d1bc77e87dfc3b1cd38385302112778f4eac3163
End-to-end voice UX iteration after DAVE E2E shipped. Each change addresses a
real symptom Marius hit in live testing today:
- Kill the 3s filler ("mă gândesc"): Claude p50 is 4-7s so the filler always
fired BEFORE the response and collided with it. Removed all filler infra
from pipeline.py + tts_stream.py (FILLER_DELAY_S, _filler_task, push_filler,
load_thinking_wav, thinking.wav cache).
- Barge-in: ttsq.clear() at the top of on_segment_done drops stale frames so
a new utterance cuts off Echo's previous response cleanly.
- DTX silence flush: Discord stops sending RTP packets when the user goes
silent (DTX), so the inline silence-check in sink.write() never fired for
the trailing audio of an utterance — STT was missed entirely. Added a
background poller thread that checks the silence-flush condition every
200ms independent of incoming packets.
- Discord audio cadence fix: EchoStreamingAudioSource.read() blocked 100ms
per call when pcm_queue was empty, wrecking Discord's 20ms frame pacing →
client interpreted the stream as stutter and discarded leading frames
(Marius heard "4 de minute în București" instead of the full sentence).
Switched to get_frame_nowait() — instant return, silence frame on empty.
- RO time expansion: "23:09" was being read as "douăzeci și trei:nouă"
with literal colon. Added expand_time() with feminine-correct minute
formatting (un minut / două minute / douăzeci de minute / una de minute).
- Supertonic Unicode sanitize centralized in tools/tts.py: Romanian curly
quotes (`„`, `"`, `"`, `—`, `…`) crash Supertonic with HTTP 500. Map them
to ASCII at the synthesize() entry so BOTH voice mode and /audio command
are covered without duplication. normalize.py re-exports for compat.
- Whisper offline: WhisperModel(..., local_files_only=True) — no more
huggingface.co metadata GET on every startup. Model is already cached.
- Diagnostic logging across the chain: sink first-packet, VAD first-speech,
voice stream block (Claude → callback), push_text (text → clauses queued),
TTS pushed (clauses → frames). Lets future "spoke but Echo silent" bugs
pinpoint exactly where the chain breaks.
- Captured Supertonic curly-quote lesson in tasks/lessons.md.
All 76 voice tests pass.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Echo Core
AI-powered personal assistant bot with Discord, Telegram, and WhatsApp bridges. Uses Claude Code CLI for conversation, with persistent sessions, cron scheduling, semantic memory search, and heartbeat monitoring.
Quick Start
# Interactive setup wizard (recommended for first install)
bash setup.sh
The wizard handles prerequisites, virtual environment, bridge tokens, config, and systemd services in 10 guided steps.
Manual Setup
# 1. Create venv and install dependencies
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
# 2. Store Discord token in keyring
./cli.py secrets set discord_token
# 3. Edit config.json (bot name, owner ID, channels)
# 4. Start
systemctl --user start echo-core
Architecture
┌─────────────┐
│ Claude CLI │
└──────┬──────┘
│
┌──────┴──────┐
│ Router │
└──────┬──────┘
┌────────────┼────────────┐
│ │ │
┌─────┴─────┐ ┌───┴───┐ ┌──────┴──────┐
│ Discord │ │Telegram│ │ WhatsApp │
│ (d.py) │ │(ptb) │ │(Baileys+py) │
└────────────┘ └────────┘ └─────────────┘
- Discord: slash commands via discord.py
- Telegram: commands + inline keyboards via python-telegram-bot
- WhatsApp: Node.js Baileys bridge + Python polling adapter
- All three run concurrently in the same asyncio event loop
Key Components
| Component | Description |
|---|---|
src/main.py |
Entry point — starts all adapters + scheduler + heartbeat |
src/router.py |
Routes messages to Claude or handles commands |
src/claude_session.py |
Claude Code CLI wrapper with --resume sessions |
src/credential_store.py |
Keyring-based secrets manager |
src/scheduler.py |
APScheduler cron jobs |
src/heartbeat.py |
Periodic health checks |
src/memory_search.py |
Ollama embeddings + SQLite semantic search |
cli.py |
CLI tool — status, doctor, logs, secrets, cron, etc. |
setup.sh |
Interactive 10-step setup wizard |
bridge/whatsapp/ |
Node.js WhatsApp bridge (Baileys + Express) |
CLI Usage
The setup wizard installs eco as a global command (~/.local/bin/eco):
eco status # Bot online/offline, uptime
eco doctor # Full diagnostic check
eco restart # Restart the service
eco restart --bridge # Restart bot + WhatsApp bridge
eco stop # Stop the service
eco logs # Tail echo-core.log (last 20 lines)
eco logs 50 # Last 50 lines
eco secrets list # Show stored credentials
eco secrets set <name> # Store a secret in keyring
eco secrets test # Check required secrets
eco sessions list # Active Claude sessions
eco sessions clear # Clear all sessions
eco channel list # Registered Discord channels
eco cron list # Show scheduled jobs
eco cron run <name> # Force-run a cron job
eco memory search "<query>" # Semantic search in memory
eco memory reindex # Rebuild search index
eco heartbeat # Run health checks
eco whatsapp status # WhatsApp bridge connection
eco whatsapp qr # QR code pairing instructions
eco send <alias> <message> # Send message via router
Configuration
config.json — runtime configuration:
{
"bot": {
"name": "Echo",
"default_model": "opus",
"owner": "DISCORD_USER_ID",
"admins": ["TELEGRAM_USER_ID"]
},
"channels": { },
"telegram_channels": { },
"whatsapp": {
"enabled": true,
"bridge_url": "http://127.0.0.1:8098",
"owner": "PHONE_NUMBER"
},
"whatsapp_channels": { }
}
Secrets (Discord/Telegram tokens) are stored in the system keyring, not in config files.
Services
Echo Core runs as systemd user services:
systemctl --user start echo-core # Start bot
systemctl --user start echo-whatsapp-bridge # Start WA bridge
systemctl --user status echo-core # Check status
journalctl --user -u echo-core -f # Follow logs
Requirements
- Python 3.12+
- Claude Code CLI
- Node.js 22+ (only for WhatsApp bridge)
Tests
source .venv/bin/activate
pytest tests/
440 tests, zero failures.
Description
Languages
Python
73%
HTML
23.3%
Shell
2.6%
JavaScript
0.6%
CSS
0.5%