Honcho Memory for SillyTavern

Persistent, personalized memory for SillyTavern AI characters via Honcho.

Install

From your SillyTavern directory:

macOS / Linux:

bash <(curl -fsSL https://raw.githubusercontent.com/plastic-labs/sillytavern-honcho/main/install.sh)

Windows (PowerShell):

irm https://raw.githubusercontent.com/plastic-labs/sillytavern-honcho/main/install.ps1 | iex

Then restart SillyTavern.

What the script does

1. Clones this repo into public/scripts/extensions/third-party/sillytavern-honcho
2. Symlinks the server plugin to plugins/honcho-proxy
3. Installs the @honcho-ai/sdk dependency
4. Checks that enableServerPlugins: true is set in config.yaml
5. Detects your ~/.honcho/config.json if it exists

Prerequisites

• SillyTavern running locally
• A Honcho API key from app.honcho.dev
• Server plugins enabled in config.yaml:

  enableServerPlugins: true

Configuration

Extension settings

Open Extensions (puzzle piece icon) and expand Honcho Memory:

1. Check Enable Honcho Memory
2. Click the API key field to set your key
3. Enter your workspace ID
4. Status indicator should show Ready

Global config (for multi-tool setups)

~/.honcho/config.json is a shared config file that other Honcho integrations (Claude Code, Cursor, Hermes) write when you first set them up. If the file already exists when you install this extension, the server plugin reads it on startup.

Resolution order (plugin-side, on startup):

1. hosts.sillytavern.apiKey (nested, host-specific)
2. root-level apiKey (flat fallback)
3. If neither resolves → plugin falls through; you must enter the key via the Extensions panel

Precedence against the Extensions panel key: the Extensions-panel key (SillyTavern's secret manager) takes priority at request time — the plugin checks SECRET_KEYS.HONCHO first, then falls back to the global-config key. So entering a key in the UI overrides the file without touching it.

Example ~/.honcho/config.json the plugin accepts:

{
  "apiKey": "your-honcho-api-key",
  "peerName": "your-name",
  "workspace": "sillytavern",
  "enabled": true
}

Nested form (when multiple tools share the file):

{
  "hosts": {
    "sillytavern": {
      "apiKey": "your-honcho-api-key",
      "workspace": "sillytavern"
    }
  }
}

How it works

The extension has two parts:

• Client extension (browser) -- hooks into SillyTavern events to inject memory context and store messages
• Server plugin (Node.js) -- proxies requests to the Honcho API

Peer observability

By default, only the user peer accumulates derived memory — Honcho observes the user's messages and derives conclusions about them across sessions. The AI character's persona comes from its character card, not from peer derivation. If you want the character to have its own Honcho-derived state, configure it as an additional peer in session setup (see the /session route in plugin/index.js).

Context modes

Every generation injects the peer representation and session summary from session.context() as a base layer (stale-while-revalidate -- zero latency after first turn, configurable refresh interval). The enrichment mode controls what happens on top:

Mode	Behavior
Context only	Base layer only -- peer representation + session summary
Reasoning (default)	Base layer + dialectic peer.chat() queries on an interval
Tool call	Base layer + function tools the LLM can call on demand (query, save conclusion, search)

Peer modes

Mode	Behavior
Single peer	One user peer shared across all personas
Per-persona	Each persona gets its own isolated memory

Event flow

Event	Action
Chat opened	Creates/gets Honcho session + peers
Before generation	Injects memory context into prompt
User sends message	Stores message in Honcho session
AI responds	Stores response in Honcho session

Architecture

Browser (Extension)                     Server (Plugin)
+-----------------------+               +------------------------------+
| index.js              |  fetch()      | plugin/index.js              |
|                       | ------------> |                              |
| - Settings UI         | /api/plugins/ | - Express router             |
| - Event hooks         |  honcho-proxy | - Honcho SDK (@honcho-ai/sdk)|
| - Prompt injection    |               | - API key from secrets or    |
| - Tool registration   |               |   ~/.honcho/config.json      |
+-----------------------+               +------------------------------+

File structure

sillytavern-honcho/
+-- manifest.json          Extension manifest
+-- index.js               Client extension
+-- settings.html          Settings panel
+-- style.css              Styles
+-- install.sh             Installer (macOS/Linux)
+-- install.ps1            Installer (Windows)
+-- plugin/
|   +-- index.js           Server plugin (9 routes: 1 GET + 8 POST)
|   +-- package.json       @honcho-ai/sdk dependency

Troubleshooting

Symptom	Fix
No "Honcho Memory" in Extensions	Check symlink: ls public/scripts/extensions/third-party/sillytavern-honcho/manifest.json
Plugin not initializing	Add enableServerPlugins: true to config.yaml and restart
Just installed SillyTavern, plugin silent on first boot	SillyTavern creates config.yaml with enableServerPlugins: false by default on first launch. Flip it to true and restart.
403 on plugin requests	Set Honcho API key in extension settings or ~/.honcho/config.json
SDK import error	Run cd plugins/honcho-proxy && npm install
Extension loads but nothing happens	Enable the checkbox and ensure workspace ID is set
API key modal saves but status never turns Ready	SillyTavern core patches missing (SECRET_KEYS.HONCHO not registered upstream). Either apply the patches (see CONTRIBUTING.md) or configure via ~/.honcho/config.json.

Development

• For local-clone + symlink dev setup, see CONTRIBUTING.md.
• For Claude Code-assisted install, invoke the setup skill at skills/setup/SKILL.md.