From 06564d843e04b4c4934fa43014bba47d05eea2d8 Mon Sep 17 00:00:00 2001 From: Ankush Malaker <43288948+AnkushMalaker@users.noreply.github.com> Date: Sun, 15 Mar 2026 22:10:39 +0000 Subject: [PATCH 01/25] edge code --- backends/advanced/Dockerfile | 4 +- backends/advanced/docker-compose.yml | 4 + backends/advanced/pyproject.toml | 1 + .../src/advanced_omi_backend/app_factory.py | 48 +++ .../controllers/system_controller.py | 96 ++++++ .../routers/modules/system_routes.py | 8 + backends/advanced/uv.lock | 14 + backends/advanced/webui/src/App.tsx | 8 + .../webui/src/components/layout/Layout.tsx | 3 +- backends/advanced/webui/src/pages/Network.tsx | 306 ++++++++++++++++++ backends/advanced/webui/src/services/api.ts | 3 + edge/Dockerfile | 5 + edge/README.md | 82 +++++ edge/agent.py | 69 ++++ edge/install.sh | 181 +++++++++++ edge/run-on-rpi.sh | 5 + edge/services.yml | 33 ++ edge/status.sh | 49 +++ edge/stop.sh | 37 +++ extras/asr-services/docker-compose.yml | 13 + extras/havpe-relay/docker-compose.yml | 13 + extras/havpe-relay/init.py | 20 ++ extras/havpe-relay/pyproject.toml | 1 + extras/havpe-relay/relay_core.py | 32 +- extras/llm-services/docker-compose.yml | 13 + extras/speaker-recognition/docker-compose.yml | 13 + 26 files changed, 1056 insertions(+), 5 deletions(-) create mode 100644 backends/advanced/webui/src/pages/Network.tsx create mode 100644 edge/Dockerfile create mode 100644 edge/README.md create mode 100644 edge/agent.py create mode 100755 edge/install.sh create mode 100755 edge/run-on-rpi.sh create mode 100644 edge/services.yml create mode 100755 edge/status.sh create mode 100755 edge/stop.sh diff --git a/backends/advanced/Dockerfile b/backends/advanced/Dockerfile index 7208e696..5da81141 100644 --- a/backends/advanced/Dockerfile +++ b/backends/advanced/Dockerfile @@ -45,7 +45,7 @@ COPY --from=builder /usr/local/lib/python3.12/site-packages /usr/local/lib/pytho COPY --from=builder /usr/local/bin /usr/local/bin # Source layout needs PYTHONPATH -ENV PYTHONPATH=/app/src +ENV PYTHONPATH=/app/src:/app # Copy application code COPY . . @@ -83,7 +83,7 @@ RUN --mount=type=cache,target=/root/.cache/uv \ uv pip install --system -r requirements.txt && \ rm /bin/uv /bin/uvx -ENV PYTHONPATH=/app/src +ENV PYTHONPATH=/app/src:/app # Copy application code COPY . . diff --git a/backends/advanced/docker-compose.yml b/backends/advanced/docker-compose.yml index e74811f4..14bd1fd4 100644 --- a/backends/advanced/docker-compose.yml +++ b/backends/advanced/docker-compose.yml @@ -42,6 +42,8 @@ services: - ./data:/app/data - ../../config:/app/config # Mount entire config directory (includes config.yml, defaults.yml, plugins.yml) - ../../plugins:/app/plugins # External plugins directory + - ../../discovery.py:/app/discovery.py:ro # Service discovery module + - /var/run/tailscale/tailscaled.sock:/var/run/tailscale/tailscaled.sock:ro # Tailscale socket for minidisc environment: - DEEPGRAM_API_KEY=${DEEPGRAM_API_KEY} - PARAKEET_ASR_URL=${PARAKEET_ASR_URL} @@ -99,6 +101,8 @@ services: - ./data:/app/data - ../../config:/app/config # Mount entire config directory (includes config.yml, defaults.yml, plugins.yml) - ../../plugins:/app/plugins # External plugins directory + - ../../discovery.py:/app/discovery.py:ro # Service discovery module + - /var/run/tailscale/tailscaled.sock:/var/run/tailscale/tailscaled.sock:ro # Tailscale socket for minidisc environment: - DEEPGRAM_API_KEY=${DEEPGRAM_API_KEY} - PARAKEET_ASR_URL=${PARAKEET_ASR_URL} diff --git a/backends/advanced/pyproject.toml b/backends/advanced/pyproject.toml index 56b5c62a..f6ca66aa 100644 --- a/backends/advanced/pyproject.toml +++ b/backends/advanced/pyproject.toml @@ -38,6 +38,7 @@ dependencies = [ "websockets>=12.0", "croniter>=1.3.0", "rich>=13.0.0", + "minidisc-python>=0.1.0", ] [project.optional-dependencies] diff --git a/backends/advanced/src/advanced_omi_backend/app_factory.py b/backends/advanced/src/advanced_omi_backend/app_factory.py index c1d56fed..8758d566 100644 --- a/backends/advanced/src/advanced_omi_backend/app_factory.py +++ b/backends/advanced/src/advanced_omi_backend/app_factory.py @@ -7,6 +7,7 @@ import asyncio import logging +import os import time from contextlib import asynccontextmanager from pathlib import Path @@ -215,12 +216,51 @@ async def _init_prompt_registry(): except Exception as e: application_logger.warning(f"Prompt registry initialization failed: {e}") + async def _init_service_discovery(): + try: + from discovery import ( + CHRONICLE_ASR, + CHRONICLE_BACKEND, + CHRONICLE_SPEAKER, + advertise_service, + is_tailscale_available, + ) + + if not is_tailscale_available(): + application_logger.debug( + "Tailscale socket not found — skipping service advertisement" + ) + return + + registry = advertise_service(8000, CHRONICLE_BACKEND) + if registry: + app.state.minidisc_registry = registry + + # Also advertise co-located optional services if configured + speaker_url = os.environ.get("SPEAKER_SERVICE_URL", "") + if speaker_url and "host.docker.internal" in speaker_url: + advertise_service(8085, CHRONICLE_SPEAKER) + + asr_url = os.environ.get("PARAKEET_ASR_URL", "") + if asr_url and "host.docker.internal" in asr_url: + advertise_service(8767, CHRONICLE_ASR) + + except ImportError: + application_logger.debug( + "discovery module not available — skipping service advertisement" + ) + except Exception as e: + application_logger.debug( + "Service discovery initialization failed (non-fatal): %s", e + ) + await asyncio.gather( _init_redis_rq(), _init_task_manager(), _init_client_manager(), _init_otel(), _init_prompt_registry(), + _init_service_discovery(), ) application_logger.info( @@ -487,6 +527,14 @@ async def _init_plugins(): except Exception as e: application_logger.error(f"Error stopping cron scheduler: {e}") + # Stop minidisc service advertisement + if hasattr(app.state, "minidisc_registry") and app.state.minidisc_registry: + try: + app.state.minidisc_registry = None + application_logger.info("Minidisc service advertisement stopped") + except Exception as e: + application_logger.debug("Error stopping minidisc registry: %s", e) + # Shutdown memory service and speaker service shutdown_memory_service() application_logger.info("Memory and speaker services shut down.") diff --git a/backends/advanced/src/advanced_omi_backend/controllers/system_controller.py b/backends/advanced/src/advanced_omi_backend/controllers/system_controller.py index dc51d061..970256b6 100644 --- a/backends/advanced/src/advanced_omi_backend/controllers/system_controller.py +++ b/backends/advanced/src/advanced_omi_backend/controllers/system_controller.py @@ -36,6 +36,102 @@ logger = logging.getLogger(__name__) audio_logger = logging.getLogger("audio_processing") + +async def get_network_discovery(app): + """Return Tailscale status and discovered minidisc services. + + The *app* parameter is the FastAPI application instance, used to read + ``app.state.minidisc_registry`` (set during startup in app_factory). + """ + import asyncio as _asyncio + + result = { + "tailscale_available": False, + "advertising": [], + "discovered_services": [], + } + + try: + from discovery import ( + CHRONICLE_ASR, + CHRONICLE_BACKEND, + CHRONICLE_OPENMEMORY, + CHRONICLE_SPEAKER, + is_tailscale_available, + list_all_services, + ) + except ImportError: + result["error"] = "discovery module not available" + return result + + result["tailscale_available"] = is_tailscale_available() + + # Report what this node is *actually* advertising via the minidisc + # registry that was initialised at startup (app_factory Phase 2). + registry = getattr(app.state, "minidisc_registry", None) + if registry is not None: + # The backend always advertises itself when the registry exists. + result["advertising"].append({"name": CHRONICLE_BACKEND, "port": 8000}) + + # Optional co-located services — app_factory only advertises them + # when the env-var URL points at host.docker.internal, so we + # mirror that same check here. + speaker_url = os.environ.get("SPEAKER_SERVICE_URL", "") + if speaker_url and "host.docker.internal" in speaker_url: + result["advertising"].append({"name": CHRONICLE_SPEAKER, "port": 8085}) + + asr_url = os.environ.get("PARAKEET_ASR_URL", "") + if asr_url and "host.docker.internal" in asr_url: + result["advertising"].append({"name": CHRONICLE_ASR, "port": 8767}) + + openmemory_url = os.environ.get("OPENMEMORY_MCP_URL", "") + if openmemory_url and "host.docker.internal" in openmemory_url: + result["advertising"].append({"name": CHRONICLE_OPENMEMORY, "port": 8765}) + + if not result["tailscale_available"]: + return result + + # Discover all chronicle-* services on the Tailnet via list_all_services() + loop = _asyncio.get_running_loop() + all_services = await loop.run_in_executor(None, list_all_services) + + async def _health_check(svc: dict): + name = svc["name"] + address = svc.get("address", "") + port = svc.get("port", 0) + labels = svc.get("labels", {}) + host = labels.get("host", address) + + url = f"http://{address}:{port}" if address and port else None + reachable = False + if url: + try: + import httpx + + async with httpx.AsyncClient(timeout=3.0) as client: + resp = await client.get(f"{url}/health") + reachable = resp.status_code < 500 + except Exception: + pass + return { + "name": name, + "url": url, + "reachable": reachable, + "labels": labels, + "host": host, + } + + if all_services: + discovered = await _asyncio.gather( + *[_health_check(svc) for svc in all_services] + ) + result["discovered_services"] = list(discovered) + else: + result["discovered_services"] = [] + + return result + + _yaml = YAML() _yaml.preserve_quotes = True diff --git a/backends/advanced/src/advanced_omi_backend/routers/modules/system_routes.py b/backends/advanced/src/advanced_omi_backend/routers/modules/system_routes.py index 0f580be7..1622b37f 100644 --- a/backends/advanced/src/advanced_omi_backend/routers/modules/system_routes.py +++ b/backends/advanced/src/advanced_omi_backend/routers/modules/system_routes.py @@ -33,6 +33,14 @@ class MemoryConfigRequest(BaseModel): config_yaml: str +@router.get("/system/network") +async def get_network_discovery( + request: Request, current_user: User = Depends(current_superuser) +): + """Get Tailscale and minidisc service discovery status. Admin only.""" + return await system_controller.get_network_discovery(request.app) + + @router.get("/config/diagnostics") async def get_config_diagnostics(current_user: User = Depends(current_superuser)): """Get configuration diagnostics including errors, warnings, and status. Admin only.""" diff --git a/backends/advanced/uv.lock b/backends/advanced/uv.lock index 61ddc861..feb1ba96 100644 --- a/backends/advanced/uv.lock +++ b/backends/advanced/uv.lock @@ -27,6 +27,7 @@ dependencies = [ { name = "langchain-neo4j" }, { name = "langfuse" }, { name = "mem0ai" }, + { name = "minidisc-python" }, { name = "motor" }, { name = "neo4j" }, { name = "ollama" }, @@ -99,6 +100,7 @@ requires-dist = [ { name = "langchain-neo4j" }, { name = "langfuse", specifier = ">=4.0.0" }, { name = "mem0ai", git = "https://github.com/AnkushMalaker/mem0.git?rev=main" }, + { name = "minidisc-python", specifier = ">=0.1.0" }, { name = "motor", specifier = ">=3.7.1" }, { name = "neo4j", specifier = ">=5.0.0,<6.0.0" }, { name = "ollama", specifier = ">=0.4.8" }, @@ -2195,6 +2197,18 @@ dependencies = [ { name = "sqlalchemy" }, ] +[[package]] +name = "minidisc-python" +version = "0.1.1" +source = { registry = "https://pypi.org/simple" } +dependencies = [ + { name = "pydantic" }, +] +sdist = { url = "https://files.pythonhosted.org/packages/0a/f0/19ed522aecb5da6921710948e48cdf564b076690413092c0236a417bca67/minidisc_python-0.1.1.tar.gz", hash = "sha256:f0edb7279179e0b6fa15d2a1c0261dbf957345bef44df23d340a49553a23c102", size = 6758, upload_time = "2025-04-25T11:35:24.938Z" } +wheels = [ + { url = "https://files.pythonhosted.org/packages/1a/3b/9bf00d28c5baf64556a96c71c82f3094b3b3bca2a6ece60957446328962a/minidisc_python-0.1.1-py3-none-any.whl", hash = "sha256:634ca246e74944f87637a38c0764ba93829f4c516ece575e8ad7219a0f600a9e", size = 6022, upload_time = "2025-04-25T11:35:23.497Z" }, +] + [[package]] name = "more-itertools" version = "10.8.0" diff --git a/backends/advanced/webui/src/App.tsx b/backends/advanced/webui/src/App.tsx index d1f1ff5c..9ae62d5e 100644 --- a/backends/advanced/webui/src/App.tsx +++ b/backends/advanced/webui/src/App.tsx @@ -23,6 +23,7 @@ const Queue = lazy(() => import('./pages/Queue')) const LiveRecord = lazy(() => import('./pages/LiveRecord')) const Plugins = lazy(() => import('./pages/Plugins')) const Finetuning = lazy(() => import('./pages/Finetuning')) +const Network = lazy(() => import('./pages/Network')) function PageSkeleton() { @@ -160,6 +161,13 @@ function App() { } /> + + }> + + + + } /> diff --git a/backends/advanced/webui/src/components/layout/Layout.tsx b/backends/advanced/webui/src/components/layout/Layout.tsx index 1dd65c60..b9358b75 100644 --- a/backends/advanced/webui/src/components/layout/Layout.tsx +++ b/backends/advanced/webui/src/components/layout/Layout.tsx @@ -1,5 +1,5 @@ import { Link, useLocation, Outlet } from 'react-router-dom' -import { Music, MessageSquare, MessageCircle, Brain, Users, Upload, Settings, LogOut, Sun, Moon, Shield, Radio, Layers, Puzzle, Zap, Activity } from 'lucide-react' +import { Music, MessageSquare, MessageCircle, Brain, Users, Upload, Settings, LogOut, Sun, Moon, Shield, Radio, Layers, Puzzle, Zap, Activity, Network } from 'lucide-react' import { useAuth } from '../../contexts/AuthContext' import { useTheme } from '../../contexts/ThemeContext' import GlobalRecordingIndicator from './GlobalRecordingIndicator' @@ -22,6 +22,7 @@ export default function Layout() { { path: '/queue', label: 'Queue & Events', icon: Layers }, { path: '/plugins', label: 'Plugins', icon: Puzzle }, { path: '/finetuning', label: 'Fine-tuning', icon: Zap }, + { path: '/network', label: 'Network', icon: Network }, { path: '/system', label: 'System Status', icon: Activity }, { path: '/settings', label: 'Settings', icon: Settings }, ] : []), diff --git a/backends/advanced/webui/src/pages/Network.tsx b/backends/advanced/webui/src/pages/Network.tsx new file mode 100644 index 00000000..202f1029 --- /dev/null +++ b/backends/advanced/webui/src/pages/Network.tsx @@ -0,0 +1,306 @@ +import { useState } from 'react' +import { Network as NetworkIcon, RefreshCw, CheckCircle, XCircle, Wifi, WifiOff, Radio, Search, Server } from 'lucide-react' +import { useQuery } from '@tanstack/react-query' +import { useAuth } from '../contexts/AuthContext' +import { systemApi } from '../services/api' + +interface DiscoveredService { + name: string + url: string | null + reachable: boolean + error?: string + labels?: Record + host?: string +} + +interface AdvertisedService { + name: string + port: number +} + +interface NetworkData { + tailscale_available: boolean + advertising: AdvertisedService[] + discovered_services: DiscoveredService[] + error?: string +} + +const SERVICE_DISPLAY: Record = { + 'chronicle-backend': { label: 'Chronicle Backend', description: 'Core API and audio processing' }, + 'chronicle-speaker': { label: 'Speaker Recognition', description: 'Voice identification service' }, + 'chronicle-asr': { label: 'Parakeet ASR', description: 'Offline speech-to-text' }, + 'chronicle-openmemory': { label: 'OpenMemory MCP', description: 'Cross-client memory server' }, + 'chronicle-llm': { label: 'Local LLM', description: 'Local LLM via llama.cpp' }, + 'chronicle-tts': { label: 'Text-to-Speech', description: 'TTS synthesis service' }, + 'chronicle-relay': { label: 'HAVPE Relay', description: 'ESP32 audio bridge' }, +} + +function getServiceDisplay(name: string) { + if (SERVICE_DISPLAY[name]) return SERVICE_DISPLAY[name] + // Derive a label from unknown chronicle-* names + const suffix = name.replace('chronicle-', '') + return { label: suffix.charAt(0).toUpperCase() + suffix.slice(1), description: '' } +} + +// Group discovered services by host +function groupByHost(services: DiscoveredService[]): Record { + const groups: Record = {} + for (const svc of services) { + const host = svc.host || svc.url?.replace(/^https?:\/\//, '').replace(/:\d+$/, '') || 'unknown' + if (!groups[host]) groups[host] = [] + groups[host].push(svc) + } + return groups +} + +export default function Network() { + const { isAdmin } = useAuth() + const [isScanning, setIsScanning] = useState(false) + + const { data, isLoading, refetch, dataUpdatedAt } = useQuery({ + queryKey: ['system', 'network'], + queryFn: async () => { + const response = await systemApi.getNetworkDiscovery() + return response.data + }, + enabled: isAdmin, + staleTime: 60_000, + }) + + const handleScan = async () => { + setIsScanning(true) + try { + await refetch() + } finally { + setIsScanning(false) + } + } + + if (!isAdmin) { + return ( +
+ +

+ Access Restricted +

+

+ You need administrator privileges to view network status. +

+
+ ) + } + + const loading = isLoading || isScanning + const lastUpdated = dataUpdatedAt ? new Date(dataUpdatedAt) : null + const advertisedNames = new Set(data?.advertising?.map(s => s.name) || []) + const nodeGroups = data?.discovered_services ? groupByHost(data.discovered_services) : {} + + return ( +
+ {/* Header */} +
+
+ +

+ Network +

+
+
+ {lastUpdated && ( + + Last scan: {lastUpdated.toLocaleTimeString()} + + )} + +
+
+ + {/* Tailscale Status */} +
+
+ {data?.tailscale_available ? ( + <> + +
+ Tailscale Connected +

+ Service discovery via minidisc is active. Services on your Tailnet can find each other automatically. +

+
+ + ) : ( + <> + +
+ Tailscale Not Detected +

+ {data?.error + ? data.error + : 'Install Tailscale and mount the socket to enable automatic service discovery across machines.'} +

+
+ + )} +
+
+ + {/* Advertising (This Node) */} +
+

+ + This Node is Advertising +

+ {data?.advertising && data.advertising.length > 0 ? ( +
+ {data.advertising.map((svc) => { + const display = getServiceDisplay(svc.name) + return ( +
+
+
{display.label}
+
Port {svc.port}
+
+ + broadcasting + +
+ ) + })} +
+ ) : ( +

+ {data?.tailscale_available + ? 'No services being advertised' + : 'Tailscale required for service advertisement'} +

+ )} +
+ + {/* Discovered Services — grouped by node */} +
+

+ + Discovered on Tailnet +

+ {Object.keys(nodeGroups).length > 0 ? ( +
+ {Object.entries(nodeGroups).map(([host, services]) => { + const hasEdge = services.some(s => s.labels?.type === 'edge') + return ( +
+ {/* Node header */} +
+
+ + {host} + {hasEdge && ( + + edge + + )} +
+ + {services.length} service{services.length !== 1 ? 's' : ''} + +
+ {/* Services on this node */} +
+ {services.map((svc) => { + const display = getServiceDisplay(svc.name) + const isLocal = advertisedNames.has(svc.name) + return ( +
+
+
+
+ + {display.label} + + {isLocal && svc.url && ( + + local + + )} +
+ {display.description && ( +
+ {display.description} +
+ )} + {svc.url && ( + + {svc.url} + + )} +
+
+ {svc.url ? ( + svc.reachable ? ( + + ) : ( + + ) + ) : ( + not found + )} +
+
+
+ ) + })} +
+
+ ) + })} +
+ ) : ( +

