prompt / Next to your message | Where the changing context goes. Next to your message caches the history too. |
See Context budget and prompt cache.
Memory
Section titled “Memory”Retrieval
Section titled “Retrieval”How much Kallamo looks up for each reply, and how close a match must be. See How Kallamo finds context.
| Option | What it does |
|---|---|
| Each search brings at least | The minimum number of passages for profile knowledge, workspace files, long-term memory and Worldbuild entities. More come in when the budget can hold them. |
| Strictness | How close in meaning a passage must be, from Broad to Strict. |
| Worldbuild breadth | How far relationships are followed: Focused, Balanced or Broad. |
| Passage length | Characters in each searchable piece of an imported file. |
Embeddings
Section titled “Embeddings”The model that turns your files and chats into something Kallamo can search by meaning.
| Option | What it does |
|---|---|
| On this computer | The default: a multilingual model that runs locally. Nothing leaves your computer to be indexed. |
| Through a connection | Use a provider’s embedding model instead. |
| Runs on | Native is the stable choice. WebAssembly is the fallback. |
| Downloaded model files | Clear them to free space. They download again the next time something is indexed. |
Appearance
Section titled “Appearance”| Option | What it does |
|---|---|
| Accent | Active items, the main action and the light at the top of each page. |
| Font | Inter, Merriweather or JetBrains Mono. |
| Text size | Small, medium or large. |
| Background blur | Turn off on slower computers. |
| Chat layout | Speech bubbles, or one continuous page. |
| Code theme and Line numbers in code | How code looks in the chat. |
| Formatting toolbar | In the Writing Desk: fixed at the top, or over the selection. |
| Smart typography | Curly quotes, dashes and ellipses as you type, in documents only. |
Data and backups
Section titled “Data and backups”See Data, backups and privacy.
Diagnostics
Section titled “Diagnostics”Extra detail under each reply. See Diagnostics.
About Kallamo
Section titled “About Kallamo”The version, updates, the license, and the people who made it.