Remove interactive ISO storage detection/selection and associated ISO_PATH logic. Default to a fixed cache directory (/var/lib/vz/template/iso) via CACHE_DIR and update CACHE_FILE and mkdir usage accordingly. Also update VM configuration to reference the local ISO store (local:iso/<file>) instead of the previously selectable storage. This simplifies ISO handling by avoiding pvesm/whiptail prompts; note this hardcodes the ISO location and removes support for alternate ISO storages.
Introduce an interactive script (vm/cachyos-vm.sh) to create and configure a CachyOS VM on Proxmox. The script validates Proxmox environment and architecture, offers default and advanced settings via whiptail, detects storage/ISO locations, downloads the CachyOS desktop ISO, allocates EFI and main disks, creates and configures the VM (network, CPU, memory, disks, boot order, OVMF), sets a rich VM description, and optionally starts the VM. Also add an ASCII header file (vm/headers/cachyos-vm) used by the script.
Introduce a telemetry data microservice under misc/data: add Dockerfile, entrypoint, migration tools, README, LICENSE and a .gitignore. Increase Docker CACHE_TTL_SECONDS to 300s. Implement extensive dashboard and analytics updates in dashboard.go: add total_all_time and sample_size, return total item counts from fetchRecords (with page/limit handling and a maxRecords guard), raise top-N limits, add a minimum-installs threshold for failed-apps, and numerous UI/style/layout improvements in the embedded DashboardHTML. Minor formatting tweak to misc/api.func.
Update dashboard UI in misc/data/dashboard.go: move the repository <select> into the quickfilter group, apply inline styling for a compact themed dropdown, and shorten option labels (remove environment suffixes). Add a separator span between the select and quickfilter buttons. Move the Export CSV button from the top controls into the footer to simplify the header layout and avoid duplicate export buttons.
Remove implicit defaults for PocketBase collection names and make them required environment variables. Replace env(..., "_telemetry_service"/"_telemetry_data") with mustEnv("PB_AUTH_COLLECTION") and mustEnv("PB_TARGET_COLLECTION") so the service fails startup if those are not provided. Also update inline comments to clarify these fields come from the environment.
Refactor telemetry backend to store all telemetry in a single collection and add repo_source-based filtering.
Key changes:
- Added detect_repo_source() in misc/api.func to auto-detect/export REPO_SOURCE (ProxmoxVE/ProxmoxVED/external) when scripts are sourced.
- Consolidated PocketBase collections into a single default collection (_telemetry_data) across service, migration, and scripts; updated defaults in migrate.go, migration.go, migrate.sh and migration shell scripts.
- Simplified PBClient to use one targetColl and removed collection resolution logic; updated create/update/find/fetch functions to use targetColl.
- Introduced repo_source field (values: "ProxmoxVE", "ProxmoxVED", "external") on telemetry records and telemetry payloads; updated validation and logging.
- Added repo filtering to dashboard endpoints, FetchDashboardData and FetchRecordsPaginated, plus a repo selector in the dashboard UI; default filter is ProxmoxVE (production), with an "all" option.
- Adjusted API handlers and callers to pass repo filters and include repo_source when upserting telemetry.
- Misc: updated comments, error messages, and logging to reflect the new model; added telemetry-service.exe binary.
Purpose: simplify data model (single collection), make telemetry attributable to repository sources, and enable dashboard filtering by repo/source.
Introduce weekly summary reports and a cleanup job, enhance dashboard UI, and adjust telemetry/build settings.
- Add REPO_SOURCE to misc/api.func and include repo_source in telemetry payloads.
- Implement weekly report generation/scheduling in alerts.go: new data types, HTML/plain templates, scheduler, SendWeeklyReport/TestWeeklyReport, and email/HTML helpers.
- Add Cleaner (misc/data/cleanup.go) to detect and mark stuck installations as 'unknown' with scheduling and manual trigger APIs.
- Enhance dashboard backend/frontend (misc/data/dashboard.go): optional days filter (allow 'All'), increase fetch page size, simplify fetchRecords, add quick filter buttons, detail & health modals, improved styles and chart options, and client-side record detail view.
- Update Dockerfile (misc/data/Dockerfile): rename binaries to telemetry-service and build migrate from ./migration/migrate.go; copy adjusted in final image.
- Add migration tooling (misc/data/migration/migrate.sh and migration.go) and other small service changes.
These changes add operational reporting and cleanup capabilities, improve observability and UX of the dashboard, and align build and telemetry identifiers for the service.
Use an absolute path for the downloaded edgetpu model to ensure it lands at /edgetpu_model.tflite. Wrap the OpenVINO build step in a conditional: only copy the generated .xml/.bin and fetch the COCO class file if build_ov_model.py exits successfully; otherwise emit a warning and allow Frigate to fall back to the CPU model (handles machines without required CPU instructions).