+ {loading + ? 'Scanning Tailnet...' + : data?.tailscale_available + ? 'No services discovered. Make sure other machines are running Chronicle services.' + : 'Tailscale required for service discovery'} +

+ )} +
+ + {/* Legend */} +
+

Status Legend

+
+
+ + Reachable +
+
+ + Found but /health unreachable +
+
+ edge + Remote edge node +
+
+ not found + Not discovered on Tailnet +
+
+
+
+ ) +} diff --git a/backends/advanced/webui/src/services/api.ts b/backends/advanced/webui/src/services/api.ts index ebca7322..981b5b1b 100644 --- a/backends/advanced/webui/src/services/api.ts +++ b/backends/advanced/webui/src/services/api.ts @@ -375,6 +375,9 @@ export const systemApi = { testLLMModel: (modelName: string | null) => api.post('/api/admin/llm-operations/test', { model_name: modelName }), + // Network discovery + getNetworkDiscovery: () => api.get('/api/system/network'), + // System restart operations restartWorkers: () => api.post('/api/admin/system/restart-workers'), restartBackend: () => api.post('/api/admin/system/restart-backend'), diff --git a/edge/Dockerfile b/edge/Dockerfile new file mode 100644 index 00000000..a24bb391 --- /dev/null +++ b/edge/Dockerfile @@ -0,0 +1,5 @@ +FROM python:3.12-slim +RUN pip install --no-cache-dir minidisc-python +COPY agent.py /app/agent.py +WORKDIR /app +CMD ["python", "agent.py"] diff --git a/edge/README.md b/edge/README.md new file mode 100644 index 00000000..e3f740b6 --- /dev/null +++ b/edge/README.md @@ -0,0 +1,82 @@ +# Chronicle Edge Deployment + +Deploy Chronicle services on remote machines (RPi, GPU VMs) with one command. Services auto-discover each other via Tailscale. + +## Prerequisites + +On the remote machine: +- Docker (with `docker compose`) +- Tailscale (connected to your Tailnet) +- `uv` (Python package manager) +- `git` + +## Deploy a Service + +```bash +curl -sSL https://raw.githubusercontent.com/SimpleOpenSoftware/chronicle/main/edge/install.sh \ + | bash -s -- +``` + +### Available Services + +| Name | Description | +|------|-------------| +| `speaker-recognition` | Speaker identification (pyannote) | +| `asr-services` | Offline speech-to-text (Parakeet/NeMo) | +| `tts` | Text-to-speech synthesis | +| `llm-services` | Local LLM via llama.cpp | +| `havpe-relay` | ESP32 audio bridge | + +### Examples + +```bash +# Deploy speaker recognition on a GPU machine +curl -sSL https://raw.githubusercontent.com/SimpleOpenSoftware/chronicle/main/edge/install.sh \ + | bash -s -- speaker-recognition + +# Deploy from a specific branch +curl -sSL https://raw.githubusercontent.com/SimpleOpenSoftware/chronicle/feat/tailscale-discovery/edge/install.sh \ + | bash -s -- havpe-relay --branch feat/tailscale-discovery + +# Custom install directory (default: ~/.chronicle) +CHRONICLE_HOME=~/my-services curl -sSL ... | bash -s -- havpe-relay +``` + +## What Happens + +1. Clones the repo to `~/.chronicle/` (or `$CHRONICLE_HOME`) +2. Runs the service's interactive config wizard (API keys, credentials, etc.) +3. Starts the service + an edge-agent sidecar container +4. The edge-agent advertises the service on your Tailnet via minidisc +5. Service appears on the **Network** page of your Chronicle dashboard + +## Manage Edge Services + +```bash +cd ~/.chronicle/extras/ + +# Status +docker compose --profile edge ps + +# Logs +docker compose --profile edge logs -f + +# Stop +docker compose --profile edge down + +# Restart +docker compose --profile edge up -d +``` + +## How It Works + +The edge-agent is a tiny Docker sidecar that advertises the service on your Tailnet using minidisc. The main Chronicle backend discovers it automatically — no manual IP configuration needed. + +``` +RPi / GPU VM Main Server +───────────── ─────────── +Docker: speaker-service Docker: chronicle-backend +Docker: edge-agent (sidecar) ←TS→ GET /api/system/network + ↓ ↓ + advertises on Tailnet Network page shows node +``` diff --git a/edge/agent.py b/edge/agent.py new file mode 100644 index 00000000..b4feae64 --- /dev/null +++ b/edge/agent.py @@ -0,0 +1,69 @@ +""" +Edge agent sidecar — advertises a Chronicle service on the Tailnet via minidisc. + +Runs as a Docker container alongside the service it advertises. +Env vars: + EDGE_SERVICE_NAME — minidisc service name (e.g. chronicle-speaker) + EDGE_SERVICE_PORT — port to advertise (e.g. 8085) +""" + +import logging +import os +import signal +import socket +import sys +import time + +logging.basicConfig( + level=logging.INFO, + format="%(asctime)s [edge-agent] %(message)s", + datefmt="%H:%M:%S", +) +logger = logging.getLogger(__name__) + +shutdown = False + + +def _handle_signal(signum, frame): + global shutdown + logger.info("Received signal %s, shutting down", signal.Signals(signum).name) + shutdown = True + + +def main(): + global shutdown + + service_name = os.environ.get("EDGE_SERVICE_NAME") + service_port = os.environ.get("EDGE_SERVICE_PORT") + + if not service_name or not service_port: + logger.error("EDGE_SERVICE_NAME and EDGE_SERVICE_PORT must be set") + sys.exit(1) + + port = int(service_port) + hostname = socket.gethostname() + + signal.signal(signal.SIGTERM, _handle_signal) + signal.signal(signal.SIGINT, _handle_signal) + + try: + import minidisc + except ImportError: + logger.error("minidisc-python not installed") + sys.exit(1) + + logger.info("Starting minidisc registry...") + registry = minidisc.start_registry() + + labels = {"type": "edge", "host": hostname} + registry.advertise_service(port, service_name, labels) + logger.info("Advertising %s on port %d (labels: %s)", service_name, port, labels) + + while not shutdown: + time.sleep(60) + + logger.info("Edge agent stopped") + + +if __name__ == "__main__": + main() diff --git a/edge/install.sh b/edge/install.sh new file mode 100755 index 00000000..2168c71a --- /dev/null +++ b/edge/install.sh @@ -0,0 +1,181 @@ +#!/usr/bin/env bash +# Chronicle Edge — one-liner service deployment on remote nodes. +# +# Usage: +# curl -sSL https://raw.githubusercontent.com/.../edge/install.sh | bash -s -- speaker-recognition +# curl -sSL ... | bash -s -- asr-services --branch dev +# +# Prerequisites: docker (with compose), tailscale (connected), uv, git +set -euo pipefail + +# ── Defaults ────────────────────────────────────────────────────────── +CHRONICLE_HOME="${CHRONICLE_HOME:-$HOME/.chronicle}" +BRANCH="main" +REPO_URL="https://github.com/SimpleOpenSoftware/chronicle.git" + +# ── Colours ─────────────────────────────────────────────────────────── +RED='\033[0;31m'; GREEN='\033[0;32m'; CYAN='\033[0;36m'; YELLOW='\033[1;33m'; NC='\033[0m' +info() { echo -e "${CYAN}[edge]${NC} $*"; } +ok() { echo -e "${GREEN}[edge]${NC} $*"; } +warn() { echo -e "${YELLOW}[edge]${NC} $*"; } +err() { echo -e "${RED}[edge]${NC} $*" >&2; } + +# ── Parse args ──────────────────────────────────────────────────────── +SERVICE_NAME="" +while [[ $# -gt 0 ]]; do + case "$1" in + --branch) BRANCH="$2"; shift 2 ;; + --repo) REPO_URL="$2"; shift 2 ;; + --help|-h) + echo "Usage: $0 [--branch ] [--repo ]" + echo "" + echo "Available services:" + echo " speaker-recognition Speaker Recognition Service" + echo " asr-services ASR Speech-to-Text Service" + echo " tts Text-to-Speech Service" + echo " llm-services Local LLM via llama.cpp" + echo " havpe-relay HAVPE Audio Relay" + exit 0 + ;; + -*) err "Unknown option: $1"; exit 1 ;; + *) SERVICE_NAME="$1"; shift ;; + esac +done + +if [[ -z "$SERVICE_NAME" ]]; then + err "Service name required. Run with --help for usage." + exit 1 +fi + +# ── Validate service name against registry ──────────────────────────── +# We'll validate after clone, but do a quick sanity check here. +KNOWN_SERVICES="speaker-recognition asr-services tts llm-services havpe-relay" +if ! echo "$KNOWN_SERVICES" | grep -qw "$SERVICE_NAME"; then + err "Unknown service: $SERVICE_NAME" + err "Available: $KNOWN_SERVICES" + exit 1 +fi + +# ── Check prerequisites ────────────────────────────────────────────── +check_cmd() { + if ! command -v "$1" &>/dev/null; then + err "Required command not found: $1" + err "Please install $1 before running this script." + exit 1 + fi +} + +info "Checking prerequisites..." +check_cmd docker +check_cmd git +check_cmd uv + +# Check docker compose (plugin or standalone) +if docker compose version &>/dev/null; then + COMPOSE="docker compose" +elif command -v docker-compose &>/dev/null; then + COMPOSE="docker-compose" +else + err "docker compose not found. Install Docker Compose." + exit 1 +fi + +# Check Tailscale +if ! command -v tailscale &>/dev/null; then + err "Tailscale not found. Install from https://tailscale.com/download" + exit 1 +fi + +if ! tailscale status &>/dev/null; then + err "Tailscale is not connected. Run: sudo tailscale up" + exit 1 +fi + +TAILSCALE_IP=$(tailscale ip -4 2>/dev/null || echo "unknown") +ok "Prerequisites OK (Tailscale IP: $TAILSCALE_IP)" + +# ── Clone / update repo ────────────────────────────────────────────── +if [[ -d "$CHRONICLE_HOME/.git" ]]; then + info "Updating existing clone at $CHRONICLE_HOME..." + cd "$CHRONICLE_HOME" + git fetch origin "$BRANCH" --depth 1 + git checkout "$BRANCH" 2>/dev/null || git checkout -b "$BRANCH" "origin/$BRANCH" + git reset --hard "origin/$BRANCH" +else + info "Cloning Chronicle (branch: $BRANCH) to $CHRONICLE_HOME..." + git clone --depth 1 --branch "$BRANCH" "$REPO_URL" "$CHRONICLE_HOME" +fi +cd "$CHRONICLE_HOME" +ok "Repository ready at $CHRONICLE_HOME" + +# ── Read service metadata from registry ─────────────────────────────── +# Parse compose_path from edge/services.yml using Python (available via uv) +COMPOSE_PATH=$(uv run python3 -c " +import yaml, sys +with open('edge/services.yml') as f: + data = yaml.safe_load(f) +svc = data.get('services', {}).get('$SERVICE_NAME') +if not svc: + print('NOT_FOUND', file=sys.stderr); sys.exit(1) +print(svc['compose_path']) +") + +if [[ -z "$COMPOSE_PATH" || ! -d "$COMPOSE_PATH" ]]; then + err "Service directory not found: $COMPOSE_PATH" + exit 1 +fi + +SERVICE_DIR="$CHRONICLE_HOME/$COMPOSE_PATH" +info "Service directory: $SERVICE_DIR" + +# ── Run init.py if it exists (interactive config) ───────────────────── +INIT_SCRIPT="$SERVICE_DIR/init.py" +if [[ -f "$INIT_SCRIPT" ]]; then + info "Running configuration wizard for $SERVICE_NAME..." + cd "$SERVICE_DIR" + uv run --with-requirements "$CHRONICLE_HOME/setup-requirements.txt" python init.py + cd "$CHRONICLE_HOME" +elif [[ -f "$SERVICE_DIR/setup.sh" ]]; then + info "Running setup script for $SERVICE_NAME..." + cd "$SERVICE_DIR" + bash setup.sh + cd "$CHRONICLE_HOME" +else + warn "No init script found — using defaults." +fi + +# ── Create Docker network ───────────────────────────────────────────── +docker network create chronicle-network 2>/dev/null || true + +# ── Start service + edge agent ──────────────────────────────────────── +cd "$SERVICE_DIR" +info "Starting $SERVICE_NAME + edge agent..." + +# Determine GPU profile if applicable (same logic as services.py) +PROFILES="--profile edge" +if [[ -f .env ]]; then + PYTORCH_VERSION=$(grep -s '^PYTORCH_CUDA_VERSION=' .env | cut -d= -f2 | tr -d "'" | tr -d '"' || echo "") + if [[ "$PYTORCH_VERSION" == "strixhalo" ]]; then + PROFILES="$PROFILES --profile strixhalo" + elif [[ "$PYTORCH_VERSION" == cu* ]]; then + PROFILES="$PROFILES --profile gpu" + elif [[ -n "$PYTORCH_VERSION" ]]; then + PROFILES="$PROFILES --profile cpu" + fi +fi + +$COMPOSE $PROFILES up --build -d + +ok "────────────────────────────────────────" +ok " $SERVICE_NAME is running!" +ok "" +ok " Tailscale IP: $TAILSCALE_IP" +ok " Service dir: $SERVICE_DIR" +ok "" +ok " Check status: cd $SERVICE_DIR && $COMPOSE ps" +ok " View logs: cd $SERVICE_DIR && $COMPOSE logs -f" +ok " Stop: cd $SERVICE_DIR && $COMPOSE $PROFILES down" +ok "" +ok " The service should appear on the Network page" +ok " of your Chronicle backend dashboard." +ok "────────────────────────────────────────" diff --git a/edge/run-on-rpi.sh b/edge/run-on-rpi.sh new file mode 100755 index 00000000..82321a33 --- /dev/null +++ b/edge/run-on-rpi.sh @@ -0,0 +1,5 @@ +#!/usr/bin/env bash +# Quick deploy havpe-relay on RPi from feat/tailscale-discovery branch +CHRONICLE_HOME=~/my-services \ + curl -sSL https://raw.githubusercontent.com/SimpleOpenSoftware/chronicle/feat/tailscale-discovery/edge/install.sh \ + | bash -s -- havpe-relay --branch feat/tailscale-discovery diff --git a/edge/services.yml b/edge/services.yml new file mode 100644 index 00000000..48ec689b --- /dev/null +++ b/edge/services.yml @@ -0,0 +1,33 @@ +# Edge service registry +# Maps CLI service names to deployment metadata for edge node deployment. +services: + speaker-recognition: + minidisc_name: chronicle-speaker + port: 8085 + health_path: /health + compose_path: extras/speaker-recognition + description: Speaker Recognition Service + asr-services: + minidisc_name: chronicle-asr + port: 8767 + health_path: /health + compose_path: extras/asr-services + description: ASR Speech-to-Text Service + tts: + minidisc_name: chronicle-tts + port: 8770 + health_path: /health + compose_path: extras/tts + description: Text-to-Speech Service + llm-services: + minidisc_name: chronicle-llm + port: 8083 + health_path: /health + compose_path: extras/llm-services + description: Local LLM via llama.cpp + havpe-relay: + minidisc_name: chronicle-relay + port: 0 + health_path: / + compose_path: extras/havpe-relay + description: HAVPE Audio Relay diff --git a/edge/status.sh b/edge/status.sh new file mode 100755 index 00000000..a7490377 --- /dev/null +++ b/edge/status.sh @@ -0,0 +1,49 @@ +#!/usr/bin/env bash +# Show status of an edge-deployed Chronicle service. +# +# Usage: ./status.sh [service-name] +# If no service name given, shows Tailscale IP only. +set -euo pipefail + +CHRONICLE_HOME="${CHRONICLE_HOME:-$HOME/.chronicle}" + +echo "=== Chronicle Edge Status ===" +echo "" + +# Tailscale info +if command -v tailscale &>/dev/null; then + TAILSCALE_IP=$(tailscale ip -4 2>/dev/null || echo "not connected") + echo "Tailscale IP: $TAILSCALE_IP" +else + echo "Tailscale: not installed" +fi +echo "" + +if [[ $# -lt 1 ]]; then + echo "Usage: $0 to see container status" + echo "Available services: speaker-recognition asr-services tts llm-services havpe-relay" + exit 0 +fi + +SERVICE_NAME="$1" + +COMPOSE_PATH=$(uv run python3 -c " +import yaml, sys +with open('$CHRONICLE_HOME/edge/services.yml') as f: + data = yaml.safe_load(f) +svc = data.get('services', {}).get('$SERVICE_NAME') +if not svc: + print('NOT_FOUND', file=sys.stderr); sys.exit(1) +print(svc['compose_path']) +" 2>/dev/null || echo "") + +SERVICE_DIR="$CHRONICLE_HOME/$COMPOSE_PATH" + +if [[ -z "$COMPOSE_PATH" || ! -d "$SERVICE_DIR" ]]; then + echo "Service not found or not installed: $SERVICE_NAME" + exit 1 +fi + +cd "$SERVICE_DIR" +echo "=== $SERVICE_NAME containers ===" +docker compose --profile edge ps diff --git a/edge/stop.sh b/edge/stop.sh new file mode 100755 index 00000000..e097767b --- /dev/null +++ b/edge/stop.sh @@ -0,0 +1,37 @@ +#!