dguiducci
efb5b1dc33
Nightly Build / build (push) Successful in 7m44s
An agent that wanted to know which MCP servers it had called `list_mcp_servers` — a tool that has never existed anywhere in this repo — and got "unknown tool". It was not a random hallucination: the prompt block says "the system prompt shows available servers", and `render_mcp_list` returned an empty string when nothing was connected. The model read a promise, found no table, and invented the discovery tool the text implied. The `mcp` kinds of `list_items`/`toggle_item` had been removed to close the §14 RCE vector, which was right for the write half and left no read half at all. So `list_items` gains `type: "mcp"` and returns the whole picture in one call, split into four buckets that each answer a different question: what is already loaded (call its tools directly), what is ready for `activate_tools`, what is installed but unusable and why, and what the user could still activate. Conflating the first two is what produced the original failure, so they stay apart. Every entry carries a derived note and a next step; when the step is a human one, it says so and names the UI page, because there is no tool for it. Read-only, and structurally so: `toggle_item` deliberately gains nothing, and the new `McpDirectory` trait exposes exactly one method. Enabling a connector from a tool is the thing §14 removed, and a wider seam here is how it would come back. Deny-by-default survives the report — an ungranted connector is not named at all, since a listing of what to ask for is itself a leak — except for a catalogue manager, who cannot administer what they cannot see. Three sources answer three questions and none is redundant: the registry says what exists and who may have it, the owner database says what was activated, and the live runtimes say what is connected right now — a row can read `ready` while its process is dead. The live half reaches the tool through the turn's extension map, alongside the pool and the fs view; with no live view the durable picture still renders, so freshness is an improvement and never a precondition. The static `__MCP_LIST__` table stays as it was, because it is frozen per conversation for prompt-cache stability. Its empty case now says so out loud and points at the tool.
45 lines
5.0 KiB
Markdown
45 lines
5.0 KiB
Markdown
# Documentation index
|
|
|
|
This folder is written for **you, the assistant**, not for the human directly. It is mounted read-only at `~/docs/` in your workspace. Read it when a user asks how the software itself works, wants help configuring something, or asks what's possible — then explain it in your own words, adapted to that person (their technical level, their language, their actual goal). Don't just paste these files back at them.
|
|
|
|
Keep answers grounded in what's actually enabled and configured for this instance — check with the relevant tool (e.g. list installed/enabled plugins) rather than assuming everything described here is turned on. A feature documented here may not be enabled on this particular instance.
|
|
|
|
This index will grow over time. Right now it covers memory, projects, background tasks, system agents, access grants, connectors, voice input and plugins; more sections (agents, security groups, shared folders…) will be added later.
|
|
|
|
## Features
|
|
|
|
| Document | What it covers |
|
|
| --- | --- |
|
|
| [memory.md](memory.md) | Private and shared memory: what goes where, the indexes and history log, why some shared facts can't be changed on request |
|
|
| [projects.md](projects.md) | Projects: shared folders with their own assistant chat, a live file explorer, and member sharing |
|
|
| [system-agents.md](system-agents.md) | Background agents that run on a schedule (event triage, the two memory lints, the nightly conversation review of a supervised account): what they watch, why they only ever report, why a run can be skipped, and their settings |
|
|
| [tasks.md](tasks.md) | Background tasks: the strip above the message box, following one live, stopping one, and how every outcome comes back to the conversation |
|
|
| [settings.md](settings.md) | The admin's Config page: interface language, the compaction model picker, debug mode |
|
|
| [access.md](access.md) | Who can use which plugin or connector: the open default, removing access per person, and the role switch that keeps children out of it |
|
|
| [connectors.md](connectors.md) | Connectors (MCP servers): shared vs per-user, setting one up in the UI, the sign-in and QR-pairing flows, and what to do when one is not working |
|
|
| [voice.md](voice.md) | Voice input: configuring a transcription model, and why the microphone button does nothing unless the page is served over HTTPS or localhost |
|
|
|
|
## Plugins
|
|
|
|
Plugins are optional add-ons an admin can enable and configure — extra voices, extra ways to reach the assistant (Telegram, a phone app), image generation, long-term memory, remote access, and so on. Each has its own document in [`plugins/`](plugins/):
|
|
|
|
| Document | What it adds |
|
|
| --- | --- |
|
|
| [plugins/comfyui.md](plugins/comfyui.md) | Local image generation via a self-hosted ComfyUI server |
|
|
| [plugins/elevenlabs.md](plugins/elevenlabs.md) | Cloud text-to-speech and transcription (ElevenLabs) |
|
|
| [plugins/whisper_local.md](plugins/whisper_local.md) | Local, private speech-to-text (no cloud, no API key) |
|
|
| [plugins/kokoro_tts.md](plugins/kokoro_tts.md) | Local, lightweight text-to-speech (CPU-only, no API key) |
|
|
| [plugins/orpheus_tts_3b.md](plugins/orpheus_tts_3b.md) | Local, expressive text-to-speech with emotion tags (needs a GPU) |
|
|
| [plugins/honcho.md](plugins/honcho.md) | Long-term cross-session memory via an external Honcho server (opt-in per user) |
|
|
| [plugins/telegram.md](plugins/telegram.md) | Chat with the assistant from Telegram |
|
|
| [plugins/mobile-connector.md](plugins/mobile-connector.md) | Companion mobile app: Inbox notifications + remote access, end-to-end encrypted |
|
|
| [plugins/remote_connectivity.md](plugins/remote_connectivity.md) | Reach the web app remotely over a Tailscale mesh network |
|
|
|
|
General plugin mechanics that apply to all of them:
|
|
|
|
- An admin enables/disables and configures each plugin from the **Plugins** page (sidebar → Plugins, admin-only): one card per plugin, an enable toggle, and a **Configure** button opening its settings form.
|
|
- Enabling a plugin hands it to everyone straight away — except to roles that opt out of that (the Children role does). The admin then *removes* it from whoever should not have it, rather than granting it person by person. Full details in [access.md](access.md). (Mobile Connector is the one exception to the whole grant model: access there is the device-pairing itself, not a grant list.)
|
|
- Access is changed **per person, from that person's own page**: sidebar → Users → click the user → the **Plugins** section, right below their Connectors. So "what may this person use?" is answered in one place, for plugins and connectors together. (The plugin's own page shows the reverse view — who currently holds it — but read-only.) Admins can use every enabled plugin without being granted anything.
|
|
- A plugin with **per-user** settings (e.g. Telegram's pairing code, Honcho's memory opt-in) gives each granted user its own dedicated **sidebar page** to manage them — separate from the admin's instance-wide config.
|
|
- A plugin can add tools the assistant calls directly (e.g. `set_secret`, `telegram_pairing`), a dedicated sidebar page, or both.
|