Settings

Open with Open twinny settings (the gear icon in the sidebar) or search for twinny in VS Code settings. Providers, devices and the workspace index are configured in the sidebar rather than here; see Providers and Workspace index.

General

Setting	Default	Description
`twinny.enabled`	`true`	Turns the extension on or off. Enable twinny and Disable twinny in the command palette flip it
`twinny.locale`	`en`	Language of the sidebar UI: `en`, `zh-CN`, `zh-HK`, `de`, `es`, `es-CL`, `ja`, `fr`, `it`, `pt`, `ru`, `ko`, `nl`
`twinny.temperature`	`0.2`	Randomness of code completions. Lower is more deterministic. Chat, inline edit and review send no temperature, so the server’s default applies
`twinny.enableLogging`	`true`	Writes one line per request to the Twinny output channel: model, timing, size and why it stopped. Set the channel’s log level to Debug to also see prompts and replies. Warnings and errors are written even when off
`twinny.githubToken`	`""`	Personal access token for listing and reviewing GitHub pull requests in private repositories
`twinny.providerStorageLocation`	`globalState`	Where provider configurations live. `globalState` is VS Code’s global state; `file` is a file in the extension’s global storage, which survives changing remotes or containers
`twinny.secretShield`	`offMachine`	Secret shield. API keys, tokens, private keys and passwords in a prompt are replaced with placeholders such as `REDACTED_GITHUB_TOKEN_1` before the request is sent, and put back wherever the reply repeats a placeholder. `offMachine` covers anything that is not on this machine: hosted APIs, a gateway, a paired device, a server elsewhere on the network. `always` adds local servers. `off` sends prompts as they are

Code completion

Setting	Default	Description
`twinny.autoSuggestEnabled`	`true`	Suggest as you type. When off, request a suggestion with `Alt+\`
`twinny.enabledLanguages`	`{ "*": true }`	Languages to complete in, by VS Code language id, e.g. `{ "*": true, "markdown": false }`
`twinny.debounceWait`	`300`	Milliseconds of quiet before a suggestion is requested
`twinny.contextLength`	`100`	Lines before and after the cursor sent in the prompt
`twinny.multilineCompletionsEnabled`	`true`	Allow suggestions of more than one line
`twinny.maxLines`	`40`	Longest multi-line suggestion
`twinny.numPredictFim`	`512`	Token budget per completion request. `-1` for no limit; twinny still stops at a sensible boundary
`twinny.recentEditsEnabled`	`true`	Include your last few changes, in any open file, as short diffs. Up to 6 changes and 2,500 characters
`twinny.lspContextEnabled`	`true`	Include matching IntelliSense symbols and the current call’s signature. Up to 20 items, 2,000 characters, 150 ms wait
`twinny.fileContextEnabled`	`false`	Include snippets from neighbouring open files. Experimental; costs latency
`twinny.completionCacheEnabled`	`false`	Cache suggestions for identical prompts
`twinny.enableSubsequentCompletions`	`true`	Request another suggestion straight after one is accepted
`twinny.warmUpModel`	`true`	Load the completion model when VS Code starts or regains focus, so the first completion does not wait for a cold model. Local model servers only (Ollama, LM Studio, llama.cpp, text-generation-webui, a paired device, or an OpenAI-compatible server on this machine); hosted APIs are never called

Chat and review

Setting	Default	Description
`twinny.chatTools`	`false`	Where the agent mode switch starts, until you first switch it in the chat. Machine setting: a workspace cannot turn it on
`twinny.chatToolsEdits`	`apply`	Agent mode edits: `apply` writes them straight away (`Ctrl+Z` undoes), though deleting a file git has no copy of still asks; `review` opens each as a diff and the model waits
`twinny.chatToolsCommands`	`ask`	Agent mode commands: `ask` shows Run, Always run and Skip; `allow` runs them; `off` lets the model run none. Sets where the auto-run switch starts
`twinny.chatToolsCommandsRunIn`	`background`	`background` runs each command in a process of its own with its output in the chat, stopped after two minutes; `terminal` runs them in the twinny tools terminal, or in the background when VS Code’s shell integration is not available
`twinny.reviewMaxDiffChars`	`16000`	Characters of diff per code-review request. Larger reviews are split into parts. At least 2,000

Workspace index

Setting	Default	Description
`twinny.embeddingIgnoredGlobs`	`[]`	Extra glob patterns to skip when indexing, e.g. `*/.md` or `/fixtures/`. Only source, config and documentation files are indexed to begin with
`twinny.embeddingUpdateOnSave`	`true`	Re-embed a file when it is saved and forget it when deleted, once the workspace has been indexed

Chunk sizes, the rerank threshold, the snippet count and automatic use are set in the Embeddings tab of the sidebar.

Ollama

Used for listing models in the provider form and device cards, and for the Ollama this machine shares with paired devices.

Setting	Default	Description
`twinny.ollamaHostname`	`0.0.0.0`	Hostname of the local Ollama. `0.0.0.0` is treated as localhost
`twinny.ollamaApiPort`	`11434`	Port of the local Ollama
`twinny.ollamaUseTls`	`false`	Use HTTPS to reach Ollama
`twinny.keepAlive`	`5m`	How long Ollama keeps the completion model loaded after a completion request: `5m`, `1h`, or `-1` for indefinitely

Devices (P2P)

Setting	Default	Description
`twinny.p2pPort`	`49737`	UDP port this computer listens on when sharing its Ollama. Allow it through the firewall, e.g. `sudo ufw allow 49737/udp`. Takes effect the next time sharing starts

Teams

Setting	Default	Description
`twinny.teamShareSlots`	`2`	How many teammates’ requests this computer runs at once while it is shared with the team (Providers → Share this computer). From 1 to 8. Takes effect the next time sharing starts

These are stored by twinny rather than in settings.json:

Where	What
Providers tab	The providers, which one is active per job, paired devices, whether this computer is sharing
Embeddings tab	Chunk sizes, rerank threshold, snippet count, use-the-index-for-every-message
Manage templates	Which templates appear as chat suggestions

Per-workspace overrides

twinny.* settings can be set at the workspace level in .vscode/settings.json, except the four agent mode settings (twinny.chatTools, twinny.chatToolsEdits, twinny.chatToolsCommands, twinny.chatToolsCommandsRunIn), which are machine settings and only take effect from user settings. For example, to disable suggestions in a documentation repository:

{
  "twinny.autoSuggestEnabled": false,
  "twinny.embeddingIgnoredGlobs": ["**/generated/**"]
}