/usr/bin/env bash +# Stop an edge-deployed Chronicle service. +# +# Usage: ./stop.sh +set -euo pipefail + +CHRONICLE_HOME="${CHRONICLE_HOME:-$HOME/.chronicle}" + +if [[ $# -lt 1 ]]; then + echo "Usage: $0 " + echo "Example: $0 speaker-recognition" + exit 1 +fi + +SERVICE_NAME="$1" + +# Parse compose_path from registry +COMPOSE_PATH=$(uv run python3 -c " +import yaml, sys +with open('$CHRONICLE_HOME/edge/services.yml') as f: + data = yaml.safe_load(f) +svc = data.get('services', {}).get('$SERVICE_NAME') +if not svc: + print('NOT_FOUND', file=sys.stderr); sys.exit(1) +print(svc['compose_path']) +" 2>/dev/null || echo "") + +SERVICE_DIR="$CHRONICLE_HOME/$COMPOSE_PATH" + +if [[ -z "$COMPOSE_PATH" || ! -d "$SERVICE_DIR" ]]; then + echo "Service not found or not installed: $SERVICE_NAME" + exit 1 +fi + +cd "$SERVICE_DIR" +docker compose --profile edge down +echo "$SERVICE_NAME stopped." diff --git a/extras/asr-services/docker-compose.yml b/extras/asr-services/docker-compose.yml index fb2c016a..98e16b52 100644 --- a/extras/asr-services/docker-compose.yml +++ b/extras/asr-services/docker-compose.yml @@ -347,3 +347,16 @@ services: - MIN_AUDIO_FOR_CHUNKING=${MIN_AUDIO_FOR_CHUNKING:-60.0} - CONFIDENCE_THRESHOLD=${CONFIDENCE_THRESHOLD:-0.8} restart: unless-stopped + + # Edge agent sidecar — advertises this service on the Tailnet + edge-agent: + build: + context: ../../edge + environment: + - EDGE_SERVICE_NAME=chronicle-asr + - EDGE_SERVICE_PORT=8767 + volumes: + - /var/run/tailscale/tailscaled.sock:/var/run/tailscale/tailscaled.sock:ro + restart: unless-stopped + profiles: + - edge diff --git a/extras/havpe-relay/docker-compose.yml b/extras/havpe-relay/docker-compose.yml index 2b4e3090..c59d5759 100644 --- a/extras/havpe-relay/docker-compose.yml +++ b/extras/havpe-relay/docker-compose.yml @@ -24,3 +24,16 @@ services: - "host.docker.internal:host-gateway" volumes: - ./audio_chunks:/app/audio_chunks + + # Edge agent sidecar — advertises this service on the Tailnet + edge-agent: + build: + context: ../../edge + environment: + - EDGE_SERVICE_NAME=chronicle-relay + - EDGE_SERVICE_PORT=8989 + volumes: + - /var/run/tailscale/tailscaled.sock:/var/run/tailscale/tailscaled.sock:ro + restart: unless-stopped + profiles: + - edge diff --git a/extras/havpe-relay/init.py b/extras/havpe-relay/init.py index efb757d8..a08e8bc6 100644 --- a/extras/havpe-relay/init.py +++ b/extras/havpe-relay/init.py @@ -91,6 +91,21 @@ def backup_existing_env(self): f"[blue][INFO][/blue] Backed up existing .env file to {backup_path}" ) + def _try_discover_backend(self) -> str | None: + """Try to auto-discover the Chronicle backend via minidisc on Tailnet.""" + try: + sys.path.insert(0, str(Path(__file__).resolve().parent.parent.parent)) + from discovery import CHRONICLE_BACKEND, discover_service + + url = discover_service(CHRONICLE_BACKEND) + if url: + self.console.print( + f"[green][SUCCESS][/green] Auto-discovered backend on Tailnet: {url}" + ) + return url + except Exception: + return None + def setup_backend_urls(self): """Configure backend URL and WebSocket URL""" self.print_section("Backend Connection") @@ -100,6 +115,11 @@ def setup_backend_urls(self): default_http = "http://host.docker.internal:8000" default_ws = "ws://host.docker.internal:8000" + # Try auto-discovery via minidisc first + discovered = self._try_discover_backend() + if discovered: + default_http = discovered + # Check CLI args first if hasattr(self.args, "backend_url") and self.args.backend_url: backend_url = self.args.backend_url diff --git a/extras/havpe-relay/pyproject.toml b/extras/havpe-relay/pyproject.toml index 5e5dd3e8..8b74d86e 100644 --- a/extras/havpe-relay/pyproject.toml +++ b/extras/havpe-relay/pyproject.toml @@ -10,6 +10,7 @@ dependencies = [ "rumps>=0.4.0; sys_platform == 'darwin'", "python-dotenv>=1.0.0", "aioesphomeapi>=30.2.0", + "minidisc-python>=0.1.0", ] [dependency-groups] diff --git a/extras/havpe-relay/relay_core.py b/extras/havpe-relay/relay_core.py index 9c073098..ae136603 100644 --- a/extras/havpe-relay/relay_core.py +++ b/extras/havpe-relay/relay_core.py @@ -33,9 +33,37 @@ class RelayConfig: @classmethod def from_env(cls) -> "RelayConfig": + backend_url = os.getenv("BACKEND_URL") + backend_ws_url = os.getenv("BACKEND_WS_URL") + + # Try minidisc service discovery when env vars are not set + if not backend_url: + try: + import sys + from pathlib import Path + + # discovery.py lives at the repo root (two levels up) + _repo_root = str(Path(__file__).resolve().parent.parent.parent) + if _repo_root not in sys.path: + sys.path.insert(0, _repo_root) + + from discovery import CHRONICLE_BACKEND, discover_service + + discovered = discover_service(CHRONICLE_BACKEND) + if discovered: + backend_url = discovered + backend_ws_url = discovered.replace("http://", "ws://").replace( + "https://", "wss://" + ) + logger.info( + "Discovered Chronicle backend via minidisc: %s", discovered + ) + except ImportError: + pass + return cls( - backend_url=os.getenv("BACKEND_URL", "http://localhost:8000"), - backend_ws_url=os.getenv("BACKEND_WS_URL", "ws://localhost:8000"), + backend_url=backend_url or "http://localhost:8000", + backend_ws_url=backend_ws_url or "ws://localhost:8000", auth_username=os.getenv("AUTH_USERNAME", ""), auth_password=os.getenv("AUTH_PASSWORD", ""), device_name=os.getenv("DEVICE_NAME", "havpe"), diff --git a/extras/llm-services/docker-compose.yml b/extras/llm-services/docker-compose.yml index 22e714ec..50fb61be 100644 --- a/extras/llm-services/docker-compose.yml +++ b/extras/llm-services/docker-compose.yml @@ -44,6 +44,19 @@ services: networks: - chronicle-network + # Edge agent sidecar — advertises this service on the Tailnet + edge-agent: + build: + context: ../../edge + environment: + - EDGE_SERVICE_NAME=chronicle-llm + - EDGE_SERVICE_PORT=8083 + volumes: + - /var/run/tailscale/tailscaled.sock:/var/run/tailscale/tailscaled.sock:ro + restart: unless-stopped + profiles: + - edge + networks: chronicle-network: external: true diff --git a/extras/speaker-recognition/docker-compose.yml b/extras/speaker-recognition/docker-compose.yml index 89cc58c0..fb92c765 100644 --- a/extras/speaker-recognition/docker-compose.yml +++ b/extras/speaker-recognition/docker-compose.yml @@ -137,6 +137,19 @@ services: condition: service_healthy restart: unless-stopped + # Edge agent sidecar — advertises this service on the Tailnet + edge-agent: + build: + context: ../../edge + environment: + - EDGE_SERVICE_NAME=chronicle-speaker + - EDGE_SERVICE_PORT=8085 + volumes: + - /var/run/tailscale/tailscaled.sock:/var/run/tailscale/tailscaled.sock:ro + restart: unless-stopped + profiles: + - edge + # Shared network for cross-project communication networks: default: From 891134ee1c3b2e852ff64328ff7f327d54ebee38 Mon Sep 17 00:00:00 2001 From: Ankush Malaker <43288948+AnkushMalaker@users.noreply.github.com> Date: Sun, 15 Mar 2026 22:15:35 +0000 Subject: [PATCH 02/25] tiny fixes --- edge/README.md | 6 +++--- edge/install.sh | 25 +++++++++++++++++++++---- edge/run-on-rpi.sh | 2 +- edge/status.sh | 2 +- edge/stop.sh | 2 +- 5 files changed, 27 insertions(+), 10 deletions(-) diff --git a/edge/README.md b/edge/README.md index e3f740b6..8a4de2e1 100644 --- a/edge/README.md +++ b/edge/README.md @@ -38,13 +38,13 @@ curl -sSL https://raw.githubusercontent.com/SimpleOpenSoftware/chronicle/main/ed curl -sSL https://raw.githubusercontent.com/SimpleOpenSoftware/chronicle/feat/tailscale-discovery/edge/install.sh \ | bash -s -- havpe-relay --branch feat/tailscale-discovery -# Custom install directory (default: ~/.chronicle) +# Custom install directory (default: ~/chronicle) CHRONICLE_HOME=~/my-services curl -sSL ... | bash -s -- havpe-relay ``` ## What Happens -1. Clones the repo to `~/.chronicle/` (or `$CHRONICLE_HOME`) +1. Clones the repo to `~/chronicle/` (or `$CHRONICLE_HOME`) 2. Runs the service's interactive config wizard (API keys, credentials, etc.) 3. Starts the service + an edge-agent sidecar container 4. The edge-agent advertises the service on your Tailnet via minidisc @@ -53,7 +53,7 @@ CHRONICLE_HOME=~/my-services curl -sSL ... | bash -s -- havpe-relay ## Manage Edge Services ```bash -cd ~/.chronicle/extras/ +cd ~/chronicle/extras/ # Status docker compose --profile edge ps diff --git a/edge/install.sh b/edge/install.sh index 2168c71a..dc7a6222 100755 --- a/edge/install.sh +++ b/edge/install.sh @@ -9,10 +9,20 @@ set -euo pipefail # ── Defaults ────────────────────────────────────────────────────────── -CHRONICLE_HOME="${CHRONICLE_HOME:-$HOME/.chronicle}" BRANCH="main" REPO_URL="https://github.com/SimpleOpenSoftware/chronicle.git" +# Resolve CHRONICLE_HOME: explicit env var > detect existing clone > default +if [[ -n "${CHRONICLE_HOME:-}" ]]; then + : # User explicitly set it +elif [[ -d "$HOME/chronicle/.git" ]]; then + CHRONICLE_HOME="$HOME/chronicle" +elif [[ -d "$PWD/.git" && -f "$PWD/edge/services.yml" ]]; then + CHRONICLE_HOME="$PWD" +else + CHRONICLE_HOME="$HOME/chronicle" +fi + # ── Colours ─────────────────────────────────────────────────────────── RED='\033[0;31m'; GREEN='\033[0;32m'; CYAN='\033[0;36m'; YELLOW='\033[1;33m'; NC='\033[0m' info() { echo -e "${CYAN}[edge]${NC} $*"; } @@ -98,9 +108,16 @@ ok "Prerequisites OK (Tailscale IP: $TAILSCALE_IP)" if [[ -d "$CHRONICLE_HOME/.git" ]]; then info "Updating existing clone at $CHRONICLE_HOME..." cd "$CHRONICLE_HOME" - git fetch origin "$BRANCH" --depth 1 - git checkout "$BRANCH" 2>/dev/null || git checkout -b "$BRANCH" "origin/$BRANCH" - git reset --hard "origin/$BRANCH" + git fetch origin "$BRANCH" + # Switch to branch if not already on it + current_branch=$(git rev-parse --abbrev-ref HEAD 2>/dev/null || echo "") + if [[ "$current_branch" != "$BRANCH" ]]; then + git checkout "$BRANCH" 2>/dev/null || git checkout -b "$BRANCH" "origin/$BRANCH" + fi + # Pull with rebase, but don't fail if there are local changes + git pull --rebase --autostash origin "$BRANCH" || { + warn "Could not auto-update (you may have local changes). Continuing with current state." + } else info "Cloning Chronicle (branch: $BRANCH) to $CHRONICLE_HOME..." git clone --depth 1 --branch "$BRANCH" "$REPO_URL" "$CHRONICLE_HOME" diff --git a/edge/run-on-rpi.sh b/edge/run-on-rpi.sh index 82321a33..0e347a34 100755 --- a/edge/run-on-rpi.sh +++ b/edge/run-on-rpi.sh @@ -1,5 +1,5 @@ #!/usr/bin/env bash # Quick deploy havpe-relay on RPi from feat/tailscale-discovery branch -CHRONICLE_HOME=~/my-services \ +CHRONICLE_HOME=~/my-services/chronicle \ curl -sSL https://raw.githubusercontent.com/SimpleOpenSoftware/chronicle/feat/tailscale-discovery/edge/install.sh \ | bash -s -- havpe-relay --branch feat/tailscale-discovery diff --git a/edge/status.sh b/edge/status.sh index a7490377..1e9be004 100755 --- a/edge/status.sh +++ b/edge/status.sh @@ -5,7 +5,7 @@ # If no service name given, shows Tailscale IP only. set -euo pipefail -CHRONICLE_HOME="${CHRONICLE_HOME:-$HOME/.chronicle}" +CHRONICLE_HOME="${CHRONICLE_HOME:-$HOME/chronicle}" echo "=== Chronicle Edge Status ===" echo "" diff --git a/edge/stop.sh b/edge/stop.sh index e097767b..1365ced1 100755 --- a/edge/stop.sh +++ b/edge/stop.sh @@ -4,7 +4,7 @@ # Usage: ./stop.sh set -euo pipefail -CHRONICLE_HOME="${CHRONICLE_HOME:-$HOME/.chronicle}" +CHRONICLE_HOME="${CHRONICLE_HOME:-$HOME/chronicle}" if [[ $# -lt 1 ]]; then echo "Usage: $0 " From 259007a210f189e24945ea9be5553d5dac4b92ef Mon Sep 17 00:00:00 2001 From: Ankush Malaker <43288948+AnkushMalaker@users.noreply.github.com> Date: Sun, 15 Mar 2026 22:19:06 +0000 Subject: [PATCH 03/25] tiny fix again --- edge/install.sh | 3 +-- edge/run-on-rpi.sh | 4 ++-- edge/status.sh | 4 ++-- edge/stop.sh | 4 ++-- 4 files changed, 7 insertions(+), 8 deletions(-) diff --git a/edge/install.sh b/edge/install.sh index dc7a6222..e72a487e 100755 --- a/edge/install.sh +++ b/edge/install.sh @@ -126,8 +126,7 @@ cd "$CHRONICLE_HOME" ok "Repository ready at $CHRONICLE_HOME" # ── Read service metadata from registry ─────────────────────────────── -# Parse compose_path from edge/services.yml using Python (available via uv) -COMPOSE_PATH=$(uv run python3 -c " +COMPOSE_PATH=$(uv run --with-requirements setup-requirements.txt python3 -c " import yaml, sys with open('edge/services.yml') as f: data = yaml.safe_load(f) diff --git a/edge/run-on-rpi.sh b/edge/run-on-rpi.sh index 0e347a34..5a9685a4 100755 --- a/edge/run-on-rpi.sh +++ b/edge/run-on-rpi.sh @@ -1,5 +1,5 @@ #!/usr/bin/env bash # Quick deploy havpe-relay on RPi from feat/tailscale-discovery branch -CHRONICLE_HOME=~/my-services/chronicle \ - curl -sSL https://raw.githubusercontent.com/SimpleOpenSoftware/chronicle/feat/tailscale-discovery/edge/install.sh \ +export CHRONICLE_HOME=~/my-services/chronicle +curl -sSL https://raw.githubusercontent.com/SimpleOpenSoftware/chronicle/feat/tailscale-discovery/edge/install.sh \ | bash -s -- havpe-relay --branch feat/tailscale-discovery diff --git a/edge/status.sh b/edge/status.sh index 1e9be004..506c6551 100755 --- a/edge/status.sh +++ b/edge/status.sh @@ -27,9 +27,9 @@ fi SERVICE_NAME="$1" -COMPOSE_PATH=$(uv run python3 -c " +COMPOSE_PATH=$(cd "$CHRONICLE_HOME" && uv run --with-requirements setup-requirements.txt python3 -c " import yaml, sys -with open('$CHRONICLE_HOME/edge/services.yml') as f: +with open('edge/services.yml') as f: data = yaml.safe_load(f) svc = data.get('services', {}).get('$SERVICE_NAME') if not svc: diff --git a/edge/stop.sh b/edge/stop.sh index 1365ced1..2616c9e1 100755 --- a/edge/stop.sh +++ b/edge/stop.sh @@ -15,9 +15,9 @@ fi SERVICE_NAME="$1" # Parse compose_path from registry -COMPOSE_PATH=$(uv run python3 -c " +COMPOSE_PATH=$(cd "$CHRONICLE_HOME" && uv run --with-requirements setup-requirements.txt python3 -c " import yaml, sys -with open('$CHRONICLE_HOME/edge/services.yml') as f: +with open('edge/services.yml') as f: data = yaml.safe_load(f) svc = data.get('services', {}).get('$SERVICE_NAME') if not svc: From 9f23f62f30a6311de8d685f85e0de79c68e86eb4 Mon Sep 17 00:00:00 2001 From: Ankush Malaker <43288948+AnkushMalaker@users.noreply.github.com> Date: Sun, 15 Mar 2026 22:25:00 +0000 Subject: [PATCH 04/25] minor fix --- edge/install.sh | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/edge/install.sh b/edge/install.sh index e72a487e..531b9fa8 100755 --- a/edge/install.sh +++ b/edge/install.sh @@ -105,7 +105,10 @@ TAILSCALE_IP=$(tailscale ip -4 2>/dev/null || echo "unknown") ok "Prerequisites OK (Tailscale IP: $TAILSCALE_IP)" # ── Clone / update repo ────────────────────────────────────────────── -if [[ -d "$CHRONICLE_HOME/.git" ]]; then +if [[ -f "$CHRONICLE_HOME/edge/services.yml" ]]; then + # Already inside a working repo — skip git entirely + info "Using existing repo at $CHRONICLE_HOME" +elif [[ -d "$CHRONICLE_HOME/.git" ]]; then info "Updating existing clone at $CHRONICLE_HOME..." cd "$CHRONICLE_HOME" git fetch origin "$BRANCH" @@ -114,7 +117,6 @@ if [[ -d "$CHRONICLE_HOME/.git" ]]; then if [[ "$current_branch" != "$BRANCH" ]]; then git checkout "$BRANCH" 2>/dev/null || git checkout -b "$BRANCH" "origin/$BRANCH" fi - # Pull with rebase, but don't fail if there are local changes git pull --rebase --autostash origin "$BRANCH" || { warn "Could not auto-update (you may have local changes). Continuing with current state." } From 338fd880d851007f0c63e8976aa4b10be486ab25 Mon Sep 17 00:00:00 2001 From: Ankush Malaker <43288948+AnkushMalaker@users.noreply.github.com> Date: Sun, 15 Mar 2026 22:30:20 +0000 Subject: [PATCH 05/25] auto discover --- edge/install.sh | 37 ++++++++++++++++++++++++++++++++++++- 1 file changed, 36 insertions(+), 1 deletion(-) diff --git a/edge/install.sh b/edge/install.sh index 531b9fa8..874ad2fb 100755 --- a/edge/install.sh +++ b/edge/install.sh @@ -146,12 +146,47 @@ fi SERVICE_DIR="$CHRONICLE_HOME/$COMPOSE_PATH" info "Service directory: $SERVICE_DIR" +# ── Resolve backend URL for edge services ───────────────────────────── +# On edge nodes, host.docker.internal doesn't reach the backend. +# Try minidisc discovery first, then prompt if needed. +info "Looking for Chronicle backend on Tailnet..." +BACKEND_URL=$(uv run --with-requirements "$CHRONICLE_HOME/setup-requirements.txt" python3 -c " +import sys +sys.path.insert(0, '$CHRONICLE_HOME') +try: + from discovery import CHRONICLE_BACKEND, discover_service + url = discover_service(CHRONICLE_BACKEND) + if url: + print(url, end='') + else: + print('', end='', file=sys.stderr) + print('minidisc found no chronicle-backend service on Tailnet', file=sys.stderr) +except Exception as e: + print(f'discovery failed: {e}', file=sys.stderr) +") + +if [[ -n "$BACKEND_URL" ]]; then + ok "Auto-discovered backend at $BACKEND_URL" +else + warn "Could not auto-discover backend. Make sure your Chronicle backend is running with Tailscale." + read -rp "[edge] Enter your Chronicle backend URL (e.g. http://100.x.x.x:8000): " BACKEND_URL + if [[ -z "$BACKEND_URL" ]]; then + err "Backend URL is required for edge deployment." + exit 1 + fi +fi + # ── Run init.py if it exists (interactive config) ───────────────────── +INIT_ARGS="" +if [[ -n "$BACKEND_URL" ]]; then + INIT_ARGS="--backend-url $BACKEND_URL" +fi + INIT_SCRIPT="$SERVICE_DIR/init.py" if [[ -f "$INIT_SCRIPT" ]]; then info "Running configuration wizard for $SERVICE_NAME..." cd "$SERVICE_DIR" - uv run --with-requirements "$CHRONICLE_HOME/setup-requirements.txt" python init.py + uv run --with-requirements "$CHRONICLE_HOME/setup-requirements.txt" python init.py $INIT_ARGS cd "$CHRONICLE_HOME" elif [[ -f "$SERVICE_DIR/setup.sh" ]]; then info "Running setup script for $SERVICE_NAME..." From c7060eba305c71dd47abde2d6a3b6bbcdc977800 Mon Sep 17 00:00:00 2001 From: Ankush Malaker <43288948+AnkushMalaker@users.noreply.github.com> Date: Sun, 15 Mar 2026 22:58:17 +0000 Subject: [PATCH 06/25] reworked --- edge/agent.py | 90 +++++++++++++++---- edge/install.sh | 61 ++++++------- edge/services.yml | 33 ------- edge/status.sh | 28 +++--- edge/stop.sh | 31 ++++--- extras/asr-services/docker-compose.yml | 6 +- extras/havpe-relay/docker-compose.yml | 6 +- extras/llm-services/docker-compose.yml | 6 +- extras/speaker-recognition/docker-compose.yml | 6 +- 9 files changed, 140 insertions(+), 127 deletions(-) delete mode 100644 edge/services.yml diff --git a/edge/agent.py b/edge/agent.py index b4feae64..0ea0d7f0 100644 --- a/edge/agent.py +++ b/edge/agent.py @@ -1,10 +1,18 @@ """ -Edge agent sidecar — advertises a Chronicle service on the Tailnet via minidisc. +Edge discovery agent — advertises Chronicle services on the Tailnet via minidisc. -Runs as a Docker container alongside the service it advertises. -Env vars: - EDGE_SERVICE_NAME — minidisc service name (e.g. chronicle-speaker) - EDGE_SERVICE_PORT — port to advertise (e.g. 8085) +Config-driven: reads what to advertise from environment variables instead of +probing ports. One agent per machine. + +Two input modes (both contribute to a single service list): + + ADVERTISE — comma-separated name:port pairs (edge nodes + overrides) + e.g. ADVERTISE=chronicle-speaker:8085,chronicle-asr:8767 + + ADVERTISE_BACKEND — when "true", always includes chronicle-backend:8000 and + inspects SPEAKER_SERVICE_URL / PARAKEET_ASR_URL / + OPENMEMORY_MCP_URL / TTS_SERVICE_URL to detect co-located + services (main server mode). """ import logging @@ -16,13 +24,21 @@ logging.basicConfig( level=logging.INFO, - format="%(asctime)s [edge-agent] %(message)s", + format="%(asctime)s [discovery-agent] %(message)s", datefmt="%H:%M:%S", ) logger = logging.getLogger(__name__) shutdown = False +# Mapping from env-var URL → (service name, port) +_BACKEND_SERVICE_MAP = { + "SPEAKER_SERVICE_URL": ("chronicle-speaker", 8085), + "PARAKEET_ASR_URL": ("chronicle-asr", 8767), + "OPENMEMORY_MCP_URL": ("chronicle-openmemory", 8765), + "TTS_SERVICE_URL": ("chronicle-tts", 8770), +} + def _handle_signal(signum, frame): global shutdown @@ -30,17 +46,56 @@ def _handle_signal(signum, frame): shutdown = True +def _collect_services() -> list[tuple[str, int]]: + """Build the list of (name, port) pairs to advertise.""" + services: list[tuple[str, int]] = [] + seen: set[str] = set() + + def _add(name: str, port: int): + if name not in seen: + services.append((name, port)) + seen.add(name) + + # Mode 1: explicit ADVERTISE env var + advertise = os.environ.get("ADVERTISE", "").strip() + if advertise: + for entry in advertise.split(","): + entry = entry.strip() + if not entry: + continue + if ":" not in entry: + logger.warning( + "Skipping invalid ADVERTISE entry (missing port): %s", entry + ) + continue + name, port_str = entry.rsplit(":", 1) + try: + _add(name.strip(), int(port_str.strip())) + except ValueError: + logger.warning("Skipping invalid ADVERTISE entry (bad port): %s", entry) + + # Mode 2: backend mode — read service URLs from .env + if os.environ.get("ADVERTISE_BACKEND", "").lower() in ("true", "1", "yes"): + _add("chronicle-backend", 8000) + + for env_var, (svc_name, svc_port) in _BACKEND_SERVICE_MAP.items(): + url = os.environ.get(env_var, "") + if url and "host.docker.internal" in url: + _add(svc_name, svc_port) + + return services + + def main(): global shutdown - service_name = os.environ.get("EDGE_SERVICE_NAME") - service_port = os.environ.get("EDGE_SERVICE_PORT") - - if not service_name or not service_port: - logger.error("EDGE_SERVICE_NAME and EDGE_SERVICE_PORT must be set") + services = _collect_services() + if not services: + logger.error( + "Nothing to advertise. Set ADVERTISE=name:port and/or ADVERTISE_BACKEND=true" + ) sys.exit(1) - port = int(service_port) hostname = socket.gethostname() signal.signal(signal.SIGTERM, _handle_signal) @@ -55,14 +110,17 @@ def main(): logger.info("Starting minidisc registry...") registry = minidisc.start_registry() - labels = {"type": "edge", "host": hostname} - registry.advertise_service(port, service_name, labels) - logger.info("Advertising %s on port %d (labels: %s)", service_name, port, labels) + labels = {"type": "discovery-agent", "host": hostname} + for name, port in services: + registry.advertise_service(port, name, labels) + logger.info("Advertising %s on port %d", name, port) + + logger.info("Discovery agent running — %d service(s) advertised", len(services)) while not shutdown: time.sleep(60) - logger.info("Edge agent stopped") + logger.info("Discovery agent stopped") if __name__ == "__main__": diff --git a/edge/install.sh b/edge/install.sh index 874ad2fb..314948e8 100755 --- a/edge/install.sh +++ b/edge/install.sh @@ -17,12 +17,21 @@ if [[ -n "${CHRONICLE_HOME:-}" ]]; then : # User explicitly set it elif [[ -d "$HOME/chronicle/.git" ]]; then CHRONICLE_HOME="$HOME/chronicle" -elif [[ -d "$PWD/.git" && -f "$PWD/edge/services.yml" ]]; then +elif [[ -d "$PWD/.git" && -d "$PWD/edge" ]]; then CHRONICLE_HOME="$PWD" else CHRONICLE_HOME="$HOME/chronicle" fi +# ── Service → compose path mapping ─────────────────────────────────── +declare -A SERVICE_PATHS=( + [speaker-recognition]=extras/speaker-recognition + [asr-services]=extras/asr-services + [tts]=extras/tts + [llm-services]=extras/llm-services + [havpe-relay]=extras/havpe-relay +) + # ── Colours ─────────────────────────────────────────────────────────── RED='\033[0;31m'; GREEN='\033[0;32m'; CYAN='\033[0;36m'; YELLOW='\033[1;33m'; NC='\033[0m' info() { echo -e "${CYAN}[edge]${NC} $*"; } @@ -40,11 +49,7 @@ while [[ $# -gt 0 ]]; do echo "Usage: $0 [--branch ] [--repo ]" echo "" echo "Available services:" - echo " speaker-recognition Speaker Recognition Service" - echo " asr-services ASR Speech-to-Text Service" - echo " tts Text-to-Speech Service" - echo " llm-services Local LLM via llama.cpp" - echo " havpe-relay HAVPE Audio Relay" + for svc in "${!SERVICE_PATHS[@]}"; do echo " $svc"; done | sort exit 0 ;; -*) err "Unknown option: $1"; exit 1 ;; @@ -57,15 +62,15 @@ if [[ -z "$SERVICE_NAME" ]]; then exit 1 fi -# ── Validate service name against registry ──────────────────────────── -# We'll validate after clone, but do a quick sanity check here. -KNOWN_SERVICES="speaker-recognition asr-services tts llm-services havpe-relay" -if ! echo "$KNOWN_SERVICES" | grep -qw "$SERVICE_NAME"; then +# ── Validate service name ──────────────────────────────────────────── +if [[ -z "${SERVICE_PATHS[$SERVICE_NAME]+_}" ]]; then err "Unknown service: $SERVICE_NAME" - err "Available: $KNOWN_SERVICES" + err "Available: ${!SERVICE_PATHS[*]}" exit 1 fi +COMPOSE_PATH="${SERVICE_PATHS[$SERVICE_NAME]}" + # ── Check prerequisites ────────────────────────────────────────────── check_cmd() { if ! command -v "$1" &>/dev/null; then @@ -105,16 +110,12 @@ TAILSCALE_IP=$(tailscale ip -4 2>/dev/null || echo "unknown") ok "Prerequisites OK (Tailscale IP: $TAILSCALE_IP)" # ── Clone / update repo ────────────────────────────────────────────── -if [[ -f "$CHRONICLE_HOME/edge/services.yml" ]]; then - # Already inside a working repo — skip git entirely +if [[ -d "$CHRONICLE_HOME/.git" && -d "$CHRONICLE_HOME/edge" ]]; then info "Using existing repo at $CHRONICLE_HOME" -elif [[ -d "$CHRONICLE_HOME/.git" ]]; then - info "Updating existing clone at $CHRONICLE_HOME..." cd "$CHRONICLE_HOME" - git fetch origin "$BRANCH" - # Switch to branch if not already on it current_branch=$(git rev-parse --abbrev-ref HEAD 2>/dev/null || echo "") if [[ "$current_branch" != "$BRANCH" ]]; then + git fetch origin "$BRANCH" git checkout "$BRANCH" 2>/dev/null || git checkout -b "$BRANCH" "origin/$BRANCH" fi git pull --rebase --autostash origin "$BRANCH" || { @@ -127,42 +128,30 @@ fi cd "$CHRONICLE_HOME" ok "Repository ready at $CHRONICLE_HOME" -# ── Read service metadata from registry ─────────────────────────────── -COMPOSE_PATH=$(uv run --with-requirements setup-requirements.txt python3 -c " -import yaml, sys -with open('edge/services.yml') as f: - data = yaml.safe_load(f) -svc = data.get('services', {}).get('$SERVICE_NAME') -if not svc: - print('NOT_FOUND', file=sys.stderr); sys.exit(1) -print(svc['compose_path']) -") +# ── Resolve compose path ───────────────────────────────────────────── +SERVICE_DIR="$CHRONICLE_HOME/$COMPOSE_PATH" -if [[ -z "$COMPOSE_PATH" || ! -d "$COMPOSE_PATH" ]]; then - err "Service directory not found: $COMPOSE_PATH" +if [[ ! -d "$SERVICE_DIR" ]]; then + err "Service directory not found: $SERVICE_DIR" exit 1 fi -SERVICE_DIR="$CHRONICLE_HOME/$COMPOSE_PATH" info "Service directory: $SERVICE_DIR" # ── Resolve backend URL for edge services ───────────────────────────── # On edge nodes, host.docker.internal doesn't reach the backend. # Try minidisc discovery first, then prompt if needed. info "Looking for Chronicle backend on Tailnet..." -BACKEND_URL=$(uv run --with-requirements "$CHRONICLE_HOME/setup-requirements.txt" python3 -c " -import sys -sys.path.insert(0, '$CHRONICLE_HOME') +BACKEND_URL=$(cd "$CHRONICLE_HOME" && PYTHONPATH="$CHRONICLE_HOME" uv run --with-requirements "$CHRONICLE_HOME/setup-requirements.txt" python3 -c " try: from discovery import CHRONICLE_BACKEND, discover_service url = discover_service(CHRONICLE_BACKEND) if url: print(url, end='') else: - print('', end='', file=sys.stderr) - print('minidisc found no chronicle-backend service on Tailnet', file=sys.stderr) + print('minidisc found no chronicle-backend service on Tailnet', file=__import__('sys').stderr) except Exception as e: - print(f'discovery failed: {e}', file=sys.stderr) + print(f'discovery failed: {e}', file=__import__('sys').stderr) ") if [[ -n "$BACKEND_URL" ]]; then diff --git a/edge/services.yml b/edge/services.yml deleted file mode 100644 index 48ec689b..00000000 --- a/edge/services.yml +++ /dev/null @@ -1,33 +0,0 @@ -# Edge service registry -# Maps CLI service names to deployment metadata for edge node deployment. -services: - speaker-recognition: - minidisc_name: chronicle-speaker - port: 8085 - health_path: /health - compose_path: extras/speaker-recognition - description: Speaker Recognition Service - asr-services: - minidisc_name: chronicle-asr - port: 8767 - health_path: /health - compose_path: extras/asr-services - description: ASR Speech-to-Text Service - tts: - minidisc_name: chronicle-tts - port: 8770 - health_path: /health - compose_path: extras/tts - description: Text-to-Speech Service - llm-services: - minidisc_name: chronicle-llm - port: 8083 - health_path: /health - compose_path: extras/llm-services - description: Local LLM via llama.cpp - havpe-relay: - minidisc_name: chronicle-relay - port: 0 - health_path: / - compose_path: extras/havpe-relay - description: HAVPE Audio Relay diff --git a/edge/status.sh b/edge/status.sh index 506c6551..2047d6a8 100755 --- a/edge/status.sh +++ b/edge/status.sh @@ -19,27 +19,31 @@ else fi echo "" +declare -A SERVICE_PATHS=( + [speaker-recognition]=extras/speaker-recognition + [asr-services]=extras/asr-services + [tts]=extras/tts + [llm-services]=extras/llm-services + [havpe-relay]=extras/havpe-relay +) + if [[ $# -lt 1 ]]; then echo "Usage: $0 to see container status" - echo "Available services: speaker-recognition asr-services tts llm-services havpe-relay" + echo "Available services: ${!SERVICE_PATHS[*]}" exit 0 fi SERVICE_NAME="$1" -COMPOSE_PATH=$(cd "$CHRONICLE_HOME" && uv run --with-requirements setup-requirements.txt python3 -c " -import yaml, sys -with open('edge/services.yml') as f: - data = yaml.safe_load(f) -svc = data.get('services', {}).get('$SERVICE_NAME') -if not svc: - print('NOT_FOUND', file=sys.stderr); sys.exit(1) -print(svc['compose_path']) -" 2>/dev/null || echo "") +if [[ -z "${SERVICE_PATHS[$SERVICE_NAME]+_}" ]]; then + echo "Unknown service: $SERVICE_NAME" + echo "Available: ${!SERVICE_PATHS[*]}" + exit 1 +fi -SERVICE_DIR="$CHRONICLE_HOME/$COMPOSE_PATH" +SERVICE_DIR="$CHRONICLE_HOME/${SERVICE_PATHS[$SERVICE_NAME]}" -if [[ -z "$COMPOSE_PATH" || ! -d "$SERVICE_DIR" ]]; then +if [[ ! -d "$SERVICE_DIR" ]]; then echo "Service not found or not installed: $SERVICE_NAME" exit 1 fi diff --git a/edge/stop.sh b/edge/stop.sh index 2616c9e1..61aae821 100755 --- a/edge/stop.sh +++ b/edge/stop.sh @@ -14,20 +14,23 @@ fi SERVICE_NAME="$1" -# Parse compose_path from registry -COMPOSE_PATH=$(cd "$CHRONICLE_HOME" && uv run --with-requirements setup-requirements.txt python3 -c " -import yaml, sys -with open('edge/services.yml') as f: - data = yaml.safe_load(f) -svc = data.get('services', {}).get('$SERVICE_NAME') -if not svc: - print('NOT_FOUND', file=sys.stderr); sys.exit(1) -print(svc['compose_path']) -" 2>/dev/null || echo "") - -SERVICE_DIR="$CHRONICLE_HOME/$COMPOSE_PATH" - -if [[ -z "$COMPOSE_PATH" || ! -d "$SERVICE_DIR" ]]; then +declare -A SERVICE_PATHS=( + [speaker-recognition]=extras/speaker-recognition + [asr-services]=extras/asr-services + [tts]=extras/tts + [llm-services]=extras/llm-services + [havpe-relay]=extras/havpe-relay +) + +if [[ -z "${SERVICE_PATHS[$SERVICE_NAME]+_}" ]]; then + echo "Unknown service: $SERVICE_NAME" + echo "Available: ${!SERVICE_PATHS[*]}" + exit 1 +fi + +SERVICE_DIR="$CHRONICLE_HOME/${SERVICE_PATHS[$SERVICE_NAME]}" + +if [[ ! -d "$SERVICE_DIR" ]]; then echo "Service not found or not installed: $SERVICE_NAME" exit 1 fi diff --git a/extras/asr-services/docker-compose.yml b/extras/asr-services/docker-compose.yml index 98e16b52..3a7f940f 100644 --- a/extras/asr-services/docker-compose.yml +++ b/extras/asr-services/docker-compose.yml @@ -352,11 +352,9 @@ services: edge-agent: build: context: ../../edge + network_mode: host environment: - - EDGE_SERVICE_NAME=chronicle-asr - - EDGE_SERVICE_PORT=8767 - volumes: - - /var/run/tailscale/tailscaled.sock:/var/run/tailscale/tailscaled.sock:ro + - ADVERTISE=chronicle-asr:8767 restart: unless-stopped profiles: - edge diff --git a/extras/havpe-relay/docker-compose.yml b/extras/havpe-relay/docker-compose.yml index c59d5759..db7d9b0a 100644 --- a/extras/havpe-relay/docker-compose.yml +++ b/extras/havpe-relay/docker-compose.yml @@ -29,11 +29,9 @@ services: edge-agent: build: context: ../../edge + network_mode: host environment: - - EDGE_SERVICE_NAME=chronicle-relay - - EDGE_SERVICE_PORT=8989 - volumes: - - /var/run/tailscale/tailscaled.sock:/var/run/tailscale/tailscaled.sock:ro + - ADVERTISE=chronicle-relay:8989 restart: unless-stopped profiles: - edge diff --git a/extras/llm-services/docker-compose.yml b/extras/llm-services/docker-compose.yml index 50fb61be..064da179 100644 --- a/extras/llm-services/docker-compose.yml +++ b/extras/llm-services/docker-compose.yml @@ -48,11 +48,9 @@ services: edge-agent: build: context: ../../edge + network_mode: host environment: - - EDGE_SERVICE_NAME=chronicle-llm - - EDGE_SERVICE_PORT=8083 - volumes: - - /var/run/tailscale/tailscaled.sock:/var/run/tailscale/tailscaled.sock:ro + - ADVERTISE=chronicle-llm:8083 restart: unless-stopped profiles: - edge diff --git a/extras/speaker-recognition/docker-compose.yml b/extras/speaker-recognition/docker-compose.yml index fb92c765..8fa2ce86 100644 --- a/extras/speaker-recognition/docker-compose.yml +++ b/extras/speaker-recognition/docker-compose.yml @@ -141,11 +141,9 @@ services: edge-agent: build: context: ../../edge + network_mode: host environment: - - EDGE_SERVICE_NAME=chronicle-speaker - - EDGE_SERVICE_PORT=8085 - volumes: - - /var/run/tailscale/tailscaled.sock:/var/run/tailscale/tailscaled.sock:ro + - ADVERTISE=chronicle-speaker:8085 restart: unless-stopped profiles: - edge From 8c265292a28dbc13c0ee8436c3422f9cc6521f15 Mon Sep 17 00:00:00 2001 From: Ankush Malaker <43288948+AnkushMalaker@users.noreply.github.com> Date: Sun, 15 Mar 2026 23:03:38 +0000 Subject: [PATCH 07/25] fix --- backends/advanced/docker-compose.yml | 12 +++++ .../src/advanced_omi_backend/app_factory.py | 48 ------------------- .../controllers/system_controller.py | 41 +++++++--------- extras/asr-services/docker-compose.yml | 2 + extras/havpe-relay/docker-compose.yml | 2 + extras/llm-services/docker-compose.yml | 2 + extras/speaker-recognition/docker-compose.yml | 2 + 7 files changed, 38 insertions(+), 71 deletions(-) diff --git a/backends/advanced/docker-compose.yml b/backends/advanced/docker-compose.yml index 14bd1fd4..e644531d 100644 --- a/backends/advanced/docker-compose.yml +++ b/backends/advanced/docker-compose.yml @@ -282,6 +282,18 @@ services: retries: 5 start_period: 30s + # Discovery agent — advertises backend + co-located services on the Tailnet + discovery-agent: + build: + context: ../../edge + network_mode: host + env_file: .env + environment: + - ADVERTISE_BACKEND=true + volumes: + - /var/run/tailscale/tailscaled.sock:/var/run/tailscale/tailscaled.sock:ro + restart: unless-stopped + # ollama: # image: ollama/ollama:latest # container_name: ollama diff --git a/backends/advanced/src/advanced_omi_backend/app_factory.py b/backends/advanced/src/advanced_omi_backend/app_factory.py index 8758d566..c1d56fed 100644 --- a/backends/advanced/src/advanced_omi_backend/app_factory.py +++ b/backends/advanced/src/advanced_omi_backend/app_factory.py @@ -7,7 +7,6 @@ import asyncio import logging -import os import time from contextlib import asynccontextmanager from pathlib import Path @@ -216,51 +215,12 @@ async def _init_prompt_registry(): except Exception as e: application_logger.warning(f"Prompt registry initialization failed: {e}") - async def _init_service_discovery(): - try: - from discovery import ( - CHRONICLE_ASR, - CHRONICLE_BACKEND, - CHRONICLE_SPEAKER, - advertise_service, - is_tailscale_available, - ) - - if not is_tailscale_available(): - application_logger.debug( - "Tailscale socket not found — skipping service advertisement" - ) - return - - registry = advertise_service(8000, CHRONICLE_BACKEND) - if registry: - app.state.minidisc_registry = registry - - # Also advertise co-located optional services if configured - speaker_url = os.environ.get("SPEAKER_SERVICE_URL", "") - if speaker_url and "host.docker.internal" in speaker_url: - advertise_service(8085, CHRONICLE_SPEAKER) - - asr_url = os.environ.get("PARAKEET_ASR_URL", "") - if asr_url and "host.docker.internal" in asr_url: - advertise_service(8767, CHRONICLE_ASR) - - except ImportError: - application_logger.debug( - "discovery module not available — skipping service advertisement" - ) - except Exception as e: - application_logger.debug( - "Service discovery initialization failed (non-fatal): %s", e - ) - await asyncio.gather( _init_redis_rq(), _init_task_manager(), _init_client_manager(), _init_otel(), _init_prompt_registry(), - _init_service_discovery(), ) application_logger.info( @@ -527,14 +487,6 @@ async def _init_plugins(): except Exception as e: application_logger.error(f"Error stopping cron scheduler: {e}") - # Stop minidisc service advertisement - if hasattr(app.state, "minidisc_registry") and app.state.minidisc_registry: - try: - app.state.minidisc_registry = None - application_logger.info("Minidisc service advertisement stopped") - except Exception as e: - application_logger.debug("Error stopping minidisc registry: %s", e) - # Shutdown memory service and speaker service shutdown_memory_service() application_logger.info("Memory and speaker services shut down.") diff --git a/backends/advanced/src/advanced_omi_backend/controllers/system_controller.py b/backends/advanced/src/advanced_omi_backend/controllers/system_controller.py index 970256b6..1cbe42bf 100644 --- a/backends/advanced/src/advanced_omi_backend/controllers/system_controller.py +++ b/backends/advanced/src/advanced_omi_backend/controllers/system_controller.py @@ -40,8 +40,8 @@ async def get_network_discovery(app): """Return Tailscale status and discovered minidisc services. - The *app* parameter is the FastAPI application instance, used to read - ``app.state.minidisc_registry`` (set during startup in app_factory). + The *app* parameter is the FastAPI application instance (kept for API + compatibility but no longer used — discovery-agent handles advertising). """ import asyncio as _asyncio @@ -66,27 +66,22 @@ async def get_network_discovery(app): result["tailscale_available"] = is_tailscale_available() - # Report what this node is *actually* advertising via the minidisc - # registry that was initialised at startup (app_factory Phase 2). - registry = getattr(app.state, "minidisc_registry", None) - if registry is not None: - # The backend always advertises itself when the registry exists. - result["advertising"].append({"name": CHRONICLE_BACKEND, "port": 8000}) - - # Optional co-located services — app_factory only advertises them - # when the env-var URL points at host.docker.internal, so we - # mirror that same check here. - speaker_url = os.environ.get("SPEAKER_SERVICE_URL", "") - if speaker_url and "host.docker.internal" in speaker_url: - result["advertising"].append({"name": CHRONICLE_SPEAKER, "port": 8085}) - - asr_url = os.environ.get("PARAKEET_ASR_URL", "") - if asr_url and "host.docker.internal" in asr_url: - result["advertising"].append({"name": CHRONICLE_ASR, "port": 8767}) - - openmemory_url = os.environ.get("OPENMEMORY_MCP_URL", "") - if openmemory_url and "host.docker.internal" in openmemory_url: - result["advertising"].append({"name": CHRONICLE_OPENMEMORY, "port": 8765}) + # Report what the discovery-agent *should* be advertising based on the + # same env-var logic it uses. The actual advertising is handled by + # the discovery-agent container (network_mode: host). + result["advertising"].append({"name": CHRONICLE_BACKEND, "port": 8000}) + + speaker_url = os.environ.get("SPEAKER_SERVICE_URL", "") + if speaker_url and "host.docker.internal" in speaker_url: + result["advertising"].append({"name": CHRONICLE_SPEAKER, "port": 8085}) + + asr_url = os.environ.get("PARAKEET_ASR_URL", "") + if asr_url and "host.docker.internal" in asr_url: + result["advertising"].append({"name": CHRONICLE_ASR, "port": 8767}) + + openmemory_url = os.environ.get("OPENMEMORY_MCP_URL", "") + if openmemory_url and "host.docker.internal" in openmemory_url: + result["advertising"].append({"name": CHRONICLE_OPENMEMORY, "port": 8765}) if not result["tailscale_available"]: return result diff --git a/extras/asr-services/docker-compose.yml b/extras/asr-services/docker-compose.yml index 3a7f940f..acb1ef4f 100644 --- a/extras/asr-services/docker-compose.yml +++ b/extras/asr-services/docker-compose.yml @@ -355,6 +355,8 @@ services: network_mode: host environment: - ADVERTISE=chronicle-asr:8767 + volumes: + - /var/run/tailscale/tailscaled.sock:/var/run/tailscale/tailscaled.sock:ro restart: unless-stopped profiles: - edge diff --git a/extras/havpe-relay/docker-compose.yml b/extras/havpe-relay/docker-compose.yml index db7d9b0a..a2d2505c 100644 --- a/extras/havpe-relay/docker-compose.yml +++ b/extras/havpe-relay/docker-compose.yml @@ -32,6 +32,8 @@ services: network_mode: host environment: - ADVERTISE=chronicle-relay:8989 + volumes: + - /var/run/tailscale/tailscaled.sock:/var/run/tailscale/tailscaled.sock:ro restart: unless-stopped profiles: - edge diff --git a/extras/llm-services/docker-compose.yml b/extras/llm-services/docker-compose.yml index 064da179..0f560550 100644 --- a/extras/llm-services/docker-compose.yml +++ b/extras/llm-services/docker-compose.yml @@ -51,6 +51,8 @@ services: network_mode: host environment: - ADVERTISE=chronicle-llm:8083 + volumes: + - /var/run/tailscale/tailscaled.sock:/var/run/tailscale/tailscaled.sock:ro restart: unless-stopped profiles: - edge diff --git a/extras/speaker-recognition/docker-compose.yml b/extras/speaker-recognition/docker-compose.yml index 8fa2ce86..bcf3c8ab 100644 --- a/extras/speaker-recognition/docker-compose.yml +++ b/extras/speaker-recognition/docker-compose.yml @@ -144,6 +144,8 @@ services: network_mode: host environment: - ADVERTISE=chronicle-speaker:8085 + volumes: + - /var/run/tailscale/tailscaled.sock:/var/run/tailscale/tailscaled.sock:ro restart: unless-stopped profiles: - edge From ed66c78de4d3b02ce5b30675441909314deda6ff Mon Sep 17 00:00:00 2001 From: Ankush Malaker <43288948+AnkushMalaker@users.noreply.github.com> Date: Sun, 15 Mar 2026 23:04:13 +0000 Subject: [PATCH 08/25] missing file --- discovery.py | 178 +++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 178 insertions(+) create mode 100644 discovery.py diff --git a/discovery.py b/discovery.py new file mode 100644 index 00000000..afb96e59 --- /dev/null +++ b/discovery.py @@ -0,0 +1,178 @@ +""" +Zero-config service discovery for Chronicle on Tailscale networks. + +Thin wrapper around minidisc-python. Services advertise themselves on the +Tailnet; consumers discover them automatically. Every function is safe to call +even when Tailscale or minidisc is unavailable — they return None gracefully. + +Resolution priority (used by resolve_service_url): + 1. Explicit environment variable (manual override always wins) + 2. Minidisc discovery (automatic if Tailscale available) + 3. Default value / disabled +""" + +import logging +import os +import stat +from typing import Optional + +logger = logging.getLogger(__name__) + +# ── Service name constants ────────────────────────────────────────────── +CHRONICLE_BACKEND = "chronicle-backend" +CHRONICLE_SPEAKER = "chronicle-speaker" +CHRONICLE_ASR = "chronicle-asr" +CHRONICLE_OPENMEMORY = "chronicle-openmemory" +CHRONICLE_LLM = "chronicle-llm" +CHRONICLE_TTS = "chronicle-tts" +CHRONICLE_RELAY = "chronicle-relay" + +_TAILSCALE_SOCKET = "/var/run/tailscale/tailscaled.sock" + + +def is_tailscale_available() -> bool: + """Check whether the Tailscale daemon socket exists on this machine. + + Uses stat to verify the path is actually a Unix socket, not just a + directory (Docker creates empty dirs for missing bind-mount sources). + """ + try: + return stat.S_ISSOCK(os.stat(_TAILSCALE_SOCKET).st_mode) + except (OSError, ValueError): + return False + + +def advertise_service( + port: int, + name: str, + labels: Optional[dict] = None, +): + """Advertise a service on the local Tailnet. + + Returns the registry handle (keep a reference to stay advertised), + or None if Tailscale/minidisc is unavailable. + """ + try: + import threading + + import minidisc + + # start_registry() blocks on ready.wait() with no timeout. + # If the server thread can't bind (e.g., Docker container without + # Tailscale interfaces), it hangs forever. Run with a timeout. + registry_holder = [] + + def _try_start(): + registry_holder.append(minidisc.start_registry()) + + t = threading.Thread(target=_try_start, daemon=True) + t.start() + t.join(timeout=5) + + if not registry_holder: + logger.debug( + "minidisc registry startup timed out after 5s — skipping advertisement" + ) + return None + + registry = registry_holder[0] + registry.advertise_service(port, name, labels or {}) + logger.info("Advertising '%s' on port %d via minidisc", name, port) + return registry + except ImportError: + logger.debug("minidisc not installed — skipping service advertisement") + except Exception as e: + logger.debug("minidisc advertisement failed (non-fatal): %s", e) + return None + + +def discover_service( + name: str, + labels: Optional[dict] = None, + timeout: int = 3, +) -> Optional[str]: + """Discover a service on the Tailnet by name. + + Returns ``"http://{addr}:{port}"`` or None if not found. + """ + try: + import minidisc + + endpoint = minidisc.find_service(name, labels or {}) + if endpoint: + url = f"http://{endpoint}" + logger.info("Discovered '%s' at %s", name, url) + return url + except ImportError: + logger.debug("minidisc not installed — skipping service discovery") + except Exception as e: + logger.debug("minidisc discovery for '%s' failed (non-fatal): %s", name, e) + return None + + +def resolve_service_url( + env_var: Optional[str], + service_name: str, + labels: Optional[dict] = None, + default: Optional[str] = None, +) -> Optional[str]: + """Resolve a service URL with graceful fallback. + + Priority: + 1. Environment variable (if env_var is set and non-empty) + 2. Minidisc discovery on Tailnet + 3. default + """ + if env_var: + value = os.getenv(env_var) + if value: + return value + + discovered = discover_service(service_name, labels) + if discovered: + return discovered + + return default + + +def list_all_services() -> list[dict]: + """List all chronicle-* services on the Tailnet via minidisc. + + Returns a list of dicts with keys: name, address, port, labels. + Gracefully returns [] on ImportError or any exception. + """ + try: + import minidisc + + raw = minidisc.list_services() + results = [] + for svc in raw: + name = getattr(svc, "name", None) or ( + svc.get("name") if isinstance(svc, dict) else None + ) + if not name or not name.startswith("chronicle-"): + continue + if isinstance(svc, dict): + results.append( + { + "name": name, + "address": svc.get("address", ""), + "port": svc.get("port", 0), + "labels": svc.get("labels", {}), + } + ) + else: + results.append( + { + "name": name, + "address": getattr(svc, "address", ""), + "port": getattr(svc, "port", 0), + "labels": getattr(svc, "labels", {}), + } + ) + return results + except ImportError: + logger.debug("minidisc not installed — cannot list services") + except Exception as e: + logger.debug("minidisc list_services failed (non-fatal): %s", e) + return [] From 8667388abf6c1e5927b3fbc1bb8f4270d76eca70 Mon Sep 17 00:00:00 2001 From: Ankush Malaker <43288948+AnkushMalaker@users.noreply.github.com> Date: Mon, 16 Mar 2026 00:24:08 +0000 Subject: [PATCH 09/25] major fix --- .gitignore | 3 + backends/advanced/docker-compose.yml | 12 --- .../controllers/system_controller.py | 34 +++----- backends/advanced/webui/src/pages/Network.tsx | 6 +- edge/agent.py | 87 ++++++++++++------- 5 files changed, 71 insertions(+), 71 deletions(-) diff --git a/.gitignore b/.gitignore index efa33d15..9526e784 100644 --- a/.gitignore +++ b/.gitignore @@ -65,6 +65,9 @@ extras/test-audios/* extras/speaker-omni-experimental/data/* extras/speaker-omni-experimental/cache/* +# Discovery agent PID file +edge/.discovery-agent.pid + # AI Stuff .claude diff --git a/backends/advanced/docker-compose.yml b/backends/advanced/docker-compose.yml index e644531d..14bd1fd4 100644 --- a/backends/advanced/docker-compose.yml +++ b/backends/advanced/docker-compose.yml @@ -282,18 +282,6 @@ services: retries: 5 start_period: 30s - # Discovery agent — advertises backend + co-located services on the Tailnet - discovery-agent: - build: - context: ../../edge - network_mode: host - env_file: .env - environment: - - ADVERTISE_BACKEND=true - volumes: - - /var/run/tailscale/tailscaled.sock:/var/run/tailscale/tailscaled.sock:ro - restart: unless-stopped - # ollama: # image: ollama/ollama:latest # container_name: ollama diff --git a/backends/advanced/src/advanced_omi_backend/controllers/system_controller.py b/backends/advanced/src/advanced_omi_backend/controllers/system_controller.py index 1cbe42bf..5206ea4f 100644 --- a/backends/advanced/src/advanced_omi_backend/controllers/system_controller.py +++ b/backends/advanced/src/advanced_omi_backend/controllers/system_controller.py @@ -4,6 +4,7 @@ import asyncio import inspect +import json import logging import os import re @@ -52,36 +53,21 @@ async def get_network_discovery(app): } try: - from discovery import ( - CHRONICLE_ASR, - CHRONICLE_BACKEND, - CHRONICLE_OPENMEMORY, - CHRONICLE_SPEAKER, - is_tailscale_available, - list_all_services, - ) + from discovery import is_tailscale_available, list_all_services except ImportError: result["error"] = "discovery module not available" return result result["tailscale_available"] = is_tailscale_available() - # Report what the discovery-agent *should* be advertising based on the - # same env-var logic it uses. The actual advertising is handled by - # the discovery-agent container (network_mode: host). - result["advertising"].append({"name": CHRONICLE_BACKEND, "port": 8000}) - - speaker_url = os.environ.get("SPEAKER_SERVICE_URL", "") - if speaker_url and "host.docker.internal" in speaker_url: - result["advertising"].append({"name": CHRONICLE_SPEAKER, "port": 8085}) - - asr_url = os.environ.get("PARAKEET_ASR_URL", "") - if asr_url and "host.docker.internal" in asr_url: - result["advertising"].append({"name": CHRONICLE_ASR, "port": 8767}) - - openmemory_url = os.environ.get("OPENMEMORY_MCP_URL", "") - if openmemory_url and "host.docker.internal" in openmemory_url: - result["advertising"].append({"name": CHRONICLE_OPENMEMORY, "port": 8765}) + # Read advertised services written by services.py's discovery agent. + # The file is at config/advertised-services.json (volume-mounted from repo root). + _advertised_path = Path("/app/config/advertised-services.json") + if _advertised_path.exists(): + try: + result["advertising"] = json.loads(_advertised_path.read_text()) + except (json.JSONDecodeError, OSError) as exc: + logger.warning("Could not read advertised-services.json: %s", exc) if not result["tailscale_available"]: return result diff --git a/backends/advanced/webui/src/pages/Network.tsx b/backends/advanced/webui/src/pages/Network.tsx index 202f1029..f4096203 100644 --- a/backends/advanced/webui/src/pages/Network.tsx +++ b/backends/advanced/webui/src/pages/Network.tsx @@ -16,6 +16,7 @@ interface DiscoveredService { interface AdvertisedService { name: string port: number + label?: string } interface NetworkData { @@ -28,7 +29,7 @@ interface NetworkData { const SERVICE_DISPLAY: Record = { 'chronicle-backend': { label: 'Chronicle Backend', description: 'Core API and audio processing' }, 'chronicle-speaker': { label: 'Speaker Recognition', description: 'Voice identification service' }, - 'chronicle-asr': { label: 'Parakeet ASR', description: 'Offline speech-to-text' }, + 'chronicle-asr': { label: 'ASR Service', description: 'Offline speech-to-text' }, 'chronicle-openmemory': { label: 'OpenMemory MCP', description: 'Cross-client memory server' }, 'chronicle-llm': { label: 'Local LLM', description: 'Local LLM via llama.cpp' }, 'chronicle-tts': { label: 'Text-to-Speech', description: 'TTS synthesis service' }, @@ -169,10 +170,11 @@ export default function Network() {
{data.advertising.map((svc) => { const display = getServiceDisplay(svc.name) + const displayLabel = svc.label || display.label return (
-
{display.label}
+
{displayLabel}
Port {svc.port}
