Voxta Chat Augmentations
Composable modules that extend an in-progress chat — reasoning helpers, vision feeds, tool integrations, document writing, game integrations. Activated per character.
Chat Augmentations are modules that don't replace a core service (like the LLM or TTS) — they sit alongside the chat and add capabilities. Read sensors. Inject context. Filter or rewrite text. Handle action inference. Pipe in vision frames. Talk to other apps and games on your behalf.
This is one of the most powerful pieces of Voxta: a small augmentation can turn your generic AI character into a cockpit voice assistant, a continuously-aware vision companion, or a document-editing collaborator — without touching the LLM or the voice.
How augmentations attach to a chat
Augmentations are opt-in per character. Every character starts with no augmentations enabled — you tick exactly the ones you want.
Augmentations are only available on Companion and Assistant chat styles. The section is hidden for Roleplay characters, and even if a Roleplay character has augmentations saved, they're ignored at chat start.
The Augmentations area on a character
Open a character → Configuration tab → scroll to Augmentations.
You'll see:
- A list of currently active augmentations, one row per enabled augmentation. Each row shows the module icon, the augmentation name, an edit button (jumps to that module's own settings page) and a delete button (removes it from this character).
- Below the list, a Configure Augmentations button that opens the picker modal.
- For a fresh character the list is empty and reads "No augmentations active. Click to add one."
The picker modal
Click Configure Augmentations to open the picker. You'll see:
- A search field at the top — filter by module or augmentation name.
- A Manage button that jumps to the global augmentations page.
- The full list of installed augmentation modules, sorted alphabetically. Each module groups its own augmentation keys together — some modules expose just one (like Reply Prefixing), others expose several (like Vision with both Continuous and Prompted).
- Each augmentation key shows as a toggle pill with a check (active) or empty circle (inactive). Click to flip.
If a module's underlying service has been disabled in the global Services configuration, the row gets a Disabled Globally warning so you know toggling it on won't take effect until you fix that.
The global Augmentations page
Open the menu → Augmentations (route /augmentations, titled Chat Augmentations in the UI).
This page lists every installed augmentation module across the whole Voxta server. Use it to:
- Toggle services on/off globally. Disabling here disables for every character, even ones that have it in their per-character list.
- Set the default augmentation set for newly created characters.
- Install / update dependencies for augmentation modules that need them (Python wheels, native binaries).
Disabled-globally augmentations still appear in the per-character selector but show a Disabled Globally warning so it's clear why nothing's happening.
Lifecycle — when an augmentation runs
An augmentation can hook into several points in a turn:
| Hook | When it fires | Used for |
|---|---|---|
| Pre-process text | Before the LLM sees the user message | Cleanup, normalization, redaction |
| Post-process text | After the LLM responds, before TTS / display | Filtering, prefixing, text replacements |
| Action inference | When the AI emits an action | Map actions to side-effects (game inputs, device commands) |
| Prefix contributions | Before generation | Inject a forced prefix into the AI's reply |
| Chat context | On chat start + on world changes | Inject live context (ship state, screen description, document content) |
| Speech replacements | Just before TTS | Substitute words to fix pronunciation per voice engine |
Most modules only use one or two of these. Each augmentation's page lists which hooks it uses.
Voxta Utilities (built-in)
Most are zero-setup. Add them via Manage Services → + Add Services.
Reply Prefixing
Prefixes character replies to encourage creativity and reduce repetition.
Continuations
Lets the AI continue talking out of turn — natural multi-line replies.
Documents
Read and write live documents from inside a chat.
Text Replacements
Apply text replacements on user input and character output.
Profanity Detector
Filter or flag profanities in AI output.
Vision
Continuous or prompted vision capture wired into the chat prompt.
Chain Of Thought
Let the AI think before answering. Experimental.
Folder Watcher
Watches a folder for new images and pulls them into chat.
MCP (HTTP/SSE)
Connect to MCP tool servers over HTTP.
MCP (STDIO)
Connect to MCP tool servers as local executables.
Integrations
External-system bridges — games, devices, third-party services.
Lovense
Connect Voxta to Lovense devices via the Remote App.
Elite Dangerous: COVAS
Full cockpit voice assistant — journal events, AI actions, auto-detected keybindings.
See also
- Web Search services (Tavily, DuckDuckGo) are Chat Augmentations under the hood; they live in their own catalog section for discoverability.
- TTS modules Orpheus and Chatterbox include a Speech Replacements augmentation that fires when those engines are active — documented on their TTS pages.
- The Minecraft Companion is an external client that consumes augmentation-style action inference over the SignalR API rather than installing as a module — same underlying pattern, different packaging.