voice.hagenow.dev // Analysis & Management Portal

CHECKING

GIGABOT // PRIVATE CONSOLE

Authentication Required

This voice telemetry and management console is private. Please sign in with Discord to verify your server membership. Only members of a Discord server where Gigabot is active are authorized to view console details.

OR ENTER ADMIN PASSCODE

Tip: Type /login in Discord for an instant 1-click magic access link.

Voice Pipeline Stage

IDLE

Channel: Not connected

Identity & Audio Gate

Gigabot

Idle: 60s | Gate: -38 dBFS | Silence: 1000ms

Turn Turnaround Latency

0 ms

Speak: 0ms Wait: 0ms STT: 0ms LLM: 0ms TTS: 0ms E2E: 0ms

Memory & Uptime

-- MB RSS

Uptime: 0s

Active AI Models

LLM: Loading...

Search Model: Loading...

Audio Synthesis & Interruption

TTS: Loading...

Cutoff: Loading...

Session Estimated API Cost

$0.0000

Tokens: 0

Multi-Speaker Room Context Memory

90s sliding window

No background room conversation recorded yet.

Voice Sessions & Event Timeline

Connecting to real-time event feed...

Bot Configuration & Voice Roster SYSTEM v1.44.0 Bot: Connecting...

Dynamic Groq LPU model routing, audio synthesis, memory limits, and voice roster management.

All changes saved

Viewer Mode (Read-Only)

> LLM

Models & Inference Routing

Groq LPU hardware acceleration, model tiers, pricing cards, and temperature tuning

Voice: openai/gpt-oss-120b Search: openai/gpt-oss-120b Temp: 0.7

Architecture Presets Groq LPU Cost ↔ Quality Routing

Primary Voice LLM Model

Provider: GROQ

Type: --

Input: --

Output: --

Context Window: --

Dedicated Web Search Model

Provider: GROQ

Type: --

Input: --

Output: --

Context Window: --

Search Provider

Select the primary provider for web search queries.

Background Task Model

Provider: GROQ

Type: --

Input: --

Output: --

Context Window: --

Decision & Router Model

Provider: GROQ

Type: --

Input: --

Output: --

Context Window: --

DeepSeek Search Tool

LLM Temperature 0.70

Max Voice Tokens (Spoken Length Ceiling)

Typical spoken turns: 120-250 tokens (~15-30 seconds of speech).

System Personality Prompt

> STT · TTS

Speech Recognition & Synthesis

Groq Whisper STT, voice synthesis engines, actor persona, and speech interruption

STT: Whisper Large v3 Turbo TTS: Orpheus (Autumn) Barge-In: Shielded

STT Recognition Provider

STT Recognition Model

Provider: GROQ

Rate: $0.0006 / min

Format: Audio Transcription

FREE $0 MSRP

Text-to-Speech (TTS) Voice Synthesis

TTS Engine Provider

Voice Actor

Playback Speed Multiplier

Vocal Style / Tone

Interruption Cutoff (Barge-In)

> Signal · Mix

Audio Gate, Latency & Volume Mix

RMS silence detection threshold, turn turnaround delays, greeting audio, and limiter volumes

Gate: medium Delay: 1000ms Mix: 100% SFX / 60% Jam

RMS Noise Gate Sensitivity

Silence Turnaround Delay (ms)

Inactivity Disconnect (secs after tasks finish, 0 = Stay Connected)

Welcome Audio Greeting on Summon

Voice (TTS) Volume % 100%

Sound Effects Volume % 100%

Jam Music Volume % 60%

YouTube Stream Volume % 80%

> Audio · Forge

Audio Generation Engines (SFX & Music)

Local procedural and General MIDI engines are always available. Cloud generation through OpenRouter is off by default.

Cloud: OFF SFX: local Music: local

Cloud Audio Generation (OpenRouter)

Cloud generation costs real money per request. It is off by default and always falls back to the local engines when a cloud model is unavailable, errors, or the daily budget is exhausted.

OpenRouter API Key

Falls back to the OPENROUTER_API_KEY environment variable when empty.

Sound Effect Model (Cloud)

Music Model (Cloud)

Both lists are read from the live OpenRouter catalog, so every model that can make audio is selectable and is shown at the price it is actually billed at. Seed Audio is billed per second of generated audio, GPT Audio per output audio token, and Lyria per clip or song. Lyria is instrumental only — a song that should have lyrics needs a chat audio model such as GPT Audio.

Model catalog not loaded yet.

Daily Cloud Audio Budget (USD / server)

Set to 0 for no cap. Spent today: $0.00

Max Generated Duration (seconds)

Upper bound for any single sound effect or music request.

Confirmation Lock (seconds)

While the bot is asking somebody to confirm a generation, other people in the channel are ignored so they cannot barge in and cancel the question. The person being asked can always carry on. Set to 0 to never lock.

Confirm Every Generation

The local engine is free and instant, so it is not held up by default.

Sound Effect Loudness Target (LUFS)

-18 LUFS sits clearly under the spoken voice.

Music Loudness Target (LUFS)

-16 LUFS matches typical streaming loudness.

Default Generation Volume

New Sound Effect Volume % 100%

Generated sounds start here so they are never silent by accident.

New Jam Volume % 100%

The floor for generated music. A per-song or per-style override still wins.

Max Cloud Prompt Characters

Longer briefs are truncated before they reach the model.

External MIDI Source (notebin.fm)

The built-in composer supersedes it; enable only for comparison.

> Social

Conversation & Persona

Bot spoken identity, wake word keyword, voice authorization mode, and room memory window

Name: Gigabot Mode: Party Ambient: IGNORED

Bot Spoken Name

Wake Word Keyword

Voice Auth Mode

Room Memory Context Window (secs)

Address Speaker by Name Chance 100%

Ambient speech is never answered on its own. Gigabot only replies when it is addressed by name. Use the Smart Response switch in Automation & Triggers to control whether ambient speech is kept as room context.