diff --git a/edge/agent.py b/edge/agent.py index 0ea0d7f0..c980a6a0 100644 --- a/edge/agent.py +++ b/edge/agent.py @@ -1,18 +1,15 @@ """ Edge discovery agent — advertises Chronicle services on the Tailnet via minidisc. -Config-driven: reads what to advertise from environment variables instead of -probing ports. One agent per machine. +Config-driven: reads what to advertise from the ADVERTISE environment variable +instead of probing ports. One agent per machine. -Two input modes (both contribute to a single service list): + ADVERTISE — comma-separated name:port pairs + e.g. ADVERTISE=chronicle-backend:8000,chronicle-speaker:8085 - ADVERTISE — comma-separated name:port pairs (edge nodes + overrides) - e.g. ADVERTISE=chronicle-speaker:8085,chronicle-asr:8767 - - ADVERTISE_BACKEND — when "true", always includes chronicle-backend:8000 and - inspects SPEAKER_SERVICE_URL / PARAKEET_ASR_URL / - OPENMEMORY_MCP_URL / TTS_SERVICE_URL to detect co-located - services (main server mode). +When launched by services.py, the ADVERTISE string is built automatically from +configured services. Per-service compose files (e.g. extras/asr-services) set +their own ADVERTISE for distributed deployments. """ import logging @@ -20,6 +17,7 @@ import signal import socket import sys +import threading import time logging.basicConfig( @@ -31,14 +29,6 @@ shutdown = False -# Mapping from env-var URL → (service name, port) -_BACKEND_SERVICE_MAP = { - "SPEAKER_SERVICE_URL": ("chronicle-speaker", 8085), - "PARAKEET_ASR_URL": ("chronicle-asr", 8767), - "OPENMEMORY_MCP_URL": ("chronicle-openmemory", 8765), - "TTS_SERVICE_URL": ("chronicle-tts", 8770), -} - def _handle_signal(signum, frame): global shutdown @@ -56,7 +46,6 @@ def _add(name: str, port: int): services.append((name, port)) seen.add(name) - # Mode 1: explicit ADVERTISE env var advertise = os.environ.get("ADVERTISE", "").strip() if advertise: for entry in advertise.split(","): @@ -74,15 +63,6 @@ def _add(name: str, port: int): except ValueError: logger.warning("Skipping invalid ADVERTISE entry (bad port): %s", entry) - # Mode 2: backend mode — read service URLs from .env - if os.environ.get("ADVERTISE_BACKEND", "").lower() in ("true", "1", "yes"): - _add("chronicle-backend", 8000) - - for env_var, (svc_name, svc_port) in _BACKEND_SERVICE_MAP.items(): - url = os.environ.get(env_var, "") - if url and "host.docker.internal" in url: - _add(svc_name, svc_port) - return services @@ -91,9 +71,7 @@ def main(): services = _collect_services() if not services: - logger.error( - "Nothing to advertise. Set ADVERTISE=name:port and/or ADVERTISE_BACKEND=true" - ) + logger.error("Nothing to advertise. Set ADVERTISE=name:port,...") sys.exit(1) hostname = socket.gethostname() @@ -107,8 +85,51 @@ def main(): logger.error("minidisc-python not installed") sys.exit(1) - logger.info("Starting minidisc registry...") - registry = minidisc.start_registry() + def _try_start_registry(): + """start_registry() hangs if the background thread can't bind. + Run it with a timeout so we can detect the failure.""" + holder = [] + + def _go(): + holder.append(minidisc.start_registry()) + + t = threading.Thread(target=_go, daemon=True) + t.start() + t.join(timeout=10) + return holder[0] if holder else None + + # Retry with backoff — on WSL2, the Tailscale interface may not be + # visible inside Docker even with network_mode:host. On real Linux + # (RPi, bare-metal) it works immediately. + registry = None + backoff = [0, 10, 30, 60] + for attempt, delay in enumerate(backoff): + if shutdown: + break + if delay: + logger.info( + "Retrying in %ds (attempt %d/%d)...", delay, attempt + 1, len(backoff) + ) + for _ in range(delay): + if shutdown: + break + time.sleep(1) + logger.info("Starting minidisc registry...") + registry = _try_start_registry() + if registry: + break + logger.warning( + "minidisc registry startup timed out (Tailscale interface not bindable)" + ) + + if not registry: + logger.error( + "Could not start minidisc registry after %d attempts. " + "Tailscale interface may not be reachable from this container " + "(common on WSL2/Docker Desktop). Exiting.", + len(backoff), + ) + sys.exit(1) labels = {"type": "discovery-agent", "host": hostname} for name, port in services: From 0b19d5cacb7b70c130303453f4d152c3fc9d3ec6 Mon Sep 17 00:00:00 2001 From: Ankush Malaker <43288948+AnkushMalaker@users.noreply.github.com> Date: Mon, 16 Mar 2026 00:28:59 +0000 Subject: [PATCH 10/25] input fix --- edge/install.sh | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/edge/install.sh b/edge/install.sh index 314948e8..37a974c6 100755 --- a/edge/install.sh +++ b/edge/install.sh @@ -158,7 +158,7 @@ if [[ -n "$BACKEND_URL" ]]; then ok "Auto-discovered backend at $BACKEND_URL" else warn "Could not auto-discover backend. Make sure your Chronicle backend is running with Tailscale." - read -rp "[edge] Enter your Chronicle backend URL (e.g. http://100.x.x.x:8000): " BACKEND_URL + read -rp "[edge] Enter your Chronicle backend URL (e.g. http://100.x.x.x:8000): " BACKEND_URL Date: Mon, 16 Mar 2026 00:32:38 +0000 Subject: [PATCH 11/25] fix --- edge/run-on-rpi.sh | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) diff --git a/edge/run-on-rpi.sh b/edge/run-on-rpi.sh index 5a9685a4..daf10255 100755 --- a/edge/run-on-rpi.sh +++ b/edge/run-on-rpi.sh @@ -1,5 +1,11 @@ #!/usr/bin/env bash # Quick deploy havpe-relay on RPi from feat/tailscale-discovery branch +# +# Downloads install.sh to a temp file first so stdin stays on the terminal +# (curl | bash breaks interactive prompts). export CHRONICLE_HOME=~/my-services/chronicle +TMPFILE=$(mktemp /tmp/chronicle-install.XXXXXX.sh) curl -sSL https://raw.githubusercontent.com/SimpleOpenSoftware/chronicle/feat/tailscale-discovery/edge/install.sh \ - | bash -s -- havpe-relay --branch feat/tailscale-discovery + -o "$TMPFILE" +bash "$TMPFILE" havpe-relay --branch feat/tailscale-discovery +rm -f "$TMPFILE" From 3149339427a3f4640f8dd2479f0c75ea092b1b60 Mon Sep 17 00:00:00 2001 From: Ankush Malaker <43288948+AnkushMalaker@users.noreply.github.com> Date: Mon, 16 Mar 2026 00:42:21 +0000 Subject: [PATCH 12/25] fix docker --- extras/havpe-relay/docker-compose.yml | 11 +---------- 1 file changed, 1 insertion(+), 10 deletions(-) diff --git a/extras/havpe-relay/docker-compose.yml b/extras/havpe-relay/docker-compose.yml index a2d2505c..af5a1fdb 100644 --- a/extras/havpe-relay/docker-compose.yml +++ b/extras/havpe-relay/docker-compose.yml @@ -4,24 +4,15 @@ services: build: context: . dockerfile: Dockerfile - ports: - - "${TCP_PORT:-8989}:8989" + network_mode: host env_file: .env environment: - TCP_PORT=${TCP_PORT:-8989} - DEBUG=${DEBUG:-0} restart: unless-stopped - healthcheck: - test: ["CMD", "netstat", "-an", "|", "grep", "8989"] - interval: 30s - timeout: 10s - retries: 3 - start_period: 10s command: ["uv", "run", "python3", "main.py", "-v", "--backend-url", "${BACKEND_URL:-http://host.docker.internal:8000}", "--backend-ws-url", "${BACKEND_WS_URL:-ws://host.docker.internal:8000}"] - extra_hosts: - - "host.docker.internal:host-gateway" volumes: - ./audio_chunks:/app/audio_chunks From 3377e76b24cf5d17dbbb6485397c3c514885cc76 Mon Sep 17 00:00:00 2001 From: Ankush Malaker <43288948+AnkushMalaker@users.noreply.github.com> Date: Mon, 16 Mar 2026 01:00:09 +0000 Subject: [PATCH 13/25] fix --- discovery.py | 50 ++++++++----- services.py | 194 +++++++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 226 insertions(+), 18 deletions(-) diff --git a/discovery.py b/discovery.py index afb96e59..f38db693 100644 --- a/discovery.py +++ b/discovery.py @@ -135,6 +135,17 @@ def resolve_service_url( return default +def _parse_endpoint(endpoint: str) -> tuple[str, int]: + """Parse a minidisc endpoint string like '100.99.62.5:8989' into (address, port).""" + if not endpoint: + return ("", 0) + try: + host, port_str = endpoint.rsplit(":", 1) + return (host, int(port_str)) + except (ValueError, AttributeError): + return (endpoint, 0) + + def list_all_services() -> list[dict]: """List all chronicle-* services on the Tailnet via minidisc. @@ -152,24 +163,27 @@ def list_all_services() -> list[dict]: ) if not name or not name.startswith("chronicle-"): continue - if isinstance(svc, dict): - results.append( - { - "name": name, - "address": svc.get("address", ""), - "port": svc.get("port", 0), - "labels": svc.get("labels", {}), - } - ) - else: - results.append( - { - "name": name, - "address": getattr(svc, "address", ""), - "port": getattr(svc, "port", 0), - "labels": getattr(svc, "labels", {}), - } - ) + labels = ( + getattr(svc, "labels", {}) + if not isinstance(svc, dict) + else svc.get("labels", {}) + ) + # minidisc Service objects use 'endpoint' (e.g. "100.x.x.x:8000"), + # not separate address/port fields. + endpoint = ( + getattr(svc, "endpoint", "") + if not isinstance(svc, dict) + else svc.get("endpoint", "") + ) + address, port = _parse_endpoint(str(endpoint)) + results.append( + { + "name": name, + "address": address, + "port": port, + "labels": labels, + } + ) return results except ImportError: logger.debug("minidisc not installed — cannot list services") diff --git a/services.py b/services.py index 0794554a..5459cdb5 100755 --- a/services.py +++ b/services.py @@ -5,14 +5,18 @@ """ import argparse +import json import os +import signal import subprocess +import time from pathlib import Path import yaml from dotenv import dotenv_values from rich.console import Console from rich.table import Table + from setup_utils import read_env_value console = Console() @@ -93,6 +97,88 @@ def load_config_yml(): } +_DISCOVERY_NAMES = { + "backend": "chronicle-backend", + "speaker-recognition": "chronicle-speaker", + "asr-services": "chronicle-asr", + "openmemory-mcp": "chronicle-openmemory", + "llm-services": "chronicle-llm", +} + + +_ADVERTISED_SERVICES_PATH = ( + Path(__file__).parent / "config" / "advertised-services.json" +) + +_ASR_PROVIDER_LABELS = { + "vibevoice": "VibeVoice ASR", + "vibevoice-strixhalo": "VibeVoice ASR", + "faster-whisper": "Faster Whisper ASR", + "transformers": "Transformers ASR", + "nemo": "NeMo ASR", + "nemo-strixhalo": "NeMo ASR", + "parakeet": "Parakeet ASR", + "qwen3-asr": "Qwen3 ASR", +} + + +def _get_advertised_services() -> list[tuple[str, int, str]]: + """Return list of (discovery_name, port, label) for configured services.""" + triples: list[tuple[str, int, str]] = [] + for svc_name, discovery_name in _DISCOVERY_NAMES.items(): + if svc_name not in SERVICES or not check_service_configured(svc_name): + continue + service = SERVICES[svc_name] + endpoints = service.get("health_endpoints", []) + if not endpoints: + continue + _label, port_env, default_port, _path = endpoints[0] + if port_env: + env_path = Path(service["path"]) / ".env" + if env_path.exists(): + port = int(dotenv_values(env_path).get(port_env, default_port)) + else: + port = int(default_port) + else: + port = int(default_port) + + # Derive display label + if svc_name == "asr-services": + env_path = Path(service["path"]) / ".env" + asr_provider = "" + if env_path.exists(): + asr_provider = ( + dotenv_values(env_path).get("ASR_PROVIDER", "").strip("'\"") + ) + label = _ASR_PROVIDER_LABELS.get(asr_provider, service["description"]) + else: + label = service["description"] + + triples.append((discovery_name, port, label)) + return triples + + +def _build_advertise_string() -> str: + """Build ADVERTISE=name:port,... string from configured services.""" + return ",".join( + f"{name}:{port}" for name, port, _label in _get_advertised_services() + ) + + +def _write_advertised_services(triples: list[tuple[str, int, str]]) -> None: + """Write advertised services to config/advertised-services.json for the backend.""" + data = [ + {"name": name, "port": port, "label": label} for name, port, label in triples + ] + _ADVERTISED_SERVICES_PATH.parent.mkdir(parents=True, exist_ok=True) + _ADVERTISED_SERVICES_PATH.write_text(json.dumps(data, indent=2) + "\n") + + +def _remove_advertised_services() -> None: + """Remove advertised-services.json when discovery agent stops.""" + _ADVERTISED_SERVICES_PATH.unlink(missing_ok=True) + + def _get_backend_env_path() -> Path: return Path(__file__).parent / "backends" / "advanced" / ".env" @@ -525,6 +611,94 @@ def ensure_docker_network(): return False +# --- Discovery agent (native process, not Docker) --- + +_DISCOVERY_PID = Path(__file__).parent / "edge" / ".discovery-agent.pid" +_DISCOVERY_LOG = Path(__file__).parent / "edge" / "discovery-agent.log" + + +def _discovery_agent_running() -> bool: + """Check if discovery agent process is alive.""" + if not _DISCOVERY_PID.exists(): + return False + try: + pid = int(_DISCOVERY_PID.read_text().strip()) + os.kill(pid, 0) + return True + except (ValueError, OSError): + _DISCOVERY_PID.unlink(missing_ok=True) + return False + + +def _start_discovery_agent(): + """Start discovery agent as a native background process. + + Runs outside Docker so it can bind to the Tailscale interface directly + (Docker Desktop VMs cannot see tailscale0). + """ + if _discovery_agent_running(): + console.print("[dim]📡 Discovery agent already running[/dim]") + return True + + agent_script = Path(__file__).parent / "edge" / "agent.py" + if not agent_script.exists(): + console.print("[yellow]⚠️ edge/agent.py not found, skipping discovery[/yellow]") + return False + + pairs = _get_advertised_services() + if not pairs: + console.print("[dim]📡 No services to advertise, skipping discovery[/dim]") + return False + + _write_advertised_services(pairs) + advertise = ",".join(f"{name}:{port}" for name, port, _label in pairs) + + env = dict(os.environ) + env["ADVERTISE"] = advertise + + log_file = open(_DISCOVERY_LOG, "a") + try: + proc = subprocess.Popen( + ["uv", "run", "--with", "minidisc-python", "python", str(agent_script)], + env=env, + stdout=log_file, + stderr=subprocess.STDOUT, + start_new_session=True, + ) + except Exception as e: + console.print(f"[red]❌ Failed to start discovery agent: {e}[/red]") + log_file.close() + return False + + log_file.close() + _DISCOVERY_PID.write_text(str(proc.pid)) + console.print(f"[green]✅ Discovery agent started (PID {proc.pid})[/green]") + return True + + +def _stop_discovery_agent(): + """Stop the discovery agent process.""" + if not _DISCOVERY_PID.exists(): + _remove_advertised_services() + return + + try: + pid = int(_DISCOVERY_PID.read_text().strip()) + os.killpg(pid, signal.SIGTERM) + for _ in range(10): + try: + os.kill(pid, 0) + time.sleep(0.5) + except OSError: + break + console.print(f"[green]✅ Discovery agent stopped (PID {pid})[/green]") + except (ValueError, OSError): + console.print("[dim]Discovery agent already stopped[/dim]") + finally: + _DISCOVERY_PID.unlink(missing_ok=True) + _remove_advertised_services() + + def start_services(services, build=False, force_recreate=False): """Start specified services""" console.print(f"🚀 [bold]Starting {len(services)} services...[/bold]") @@ -561,6 +735,10 @@ def start_services(services, build=False, force_recreate=False): f"\n[green]🎉 {success_count}/{len(services)} services started successfully[/green]" ) + # Start discovery agent alongside backend + if "backend" in services and check_service_configured("backend"): + _start_discovery_agent() + # Show access URLs if backend was started if "backend" in services and check_service_configured("backend"): backend_env = _get_backend_env_path() @@ -596,6 +774,10 @@ def stop_services(services): """Stop specified services""" console.print(f"🛑 [bold]Stopping {len(services)} services...[/bold]") + # Stop discovery agent when stopping backend + if "backend" in services: + _stop_discovery_agent() + success_count = 0 for service_name in services: if service_name not in SERVICES: @@ -664,6 +846,11 @@ def restart_services(services, recreate=False): f"\n[green]🎉 {success_count}/{len(services)} services restarted successfully[/green]" ) + # Restart discovery agent alongside backend + if "backend" in services and check_service_configured("backend"): + _stop_discovery_agent() + _start_discovery_agent() + def show_status(): """Show status of all services""" @@ -697,6 +884,13 @@ def show_status(): console.print(table) + # Discovery agent status + if _discovery_agent_running(): + pid = int(_DISCOVERY_PID.read_text().strip()) + console.print(f"\n[green]📡 Discovery agent running (PID {pid})[/green]") + else: + console.print("\n[dim]📡 Discovery agent not running[/dim]") + console.print("\n💡 [dim]Use './start.sh' to start all configured services[/dim]") From 3d8b43b940574f8f340443eebe94cda3da3303c2 Mon Sep 17 00:00:00 2001 From: Ankush Malaker <43288948+AnkushMalaker@users.noreply.github.com> Date: Mon, 16 Mar 2026 02:43:39 +0000 Subject: [PATCH 14/25] ui tweak --- .../controllers/system_controller.py | 19 +++++++ backends/advanced/webui/src/pages/Network.tsx | 54 ++++++++++++++++++- 2 files changed, 72 insertions(+), 1 deletion(-) diff --git a/backends/advanced/src/advanced_omi_backend/controllers/system_controller.py b/backends/advanced/src/advanced_omi_backend/controllers/system_controller.py index 5206ea4f..b0d63a13 100644 --- a/backends/advanced/src/advanced_omi_backend/controllers/system_controller.py +++ b/backends/advanced/src/advanced_omi_backend/controllers/system_controller.py @@ -110,6 +110,25 @@ async def _health_check(svc: dict): else: result["discovered_services"] = [] + # Connected WebSocket clients (phones, relays, etc.) + from advanced_omi_backend.client_manager import get_client_manager + + mgr = get_client_manager() + connected_devices = [] + for client_id, state in mgr.get_all_clients().items(): + device_name = client_id.split("-", 1)[1] if "-" in client_id else client_id + connected_devices.append( + { + "client_id": client_id, + "device_name": device_name, + "user_email": getattr(state, "user_email", None), + "connected": getattr(state, "connected", False), + "has_active_conversation": getattr(state, "current_audio_uuid", None) + is not None, + } + ) + result["connected_devices"] = connected_devices + return result diff --git a/backends/advanced/webui/src/pages/Network.tsx b/backends/advanced/webui/src/pages/Network.tsx index f4096203..a0537482 100644 --- a/backends/advanced/webui/src/pages/Network.tsx +++ b/backends/advanced/webui/src/pages/Network.tsx @@ -1,5 +1,5 @@ import { useState } from 'react' -import { Network as NetworkIcon, RefreshCw, CheckCircle, XCircle, Wifi, WifiOff, Radio, Search, Server } from 'lucide-react' +import { Network as NetworkIcon, RefreshCw, CheckCircle, XCircle, Wifi, WifiOff, Radio, Search, Server, Smartphone } from 'lucide-react' import { useQuery } from '@tanstack/react-query' import { useAuth } from '../contexts/AuthContext' import { systemApi } from '../services/api' @@ -19,10 +19,19 @@ interface AdvertisedService { label?: string } +interface ConnectedDevice { + client_id: string + device_name: string + user_email?: string + connected: boolean + has_active_conversation: boolean +} + interface NetworkData { tailscale_available: boolean advertising: AdvertisedService[] discovered_services: DiscoveredService[] + connected_devices?: ConnectedDevice[] error?: string } @@ -281,6 +290,49 @@ export default function Network() { )}
+ {/* Connected Devices */} +
+

+ + Connected Devices +

+ {data?.connected_devices && data.connected_devices.length > 0 ? ( +
+ {data.connected_devices.map((device) => ( +
+
+
{device.device_name}
+ {device.user_email && ( +
{device.user_email}
+ )} + {device.client_id} +
+
+ {device.connected ? ( + + connected + + ) : ( + + disconnected + + )} + {device.has_active_conversation && ( + + streaming + + )} +
+
+ ))} +
+ ) : ( +

+ No devices connected via WebSocket +

+ )} +
+ {/* Legend */}

Status Legend

From fca6fcf05456d6897c19d507b0c84c62b58e56fc Mon Sep 17 00:00:00 2001 From: Ankush Malaker <43288948+AnkushMalaker@users.noreply.github.com> Date: Mon, 16 Mar 2026 02:45:59 +0000 Subject: [PATCH 15/25] Add support for optional static ESPHome device IP configuration --- extras/havpe-relay/.env.template | 5 +++++ extras/havpe-relay/docker-compose.yml | 1 + extras/havpe-relay/init.py | 24 ++++++++++++++++++++++++ extras/havpe-relay/main.py | 7 +++++++ extras/havpe-relay/relay_core.py | 21 ++++++++++++++++++--- 5 files changed, 55 insertions(+), 3 deletions(-) diff --git a/extras/havpe-relay/.env.template b/extras/havpe-relay/.env.template index b0559acb..36cd444e 100644 --- a/extras/havpe-relay/.env.template +++ b/extras/havpe-relay/.env.template @@ -9,3 +9,8 @@ AUTH_PASSWORD= # Device Configuration DEVICE_NAME=havpe TCP_PORT=8989 + +# ESPHome device IP (optional). If set, the relay connects ESPHome API to this +# IP for LED/speaker/button control instead of using the TCP client's IP. +# Set this to your ESP32's LAN IP for reliable ESPHome connectivity. +# ESPHOME_DEVICE_IP=192.168.0.149 diff --git a/extras/havpe-relay/docker-compose.yml b/extras/havpe-relay/docker-compose.yml index af5a1fdb..58d75fcc 100644 --- a/extras/havpe-relay/docker-compose.yml +++ b/extras/havpe-relay/docker-compose.yml @@ -8,6 +8,7 @@ services: env_file: .env environment: - TCP_PORT=${TCP_PORT:-8989} + - ESPHOME_DEVICE_IP=${ESPHOME_DEVICE_IP:-} - DEBUG=${DEBUG:-0} restart: unless-stopped command: ["uv", "run", "python3", "main.py", "-v", diff --git a/extras/havpe-relay/init.py b/extras/havpe-relay/init.py index a08e8bc6..f16b1a93 100644 --- a/extras/havpe-relay/init.py +++ b/extras/havpe-relay/init.py @@ -229,6 +229,25 @@ def setup_device_config(self): self.config["TCP_PORT"] = tcp_port + def setup_esphome_config(self): + """Configure optional static ESPHome device IP""" + self.print_section("ESPHome Device IP (Optional)") + self.console.print( + "If set, the relay always connects ESPHome API to this IP instead of" + ) + self.console.print( + "the TCP client's IP. Recommended when non-ESP32 clients also connect." + ) + self.console.print() + + existing = self.read_existing_env_value("ESPHOME_DEVICE_IP") + esphome_ip = self.prompt_value( + "ESPHome device IP (leave blank to auto-detect from TCP client)", + existing or "", + ) + if esphome_ip: + self.config["ESPHOME_DEVICE_IP"] = esphome_ip + def setup_firmware_secrets(self): """Optionally configure ESP32 firmware secrets""" self.print_section("ESP32 Firmware Configuration (Optional)") @@ -352,6 +371,10 @@ def show_summary(self): ) self.console.print(f" Device Name: {self.config.get('DEVICE_NAME', '')}") self.console.print(f" TCP Port: {self.config.get('TCP_PORT', '')}") + esphome_ip = self.config.get("ESPHOME_DEVICE_IP", "") + self.console.print( + f" ESPHome IP: {esphome_ip or '(auto-detect from TCP client)'}" + ) def show_next_steps(self): """Show next steps""" @@ -377,6 +400,7 @@ def run(self): self.setup_backend_urls() self.setup_auth_credentials() self.setup_device_config() + self.setup_esphome_config() self.setup_firmware_secrets() # Generate files diff --git a/extras/havpe-relay/main.py b/extras/havpe-relay/main.py index e16a5c3f..2b7970a2 100644 --- a/extras/havpe-relay/main.py +++ b/extras/havpe-relay/main.py @@ -135,6 +135,12 @@ async def main(): parser.add_argument( "--device-name", type=str, default=os.getenv("DEVICE_NAME", "havpe") ) + parser.add_argument( + "--esphome-device-ip", + type=str, + default=os.getenv("ESPHOME_DEVICE_IP", ""), + help="Static ESPHome device IP (default: use TCP client IP)", + ) parser.add_argument("-v", "--verbose", action="count", default=0) parser.add_argument( "--dump-audio", @@ -150,6 +156,7 @@ async def main(): auth_username=args.username or "", auth_password=args.password or "", device_name=args.device_name, + esphome_device_ip=args.esphome_device_ip or "", ) level = logging.WARNING - (10 * min(args.verbose, 2)) diff --git a/extras/havpe-relay/relay_core.py b/extras/havpe-relay/relay_core.py index ae136603..ce147cb3 100644 --- a/extras/havpe-relay/relay_core.py +++ b/extras/havpe-relay/relay_core.py @@ -30,6 +30,7 @@ class RelayConfig: auth_username: str auth_password: str device_name: str + esphome_device_ip: str = "" @classmethod def from_env(cls) -> "RelayConfig": @@ -67,6 +68,7 @@ def from_env(cls) -> "RelayConfig": auth_username=os.getenv("AUTH_USERNAME", ""), auth_password=os.getenv("AUTH_PASSWORD", ""), device_name=os.getenv("DEVICE_NAME", "havpe"), + esphome_device_ip=os.getenv("ESPHOME_DEVICE_IP", ""), ) @@ -261,11 +263,24 @@ async def run_device_session( async with websockets.connect(backend_uri) as ws: logger.info("Backend WS connected, starting bidirectional bridge") - api_ok = await device.connect(device_ip) + esphome_ip = config.esphome_device_ip or device_ip + try: + api_ok = await asyncio.wait_for(device.connect(esphome_ip), timeout=5.0) + except asyncio.TimeoutError: + logger.warning( + "ESPHome API connect to %s timed out (5s) — audio-only mode", + esphome_ip, + ) + api_ok = False if api_ok: - logger.info("ESPHome API connected — button/dial/LED/speaker enabled") + logger.info( + "ESPHome API connected at %s — button/dial/LED/speaker enabled", + esphome_ip, + ) else: - logger.info("ESPHome API unavailable — audio-only mode") + logger.info( + "ESPHome API unavailable at %s — audio-only mode", esphome_ip + ) ws_lock = asyncio.Lock() tasks = [ From 19a8c5d341cafd91c7f2fdda631075e411daa528 Mon Sep 17 00:00:00 2001 From: Ankush Malaker <43288948+AnkushMalaker@users.noreply.github.com> Date: Sat, 28 Mar 2026 12:14:03 +0000 Subject: [PATCH 16/25] Refactor audio streaming and authentication handling - Updated `useAudioStreamer` to accept an options parameter for token refresh handling. - Implemented automatic re-login on token expiration, updating the WebSocket connection with a new token. - Enhanced error handling in WebSocket authentication to provide specific failure messages. - Modified audio retrieval endpoints to support both Opus and WAV formats, improving efficiency and flexibility. - Updated Docker configurations to replace Qdrant with Neo4j, adjusting environment variables and service dependencies accordingly. - Cleaned up unused code and improved logging for better traceability. --- app/app/index.tsx | 15 +- app/src/hooks/useAudioStreamer.ts | 102 +- backends/advanced/docker-compose-test.yml | 44 +- backends/advanced/docker-compose.yml | 69 +- backends/advanced/init.py | 9 +- .../src/advanced_omi_backend/app_config.py | 2 - .../src/advanced_omi_backend/app_factory.py | 7 +- .../advanced/src/advanced_omi_backend/auth.py | 59 +- .../controllers/websocket_controller.py | 26 +- .../middleware/app_middleware.py | 9 +- .../advanced_omi_backend/openai_factory.py | 25 +- .../advanced_omi_backend/prompt_defaults.py | 135 +- .../routers/modules/audio_routes.py | 149 +- .../routers/modules/conversation_routes.py | 55 +- .../routers/modules/health_routes.py | 10 +- .../routers/modules/knowledge_graph_routes.py | 110 +- .../services/knowledge_graph/__init__.py | 14 +- .../knowledge_graph/entity_extractor.py | 43 +- .../services/knowledge_graph/models.py | 69 +- .../services/knowledge_graph/queries.py | 116 +- .../services/knowledge_graph/service.py | 226 +-- .../services/memory/README.md | 912 ---------- .../services/memory/__init__.py | 5 +- .../services/memory/base.py | 153 +- .../services/memory/config.py | 73 +- .../services/memory/neo4j_utils.py | 263 +++ .../services/memory/prompts.py | 200 +-- .../services/memory/providers/__init__.py | 5 +- .../services/memory/providers/chronicle.py | 1531 ++++++----------- .../memory/providers/llm_providers.py | 1 - .../memory/providers/vector_stores.py | 584 ------- .../services/memory/vault_manager.py | 126 ++ .../transcription/streaming_consumer.py | 10 +- .../utils/audio_chunk_utils.py | 390 ++--- .../workers/conversation_jobs.py | 20 +- .../workers/finetuning_jobs.py | 16 +- .../workers/memory_jobs.py | 3 +- .../workers/waveform_jobs.py | 273 +-- .../knowledge-graph/PromisesList.tsx | 308 ---- .../src/components/knowledge-graph/index.ts | 2 - .../webui/src/pages/ConversationDetail.tsx | 15 +- .../webui/src/pages/Conversations.tsx | 13 +- .../advanced/webui/src/pages/Memories.tsx | 851 ++++----- backends/advanced/webui/src/services/api.ts | 15 +- config/README.md | 3 +- config/advertised-services.json | 17 + config/config.yml.template | 11 - config/defaults.yml | 15 - extras/ml-training/README.md | 12 + plugins/email_summarizer/plugin.py | 5 +- plugins/homeassistant/plugin.py | 34 +- plugins/hourly_recap/plugin.py | 14 +- tests/configs/README.md | 1 - tests/configs/deepgram-openai.yml | 11 - tests/configs/mock-services.yml | 11 - tests/configs/mock-transcription-failure.yml | 11 - tests/configs/mock-vibevoice.yml | 12 - tests/configs/parakeet-ollama.yml | 12 - tests/configs/parakeet-openai.yml | 12 - tests/endpoints/health_tests.robot | 4 +- tests/endpoints/memory_tests.robot | 13 +- tests/integration/integration_test.robot | 14 +- 62 files changed, 2324 insertions(+), 4951 deletions(-) delete mode 100644 backends/advanced/src/advanced_omi_backend/services/memory/README.md create mode 100644 backends/advanced/src/advanced_omi_backend/services/memory/neo4j_utils.py delete mode 100644 backends/advanced/src/advanced_omi_backend/services/memory/providers/vector_stores.py create mode 100644 backends/advanced/src/advanced_omi_backend/services/memory/vault_manager.py delete mode 100644 backends/advanced/webui/src/components/knowledge-graph/PromisesList.tsx create mode 100644 config/advertised-services.json diff --git a/app/app/index.tsx b/app/app/index.tsx index 1c4dc58e..a00ea9f9 100644 --- a/app/app/index.tsx +++ b/app/app/index.tsx @@ -37,8 +37,18 @@ export default function App() { // Bluetooth const { bleManager, bluetoothState, permissionGranted, requestBluetoothPermission, isPermissionsLoading } = useBluetoothManager(); + // Settings (must be before audioStreamer so the token refresh callback can reference it) + const settings = useAppSettings(); + // Audio - const audioStreamer = useAudioStreamer(); + const audioStreamer = useAudioStreamer({ + onTokenRefreshed: (newToken) => { + // Update app-level auth state when auto-re-login refreshes the token + if (settings.currentUserEmail) { + settings.handleAuthStatusChange(true, settings.currentUserEmail, newToken); + } + }, + }); const phoneAudioRecorder = usePhoneAudioRecorder(); const { isListeningAudio: isOmiAudioListenerActive, audioPacketsReceived, startAudioListener: originalStartAudioListener, stopAudioListener: originalStopAudioListener, isRetrying: isAudioListenerRetrying, retryAttempts: audioListenerRetryAttempts } = useAudioListener(omiConnection, () => !!deviceConnection.connectedDeviceId); @@ -49,9 +59,6 @@ export default function App() { useEffect(() => { isOmiAudioListenerActiveRef.current = isOmiAudioListenerActive; }, [isOmiAudioListenerActive]); useEffect(() => { isAudioStreamingRef.current = audioStreamer.isStreaming; }, [audioStreamer.isStreaming]); - // Settings - const settings = useAppSettings(); - // Device callbacks const onDeviceConnect = useCallback(async () => { const deviceIdToSave = omiConnection.connectedDeviceId; diff --git a/app/src/hooks/useAudioStreamer.ts b/app/src/hooks/useAudioStreamer.ts index f22371c0..ce09a538 100644 --- a/app/src/hooks/useAudioStreamer.ts +++ b/app/src/hooks/useAudioStreamer.ts @@ -3,6 +3,12 @@ import { useState, useRef, useCallback, useEffect } from 'react'; import { PermissionsAndroid, Platform } from 'react-native'; import notifee, { AndroidImportance } from '@notifee/react-native'; import NetInfo from '@react-native-community/netinfo'; +import { getAuthEmail, getAuthPassword, getWebSocketUrl, saveJwtToken } from '../utils/storage'; + +interface UseAudioStreamerOptions { + /** Called when a new JWT token is obtained via auto-re-login */ + onTokenRefreshed?: (token: string) => void; +} interface UseAudioStreamer { isStreaming: boolean; @@ -90,7 +96,7 @@ async function stopForegroundServiceNotification() { /** -------------------- Hook -------------------- */ -export const useAudioStreamer = (): UseAudioStreamer => { +export const useAudioStreamer = (options?: UseAudioStreamerOptions): UseAudioStreamer => { const [isStreaming, setIsStreaming] = useState(false); const [isConnecting, setIsConnecting] = useState(false); const [error, setError] = useState(null); @@ -108,6 +114,9 @@ export const useAudioStreamer = (): UseAudioStreamer => { const MAX_RECONNECT_MS = 30000; const HEARTBEAT_MS = 25000; + // Track if we received an auth error so onclose doesn't blindly reconnect + const authFailedRef = useRef(false); + // Guard state updates after unmount const mountedRef = useRef(true); useEffect(() => { @@ -133,6 +142,57 @@ export const useAudioStreamer = (): UseAudioStreamer => { } }, []); + // Helper: re-login with saved credentials when token expires, then rebuild the WebSocket URL and reconnect + const attemptReLogin = useCallback(async (): Promise => { + try { + const email = await getAuthEmail(); + const password = await getAuthPassword(); + const wsUrl = await getWebSocketUrl(); + if (!email || !password || !wsUrl) { + console.log('[AudioStreamer] Cannot re-login: missing saved credentials'); + return false; + } + + // Derive base HTTP URL from WebSocket URL + let baseUrl = wsUrl.replace('ws://', 'http://').replace('wss://', 'https://').split('/ws')[0]; + const loginUrl = `${baseUrl}/auth/jwt/login`; + + const formData = `username=${encodeURIComponent(email)}&password=${encodeURIComponent(password)}`; + const response = await fetch(loginUrl, { + method: 'POST', + headers: { 'Content-Type': 'application/x-www-form-urlencoded' }, + body: formData, + }); + + if (!response.ok) { + console.warn(`[AudioStreamer] Re-login failed: HTTP ${response.status}`); + return false; + } + + const data = await response.json(); + const newToken = data.access_token; + if (!newToken) { + console.warn('[AudioStreamer] Re-login: no access_token in response'); + return false; + } + + await saveJwtToken(newToken); + options?.onTokenRefreshed?.(newToken); + + // Rebuild the current URL with the new token + currentUrlRef.current = currentUrlRef.current.replace( + /([?&])token=[^&]*/, + `$1token=${encodeURIComponent(newToken)}` + ); + + console.log('[AudioStreamer] Re-login successful, token refreshed'); + return true; + } catch (e) { + console.error('[AudioStreamer] Re-login error:', e); + return false; + } + }, [options]); + // Helper: send Wyoming protocol events (UNCHANGED logic) const sendWyomingEvent = useCallback(async (event: WyomingEvent, payload?: Uint8Array) => { if (!websocketRef.current || websocketRef.current.readyState !== WebSocket.OPEN) { @@ -231,6 +291,7 @@ export const useAudioStreamer = (): UseAudioStreamer => { currentUrlRef.current = trimmed; manuallyStoppedRef.current = false; + authFailedRef.current = false; // Network gate const netState = await NetInfo.fetch(); @@ -284,8 +345,19 @@ export const useAudioStreamer = (): UseAudioStreamer => { }; ws.onmessage = (event) => { - // Handle server messages if needed - console.log('[AudioStreamer] Message:', event.data); + // Parse server messages to detect auth errors + try { + const msg = JSON.parse(event.data); + if (msg.type === 'error' && (msg.error === 'token_expired' || msg.error === 'authentication_failed' || msg.error === 'user_not_found')) { + console.warn(`[AudioStreamer] Auth error from server: ${msg.error} — ${msg.message}`); + authFailedRef.current = true; + setStateSafe(setError, msg.message || 'Session expired. Re-authenticating...'); + return; + } + } catch { + // Not JSON, that's fine (e.g. binary messages) + } + console.log('[AudioStreamer] Message:', typeof event.data === 'string' ? event.data.substring(0, 200) : '(binary)'); }; ws.onerror = (e) => { @@ -307,6 +379,28 @@ export const useAudioStreamer = (): UseAudioStreamer => { if (websocketRef.current === ws) websocketRef.current = null; + // Auth failure: try re-login instead of blind reconnect + if (authFailedRef.current && !manuallyStoppedRef.current) { + authFailedRef.current = false; + console.log('[AudioStreamer] Auth failure detected, attempting re-login...'); + setStateSafe(setError, 'Session expired. Re-authenticating...'); + attemptReLogin().then(success => { + if (success && currentUrlRef.current) { + console.log('[AudioStreamer] Re-login succeeded, reconnecting...'); + reconnectAttemptsRef.current = 0; + startStreaming(currentUrlRef.current).catch(err => { + console.error('[AudioStreamer] Reconnect after re-login failed:', err); + setStateSafe(setError, 'Re-authentication succeeded but reconnection failed.'); + }); + } else { + console.warn('[AudioStreamer] Re-login failed. Please log in manually.'); + setStateSafe(setError, 'Session expired. Please log in again in Settings.'); + notifyInfo('Session Expired', 'Please open the app and log in again.'); + } + }); + return; + } + if (!isManual && !manuallyStoppedRef.current) { setStateSafe(setError, 'Connection closed; attempting to reconnect.'); attemptReconnect(); @@ -321,7 +415,7 @@ export const useAudioStreamer = (): UseAudioStreamer => { reject(new Error(msg)); } }); - }, [attemptReconnect, sendWyomingEvent, setStateSafe, stopStreaming]); + }, [attemptReconnect, attemptReLogin, notifyInfo, sendWyomingEvent, setStateSafe, stopStreaming]); const sendAudio = useCallback(async (audioBytes: Uint8Array) => { if (websocketRef.current && websocketRef.current.readyState === WebSocket.OPEN && audioBytes.length > 0) { diff --git a/backends/advanced/docker-compose-test.yml b/backends/advanced/docker-compose-test.yml index 88604400..6cb88848 100644 --- a/backends/advanced/docker-compose-test.yml +++ b/backends/advanced/docker-compose-test.yml @@ -25,8 +25,9 @@ services: environment: # Override with test-specific settings - MONGODB_URI=mongodb://mongo-test:27017/test_db - - QDRANT_BASE_URL=qdrant-test - - QDRANT_PORT=6333 + - NEO4J_HOST=neo4j-test + - NEO4J_USER=neo4j + - NEO4J_PASSWORD=testpassword - REDIS_URL=redis://redis-test:6379/0 - DEBUG_DIR=/app/debug # Fixed: match plugin database mount path # Test configuration file @@ -66,8 +67,8 @@ services: - LANGFUSE_PUBLIC_KEY=pk-lf-test-public-key - LANGFUSE_SECRET_KEY=sk-lf-test-secret-key depends_on: - qdrant-test: - condition: service_started + neo4j-test: + condition: service_healthy mongo-test: condition: service_healthy redis-test: @@ -98,18 +99,30 @@ services: condition: service_healthy mongo-test: condition: service_healthy - qdrant-test: - condition: service_started redis-test: condition: service_started - qdrant-test: - image: qdrant/qdrant:latest + neo4j-test: + image: neo4j:2026-community ports: - - "6337:6333" # gRPC - avoid conflict with dev 6333 - - "6338:6334" # HTTP - avoid conflict with dev 6334 + - "7475:7474" # HTTP - avoid conflict with dev 7474 + - "7688:7687" # Bolt - avoid conflict with dev 7687 + environment: + - NEO4J_AUTH=neo4j/testpassword + - NEO4J_PLUGINS=["apoc"] + - NEO4J_dbms_security_procedures_unrestricted=apoc.* + - NEO4J_dbms_security_procedures_allowlist=apoc.