> Automation

Visual Action Builder

Wire a Trigger Block to one or more Set Action Blocks â€” no freeform LLM research

SMART RESPONSE: OFF 0 WORKFLOWS

Smart Response

Enabled: Gigabot answers when you address it by name, and ambient speech (no name) is kept as room context. Disabled: soundboard / automation only — wake events are discarded and no reply is generated. Ambient speech is never answered on its own either way.

Trigger Tips

Matching ignores casing, punctuation, spacing and trailing plurals, so "Air Horn" also fires for "Airhorn." and "air horn". Add aliases to a trigger to cover several ways of saying the same thing without duplicating its action blocks.

How voice automation works

A Trigger Block (keyword heard, wake command, or Discord event) fires a Set Action â€” a deterministic, typed block such as Play SFX, Say (TTS), Send Discord Message or HTTP Request. The runtime walks the bound action list in order and never calls the model or the web unless you explicitly place a Run Research Task or LLM Prompt block. Voice-created triggers ("when you hear Taco Bell, play the Taco Bell bong") land here automatically.

Loading action workflows...

> Storage

Data, Retention & Privacy

Sliding context window compaction, audit log purge retention, and memory controls

Compact: MANUAL Retention: 3d Max: 1000

Context window

Automatic context condensing is disabled. The rolling summary it produced bled stale context into later turns and cost an extra background model call per compaction. Condense manually any time with Condense Context Window Now or with /compact.

Audit Log Retention (Days)

Task Retention (Days)

Max Stored Log Events

You are in viewer mode. Sign in with Discord or run /settings in Discord to authenticate and edit settings.

> VC Roster

Voice Channel Roster & Nicknames

Live participants in the connected call, custom AI nicknames, and speech identity

Not connected

Connected Voice Channel: Not connected

Participant	Discord Tag	Custom Bot Nickname	Action
Loading active participants...

Spoken Recognition: The bot learns nicknames dynamically when users speak in call (e.g. "Gigabot, call me Ace") or via /nickname set.

Service	Calls	Errors	Tokens (Prompt/Comp)	Est Cost	Avg Duration
Loading service analytics...

Time	Service	Operation	Status	Tokens	Duration	Cost	Details
No API usage logs found for this window.

Public AI Data Retention & Memory Vision

Full transparency into what information the AI assistant retains across conversations. Anyone visiting voice.hagenow.dev can inspect learned facts, rolling context turns, and privacy policies.

Personal Account Privacy & Complete Opt-Out

You have full control over your data. Opting out will immediately ignore your speech in all voice channels, ignore all commands, and permanently delete all stored memories, profiles, room logs, and preferences.

Status: ACTIVE (Interactions and speech recognition enabled)

Long-Term Facts & Memories

Permanent vector facts

Active Rolling Turns

Context turns retained

Recognized User Profiles

Known Discord member names

Storage Volume Health

PERSISTENT

Unified /data mount

Data Retention & Privacy Lifecycle Guarantees

[1] Ephemeral Ambient Audio

Background voice chat is kept in a rolling RAM circular buffer (max 90 seconds) to resolve conversation references. It is never written to disk and is completely erased on channel exit.

[2] Audit Logs & Turn Recordings

Audit events and audio clips are retained for 3 days (up to 1,000 events) and automatically pruned. Configurable by administrators in Settings.

[3] Rolling Context Compaction

Dialog turns auto-compact after 8 turns into dense summaries. Permanent facts, decisions, and preferences are indexed into vector memory.

[4] Persistent Volume Storage

Long-term memories, user nicknames, settings, and to-do items persist in the unified /data volume, surviving updates and redeployments.

Learned Long-Term Semantic Memories

Loading retained memories...

Recent Context Window Turns

Loading context turns...

Recognized User Identities & Nicknames

Loading user profiles...

Users & Access Control

Root administrators in ALLOWED_USER_IDS have permanent system access and can customize their display name. Voice call participants are automatically enrolled with view access. Manage user accounts, customize names, and assign granular capabilities below.

Total Users Enrolled

Active & auto-enrolled accounts

System Administrators

Permanent & delegated admins

Voice Call Auto-Opt-In

ACTIVE

New callers auto-granted view access

Configure User Account & Access

Discord User ID *

Display Name / Nickname

System Role

Granular Capabilities

can_view - View telemetry, events, and memory can_change_settings - Modify bot settings and models can_manage_memory - Add, edit, and delete memories can_manage_tasks - Add, check off, and delete tasks can_trigger_sfx - Play sounds and countdown timers can_use_slash_commands - Execute restricted slash commands

Active Administrators (ALLOWED_USER_IDS)

Root administrators defined via environment variables maintain permanent full control. You can change your displayed name anytime by clicking Edit Name in the directory table below.

Loading allowlist...

Automatic Voice Call Enrollment

Whenever a Discord member joins voice with Gigabot, they are automatically enrolled as an active user with can_view permissions so they can log in without manual onboarding.

User Accounts & Access Directory

User ID	Display Name	Role	Capabilities	Enrolled Via	Actions
Loading users...

Discord ↔ Sleeper Fantasy Football Mappings

Match Discord user IDs, usernames, or voice caller names directly to their Sleeper accounts. Sleeper queries automatically prioritize the user's mapped account and the configured league.

Active Sleeper League ID

Discord Identifier / Name	Linked Sleeper Username	Status	Actions
Loading Sleeper mappings...

Authentication Required

You Are Currently Opted Out

All Groq Models (Loading...)

Browse & Select Models (0 Available)

Gigabot Updates & Patch Notes

New Workflow