* + - NEO4J_server_default__listen__address=0.0.0.0 + - NEO4J_server_bolt_listen__address=0.0.0.0:7687 + - NEO4J_server_http_listen__address=0.0.0.0:7474 volumes: - - ./data/test_qdrant_data:/qdrant/storage + - ./data/test_neo4j_data:/data + healthcheck: + test: ["CMD", "wget", "-q", "--spider", "http://localhost:7474"] + interval: 10s + timeout: 5s + retries: 10 + start_period: 30s mongo-test: image: mongo:8.0.14 @@ -373,8 +386,9 @@ services: environment: # Same environment as backend - MONGODB_URI=mongodb://mongo-test:27017/test_db - - QDRANT_BASE_URL=qdrant-test - - QDRANT_PORT=6333 + - NEO4J_HOST=neo4j-test + - NEO4J_USER=neo4j + - NEO4J_PASSWORD=testpassword - REDIS_URL=redis://redis-test:6379/0 - DEBUG_DIR=/app/debug # Fixed: match plugin database mount path # Test configuration file @@ -415,8 +429,8 @@ services: condition: service_healthy redis-test: condition: service_started - qdrant-test: - condition: service_started + neo4j-test: + condition: service_healthy restart: unless-stopped # caddy: diff --git a/backends/advanced/docker-compose.yml b/backends/advanced/docker-compose.yml index 04d2d980..aa7c09dd 100644 --- a/backends/advanced/docker-compose.yml +++ b/backends/advanced/docker-compose.yml @@ -1,30 +1,4 @@ services: - tailscale: - image: tailscale/tailscale:latest - container_name: advanced-tailscale - hostname: chronicle-tailscale - environment: - - TS_AUTHKEY=${TS_AUTHKEY} - - TS_STATE_DIR=/var/lib/tailscale - - TS_USERSPACE=false - - TS_ACCEPT_DNS=true - volumes: - - tailscale-state:/var/lib/tailscale - devices: - - /dev/net/tun:/dev/net/tun - cap_add: - - NET_ADMIN - restart: unless-stopped - profiles: - - tailscale # Optional profile - ports: - - "18123:18123" # HA proxy port - command: > - sh -c "tailscaled & - tailscale up --authkey=$${TS_AUTHKEY} --accept-dns=true && - apk add --no-cache socat 2>/dev/null || true && - socat TCP-LISTEN:18123,fork,reuseaddr TCP:100.99.62.5:8123" - chronicle-backend: image: ${CHRONICLE_REGISTRY:-}chronicle-backend:${CHRONICLE_TAG:-latest} build: @@ -62,8 +36,8 @@ services: - REDIS_URL=redis://redis:6379/0 - MONGODB_URI=mongodb://mongo:27017 depends_on: - qdrant: - condition: service_started + neo4j: + condition: service_healthy mongo: condition: service_healthy redis: @@ -127,8 +101,8 @@ services: condition: service_healthy mongo: condition: service_healthy - qdrant: - condition: service_started + neo4j: + condition: service_healthy restart: unless-stopped # Annotation Cron Scheduler @@ -210,7 +184,11 @@ services: context: ./webui dockerfile: Dockerfile.dev ports: - - "${WEBUI_PORT:-3010}:5173" + - "${WEBUI_DEV_PORT:-5173}:5173" + environment: + - VITE_BACKEND_URL=${VITE_BACKEND_URL-http://${HOST_IP}:${BACKEND_PUBLIC_PORT:-8000}} + - VITE_ALLOWED_HOSTS=${VITE_ALLOWED_HOSTS:-localhost 127.0.0.1} + - VITE_HMR_PORT=${VITE_HMR_PORT:-5173} volumes: - ./webui/src:/app/src - ./webui/public:/app/public @@ -227,6 +205,8 @@ services: volumes: - ./data/qdrant_data:/qdrant/storage restart: unless-stopped + profiles: + - qdrant # Optional — Chronicle now uses Neo4j for memory search mongo: image: mongo:8.0.14 @@ -257,7 +237,7 @@ services: retries: 5 neo4j: - image: neo4j:5.15-community + image: neo4j:2026-community hostname: neo4j ports: - "7474:7474" # HTTP @@ -283,6 +263,29 @@ services: retries: 5 start_period: 30s + tailscale: + image: tailscale/tailscale:latest + container_name: advanced-tailscale + hostname: chronicle-tailscale + environment: + - TS_AUTHKEY=${TS_AUTHKEY} + - TS_STATE_DIR=/var/lib/tailscale + - TS_USERSPACE=false + - TS_ACCEPT_DNS=true + volumes: + - tailscale-state:/var/lib/tailscale + devices: + - /dev/net/tun:/dev/net/tun + cap_add: + - NET_ADMIN + restart: unless-stopped + profiles: + - tailscale # Optional profile + command: > + sh -c "tailscaled & + tailscale up --authkey=$${TS_AUTHKEY} --accept-dns=true && + sleep infinity" + # ollama: # image: ollama/ollama:latest # container_name: ollama @@ -307,6 +310,8 @@ networks: external: true volumes: + ollama_data: + driver: local caddy_data: driver: local caddy_config: diff --git a/backends/advanced/init.py b/backends/advanced/init.py index 49cc706f..5c33dc29 100644 --- a/backends/advanced/init.py +++ b/backends/advanced/init.py @@ -1230,6 +1230,13 @@ def setup_https(self): self.config["HTTPS_ENABLED"] = "true" self.config["SERVER_IP"] = server_ip + # Configure webui-dev for same-origin API calls through Caddy + self.config["VITE_BACKEND_URL"] = "" + self.config["VITE_HMR_PORT"] = "443" + self.config["VITE_ALLOWED_HOSTS"] = ( + f"localhost 127.0.0.1 {server_ip}" + ) + except Exception as e: self.console.print( f"[red]❌ ERROR: Caddyfile generation failed: {e}[/red]" @@ -1283,7 +1290,7 @@ def generate_env_file(self): merged = {**preserved_values, **self.config} for key, value in merged.items(): - if value: # Only set non-empty values + if value is not None: # Only set values that were explicitly configured set_key(env_path_str, key, value) # Ensure secure permissions diff --git a/backends/advanced/src/advanced_omi_backend/app_config.py b/backends/advanced/src/advanced_omi_backend/app_config.py index d2f50d8e..3433dc2e 100644 --- a/backends/advanced/src/advanced_omi_backend/app_config.py +++ b/backends/advanced/src/advanced_omi_backend/app_config.py @@ -62,8 +62,6 @@ def __init__(self): ) # External Services Configuration - self.qdrant_base_url = os.getenv("QDRANT_BASE_URL", "qdrant") - self.qdrant_port = os.getenv("QDRANT_PORT", "6333") # Memory provider from registry _reg = get_models_registry() _mem = _reg.memory if _reg else {} diff --git a/backends/advanced/src/advanced_omi_backend/app_factory.py b/backends/advanced/src/advanced_omi_backend/app_factory.py index c1d56fed..1b3825fd 100644 --- a/backends/advanced/src/advanced_omi_backend/app_factory.py +++ b/backends/advanced/src/advanced_omi_backend/app_factory.py @@ -7,6 +7,7 @@ import asyncio import logging +import os import time from contextlib import asynccontextmanager from pathlib import Path @@ -500,7 +501,11 @@ def create_app() -> FastAPI: app = FastAPI(lifespan=lifespan) # Set up middleware (CORS, exception handlers) - setup_middleware(app) + setup_middleware( + app, + disable_request_logging=os.getenv("DISABLE_REQUEST_LOGGING", "").lower() + == "true", + ) # Include all routers app.include_router(api_router) diff --git a/backends/advanced/src/advanced_omi_backend/auth.py b/backends/advanced/src/advanced_omi_backend/auth.py index 3bd6ea88..3c858e33 100644 --- a/backends/advanced/src/advanced_omi_backend/auth.py +++ b/backends/advanced/src/advanced_omi_backend/auth.py @@ -253,10 +253,36 @@ async def create_admin_user_if_needed(): logger.error(f"Failed to create admin user: {e}", exc_info=True) -async def websocket_auth(websocket, token: Optional[str] = None) -> Optional[User]: +def _check_token_expired(token_str: str) -> bool: + """Check if a JWT token is expired without full verification. + + Decodes without signature verification to inspect the exp claim. + Returns True if the token is expired, False otherwise. + """ + try: + payload = jwt.decode( + token_str, options={"verify_signature": False, "verify_exp": False} + ) + exp = payload.get("exp") + if exp is not None: + from datetime import datetime, timezone + + return datetime.now(timezone.utc).timestamp() > exp + except Exception: + pass + return False + + +async def websocket_auth( + websocket, token: Optional[str] = None +) -> tuple[Optional[User], str]: """ WebSocket authentication that supports both cookie and token-based auth. - Returns None if authentication fails (allowing graceful handling). + + Returns: + tuple of (User or None, failure_reason string). + failure_reason is empty string on success, or one of: + "token_expired", "user_not_found", "token_invalid", "no_credentials" """ strategy = get_jwt_strategy() @@ -274,16 +300,25 @@ async def websocket_auth(websocket, token: Optional[str] = None) -> Optional[Use logger.info( f"WebSocket auth successful for user {user.user_id} using query token." ) - return user + return user, "" else: - logger.warning( - f"Token validated but user inactive or not found: user={user}" - ) + # read_token returned None — figure out why + if _check_token_expired(token): + logger.warning("WebSocket auth failed: JWT token is expired") + return None, "token_expired" + else: + logger.warning( + f"WebSocket auth failed: user inactive or not found (user={user})" + ) + return None, "user_not_found" except Exception as e: logger.error( f"WebSocket auth with query token failed: {type(e).__name__}: {e}", exc_info=True, ) + if _check_token_expired(token): + return None, "token_expired" + return None, "token_invalid" # Try cookie authentication logger.debug("Attempting WebSocket auth with cookie.") @@ -299,17 +334,21 @@ async def websocket_auth(websocket, token: Optional[str] = None) -> Optional[Use if cookie_header: match = re.search(r"fastapiusersauth=([^;]+)", cookie_header) if match: + cookie_token = match.group(1) user_db_gen = get_user_db() user_db = await user_db_gen.__anext__() user_manager = UserManager(user_db) - user = await strategy.read_token(match.group(1), user_manager) + user = await strategy.read_token(cookie_token, user_manager) if user and user.is_active: logger.info( f"WebSocket auth successful for user {user.user_id} using cookie." ) - return user + return user, "" + elif _check_token_expired(cookie_token): + logger.warning("WebSocket auth failed: cookie JWT token is expired") + return None, "token_expired" except Exception as e: logger.warning(f"WebSocket auth with cookie failed: {e}") - logger.warning("WebSocket authentication failed.") - return None + logger.warning("WebSocket authentication failed: no valid credentials provided.") + return None, "no_credentials" diff --git a/backends/advanced/src/advanced_omi_backend/controllers/websocket_controller.py b/backends/advanced/src/advanced_omi_backend/controllers/websocket_controller.py index 80ecb5f1..10e7389e 100644 --- a/backends/advanced/src/advanced_omi_backend/controllers/websocket_controller.py +++ b/backends/advanced/src/advanced_omi_backend/controllers/websocket_controller.py @@ -402,28 +402,44 @@ async def _setup_websocket_connection( await ws.accept() # Authenticate user after accepting connection - user = await websocket_auth(ws, token) + user, auth_failure_reason = await websocket_auth(ws, token) if not user: + # Build specific error message based on failure reason + if auth_failure_reason == "token_expired": + error_type = "token_expired" + message = "Your session has expired. Please log in again." + close_reason = "Token expired" + elif auth_failure_reason == "user_not_found": + error_type = "user_not_found" + message = "User account not found or inactive. Please log in again." + close_reason = "User not found" + else: + error_type = "authentication_failed" + message = "Authentication failed. Please log in again and ensure your token is valid." + close_reason = "Authentication failed" + # Send error message to client before closing try: error_msg = ( json.dumps( { "type": "error", - "error": "authentication_failed", - "message": "Authentication failed. Please log in again and ensure your token is valid.", + "error": error_type, + "message": message, "code": 1008, } ) + "\n" ) await ws.send_text(error_msg) - application_logger.info("Sent authentication error message to client") + application_logger.info( + f"Sent auth error to client: {error_type} ({close_reason})" + ) except Exception as send_error: application_logger.warning(f"Failed to send error message: {send_error}") # Close connection with appropriate code - await ws.close(code=1008, reason="Authentication failed") + await ws.close(code=1008, reason=close_reason) return None, None, None # Generate proper client_id using user and device_name diff --git a/backends/advanced/src/advanced_omi_backend/middleware/app_middleware.py b/backends/advanced/src/advanced_omi_backend/middleware/app_middleware.py index 693ce0da..96100e88 100644 --- a/backends/advanced/src/advanced_omi_backend/middleware/app_middleware.py +++ b/backends/advanced/src/advanced_omi_backend/middleware/app_middleware.py @@ -230,11 +230,14 @@ async def http_exception_handler(request: Request, exc: HTTPException): return JSONResponse(status_code=exc.status_code, content={"detail": exc.detail}) -def setup_middleware(app: FastAPI) -> None: +def setup_middleware(app: FastAPI, *, disable_request_logging: bool = False) -> None: """Set up all middleware for the FastAPI application.""" # Add request logging middleware - app.add_middleware(RequestLoggingMiddleware) - logger.info("📝 Request logging middleware enabled") + if not disable_request_logging: + app.add_middleware(RequestLoggingMiddleware) + logger.info("📝 Request logging middleware enabled") + else: + logger.info("📝 Request logging middleware DISABLED (profiling)") setup_cors_middleware(app) setup_exception_handlers(app) diff --git a/backends/advanced/src/advanced_omi_backend/openai_factory.py b/backends/advanced/src/advanced_omi_backend/openai_factory.py index b10f72f0..10fc2aa1 100644 --- a/backends/advanced/src/advanced_omi_backend/openai_factory.py +++ b/backends/advanced/src/advanced_omi_backend/openai_factory.py @@ -4,6 +4,9 @@ modules that need an OpenAI client should use this factory instead of creating clients directly. +Clients are cached by (api_key, base_url, is_async) to avoid repeated +SSL context creation (~400ms per instantiation). + Tracing is handled by the OTEL instrumentor (see observability/otel_setup.py), which auto-instruments all OpenAI calls at startup. No per-client wrapping needed. """ @@ -14,9 +17,14 @@ logger = logging.getLogger(__name__) +_client_cache: dict[tuple[str, str, bool], openai.OpenAI | openai.AsyncOpenAI] = {} + def create_openai_client(api_key: str, base_url: str, is_async: bool = False): - """Create an OpenAI client. + """Get or create a cached OpenAI client. + + Clients are cached by (api_key, base_url, is_async). If the API key or + base URL changes (e.g. config reload), a new client is created automatically. Args: api_key: OpenAI API key @@ -26,7 +34,18 @@ def create_openai_client(api_key: str, base_url: str, is_async: bool = False): Returns: OpenAI or AsyncOpenAI client instance """ + cache_key = (api_key, base_url, is_async) + client = _client_cache.get(cache_key) + if client is not None: + return client + if is_async: - return openai.AsyncOpenAI(api_key=api_key, base_url=base_url) + client = openai.AsyncOpenAI(api_key=api_key, base_url=base_url) else: - return openai.OpenAI(api_key=api_key, base_url=base_url) + client = openai.OpenAI(api_key=api_key, base_url=base_url) + + _client_cache[cache_key] = client + logger.info( + f"Created {'async' if is_async else 'sync'} OpenAI client for {base_url}" + ) + return client diff --git a/backends/advanced/src/advanced_omi_backend/prompt_defaults.py b/backends/advanced/src/advanced_omi_backend/prompt_defaults.py index 37dfcd24..6a5a03c3 100644 --- a/backends/advanced/src/advanced_omi_backend/prompt_defaults.py +++ b/backends/advanced/src/advanced_omi_backend/prompt_defaults.py @@ -314,72 +314,6 @@ def register_all_defaults(registry: PromptRegistry) -> None: category="memory", ) - # ------------------------------------------------------------------ - # memory.temporal_extraction - # ------------------------------------------------------------------ - registry.register_default( - "memory.temporal_extraction", - template="""\ -You are an expert at extracting temporal and entity information from memory facts. - -Your task is to analyze a memory fact and extract structured information in JSON format: -1. **Entity Types**: Determine if the memory is about events, people, places, promises, or relationships -2. **Temporal Information**: Extract and resolve any time references to actual ISO 8601 timestamps -3. **Named Entities**: List all people, places, and things mentioned -4. **Representation**: Choose a single emoji that captures the essence of the memory - -You must return a valid JSON object with the following structure. - -**Current Date Context:** -- Today's date: {{current_date}} -- Current time: {{current_time}} -- Day of week: {{day_of_week}} - -**Time Resolution Guidelines:** - -Relative Time References: -- "tomorrow" -> Add 1 day to current date -- "next week" -> Add 7 days to current date -- "in X days/weeks/months" -> Add X time units to current date -- "yesterday" -> Subtract 1 day from current date - -Time of Day: -- "4pm" or "16:00" -> Use current date with that time -- "tomorrow at 4pm" -> Use tomorrow's date at 16:00 -- "morning" -> 09:00 on the referenced day -- "afternoon" -> 14:00 on the referenced day -- "evening" -> 18:00 on the referenced day -- "night" -> 21:00 on the referenced day - -Duration Estimation (when only start time is mentioned): -- Events like "wedding", "meeting", "party" -> Default 2 hours duration -- "lunch", "dinner", "breakfast" -> Default 1 hour duration -- "class", "workshop" -> Default 1.5 hours duration -- "appointment", "call" -> Default 30 minutes duration - -**Entity Type Guidelines:** - -- **isEvent**: True for scheduled activities, appointments, meetings, parties, ceremonies, classes, etc. -- **isPerson**: True when the primary focus is on a person (e.g., "Met John", "Sarah is my friend") -- **isPlace**: True when the primary focus is a location (e.g., "Botanical Gardens is beautiful", "Favorite restaurant is...") -- **isPromise**: True for commitments, promises, or agreements (e.g., "I'll call you tomorrow", "We agreed to meet") -- **isRelationship**: True for statements about relationships (e.g., "John is my brother", "We're getting married") - -**Instructions:** -- Return structured data following the TemporalEntity schema -- Convert all temporal references to ISO 8601 format -- Be conservative: if there's no temporal information, leave timeRanges empty -- Multiple tags can be true (e.g., isEvent and isPerson both true for "meeting with John") -- Extract all meaningful entities (people, places, things) mentioned in the fact -- Choose an emoji that best represents the core meaning of the memory -""", - name="Temporal Extraction", - description="Extracts temporal and entity information from memory facts with date resolution.", - category="memory", - variables=["current_date", "current_time", "day_of_week"], - is_dynamic=True, - ) - # ------------------------------------------------------------------ # chat.system # ------------------------------------------------------------------ @@ -458,7 +392,7 @@ def register_all_defaults(registry: PromptRegistry) -> None: registry.register_default( "knowledge_graph.entity_extraction", template="""\ -You are an entity extraction system. Extract entities, relationships, and promises from conversation transcripts. +You are an entity extraction system. Extract entities and relationships from conversation transcripts. ENTITY TYPES: - person: Named individuals (not generic roles) @@ -480,10 +414,9 @@ def register_all_defaults(registry: PromptRegistry) -> None: 1. Only extract NAMED entities (not "my friend" but "John") 2. Use "speaker" as the subject when the user mentions themselves 3. Extract temporal info for events (dates, times) -4. Capture promises/commitments with deadlines -5. Skip filler words, small talk, and vague references -6. Normalize names (capitalize properly) -7. Assign appropriate emoji icons to entities +4. Skip filler words, small talk, and vague references +5. Normalize names (capitalize properly) +6. Assign appropriate emoji icons to entities Return a JSON object with this structure: { @@ -502,20 +435,13 @@ def register_all_defaults(registry: PromptRegistry) -> None: "relation": "works_at|lives_in|knows|attended|located_at|part_of|related_to", "object": "Target entity name" } - ], - "promises": [ - { - "action": "What was promised", - "to": "Person promised to (optional)", - "deadline": "When it should be done (optional)" - } ] } -If no entities, relationships, or promises are found, return empty arrays. +If no entities or relationships are found, return empty arrays. Only return valid JSON, no additional text.""", name="Entity Extraction", - description="Extracts entities, relationships, and promises from conversation transcripts.", + description="Extracts entities and relationships from conversation transcripts.", category="knowledge_graph", ) @@ -715,6 +641,55 @@ def register_all_defaults(registry: PromptRegistry) -> None: is_dynamic=True, ) + # ------------------------------------------------------------------ + # memory.generate_conversation_doc + # ------------------------------------------------------------------ + registry.register_default( + "memory.generate_conversation_doc", + template="""\ +You are generating a structured conversation document from a transcript. + +Given a transcript with speaker labels, produce a markdown document with this EXACT structure: + +--- +conversation_id: {{conversation_id}} +date: {{date}} +speakers: [{{speakers}}] +duration_minutes: {{duration}} +--- + +## {Title - descriptive, 3-8 words} + +### Summary +{2-3 sentence summary of what was discussed} + +### Key Facts +{Bulleted list of specific facts, decisions, numbers, dates mentioned} + +### People +{Bulleted list in format: - Name (role/relationship, context)} +Include ALL named individuals — speakers, people mentioned, people referenced. +If a speaker is identified by name (e.g., "John" not "Speaker 0"), they MUST appear here. +Do not include unnamed roles or generic labels like "Speaker 0". + +### Action Items +{Bulleted list in format: - [ ] Action item description} +Use [x] for items already completed in the conversation. + +Rules: +- Every ### section MUST be present, even if empty (use "- None" for empty sections) +- People section: ONLY named individuals, format MUST be "- Name (description)" +- Key Facts: be specific — include dates, numbers, names +- Action Items: use checkbox format [ ] or [x] +- Do NOT add any sections beyond the four listed above +""", + name="Conversation Document Generator", + description="Generates a structured markdown conversation document from a transcript with frontmatter, summary, key facts, people, and action items.", + category="memory", + variables=["conversation_id", "date", "speakers", "duration"], + is_dynamic=True, + ) + # ------------------------------------------------------------------ # memory.consolidate_basic_memory # ------------------------------------------------------------------ diff --git a/backends/advanced/src/advanced_omi_backend/routers/modules/audio_routes.py b/backends/advanced/src/advanced_omi_backend/routers/modules/audio_routes.py index 81065a8e..e61b6a05 100644 --- a/backends/advanced/src/advanced_omi_backend/routers/modules/audio_routes.py +++ b/backends/advanced/src/advanced_omi_backend/routers/modules/audio_routes.py @@ -24,6 +24,8 @@ from advanced_omi_backend.utils.audio_chunk_utils import ( build_wav_from_pcm, concatenate_chunks_to_pcm, + get_opus_for_conversation, + get_opus_for_time_range, reconstruct_wav_from_conversation, retrieve_audio_chunks, ) @@ -70,6 +72,7 @@ async def upload_audio_from_drive_folder( async def get_conversation_audio( conversation_id: str, request: Request, + format: str = Query(default="opus", description="Audio format: opus or wav"), token: Optional[str] = Query( default=None, description="JWT token for audio element access" ), @@ -78,27 +81,17 @@ async def get_conversation_audio( """ Serve complete audio file for a conversation from MongoDB chunks. - Reconstructs audio by: - 1. Retrieving all Opus-compressed chunks from MongoDB - 2. Decoding each chunk to PCM - 3. Concatenating PCM data - 4. Building complete WAV file with headers - - Supports both header-based auth (Authorization: Bearer) and query param token - for