cc2-dash

cc2-dash is a lightweight local dashboard and portal shell for the Elegoo Centauri Carbon 2 / CC2 ecosystem. It gives you a clean LAN dashboard, printer discovery and pairing, camera relay/fanout, a stock Elegoo portal bridge, optional Ollama-powered visual monitoring, feedback-aware AI review tools, G-code upload staging, kiosk mode, stock-style control tools, and a themeable mobile-friendly UI. Experimental File Manager and Filament Manager code is retained in the source tree but locked off for this public test build; the Control page remains enabled with runtime safety gates.

It is designed for a Raspberry Pi-style board sitting on your trusted home network. It is intended as a printer-room companion dashboard, not an enterprise print-farm controller.

Warning

Private, home, hobbyist use only. cc2-dash is not designed, tested, or recommended for production environments, commercial print farms, safety-critical workflows, unattended remote operation, or any situation where missed detection, a failed command, or an incorrect AI result could cause damage. Keep physical access to your printer and use the stock printer controls as the final authority.

Important

This is an unofficial project. It is not affiliated with, endorsed by, or supported by Elegoo, OctoEverywhere, or any printer vendor. Firmware behavior can change. Some stock command paths behave differently across firmware versions.

Note

In this version, Failure Detection can optionally pause a print after a high-risk warning countdown. Auto-pause is off by default, has a configurable countdown/cancel window, uses a shared pause-permission gate, and performs a fresh telemetry/vision recheck immediately before sending PAUSE_PRINT. Cancel print remains locked behind manual controls; AI never cancels, resumes, loads/unloads filament, or overrides the stock printer controls.

Current status
Tested hardware and platform notes
Feature overview
What cc2-dash does not do
Install from GitHub on Raspberry Pi OS
Run manually
Install as a systemd service
Update from GitHub
First-run setup
Using the dashboard
Multi-View
Printer Manager
Camera Relay / stream protection
Kiosk mode
Portal AI and Ollama vision
AI feedback and false-alarm suppression
Persistent AI learning
AI Training review page
Upload page
File Manager
Filament Manager / CANVAS controls
Control page
Stock Elegoo portal bridge
Themes and appearance
Logs and diagnostics
Safety gates and command behavior
Configuration and data paths
Useful API endpoints
Troubleshooting
Known limitations
Uninstall
Project layout
Release notes
Development checks

Current status

Current documented version:

1.2.76 klipper-device-portal-controls
1.2.73 multi-view-dashboard
1.2.72 dummy-printer-simulator
1.2.71 multi-printer-view-context
1.2.70 chamber-readout-only
1.2.69 chamber-temperature-status
1.2.68 ai-training-backup-restore
1.2.67 file-manager-info-modal
1.2.66 file-manager-search-sort
1.2.64 file-manager-multiselect
1.2.63 timelapse-export-confirmation
1.2.62 async-timelapse-export
1.2.61 roi-feedback-mobile-cache-fix
1.2.60 roi-missed-failure-feedback

Major current capabilities:

Area	Status
Printer discovery / pairing	Working, verified Centauri discovery filtering, with optional dummy/simulator printer entries for multi-printer UI testing
Dashboard status	Working, mobile-first, active/idle aware, with active-printer context, global warning banner, Multi-View overview, and basic Klipper/Moonraker status support
Stock portal bridge	Working as fallback/reference portal
Camera relay	Working, reduces direct camera connection pileups
Kiosk mode	Working, camera-first fullscreen view
Failure Detection telemetry checks	Working, with optional guarded auto-pause off by default
Ollama vision checks	Working, active-print-only by default
AI feedback dataset	Working, includes fresh-frame capture, missed-failure ROI annotation/crops, optional reason chips, JSONL audit log, SQLite mirror/import, outcome interpretation, AI Training review tools, lightweight dataset export, and full backup/restore ZIP import/export
False-alarm suppression	Working for similar low/severity warnings on the same active print
Persistent AI learning	Working foundation plus Settings UI visibility and optional safe auto-adjustment of live vision thresholds
Upload page	Working, stages local `.gcode` files in cc2-dash, extracts metadata/thumbnails where possible, then uploads or uploads-and-prints
File Manager	Experimental code retained, disabled and locked off for this public test build; dev branch includes async timelapse export, confirmed generated/download-ready status before download, friendly timelapse composing status labels, themed export controls, mobile-friendly multi-select deletion, search/filter/sort controls, themed mobile-friendly info modals with thumbnail previews, and stock-shaped local printer-file delete payloads
Filament Manager / CANVAS	Experimental code retained, disabled and locked off for this public test build
Control page	Enabled, stock-portal-style controls with offline/active-print lockouts, command permissions, fans, speed, light, jog/home, bed/extruder temperature controls, and read-only chamber temperature readout
Themes	Built-in theme library with preview cards
Windows support	Not tested; may work manually, but scripts are Linux/systemd focused

Tested hardware and platform notes

Tested

Raspberry Pi Zero 2 W running Raspberry Pi OS-style Linux.

Expected to work better

Raspberry Pi 4.
Raspberry Pi 5.
Other Debian/Ubuntu-like Linux boxes.
Small x86 Linux mini-PCs.

A Pi Zero 2 W can run the dashboard, but a Pi 4 or Pi 5 is a much nicer target if you plan to use camera relay, logs, browser clients, and Ollama-related network calls heavily. Ollama itself should usually run on a stronger LAN machine, not on the Zero 2 W.

Windows

Windows has not been tested. The backend is Python/FastAPI, so it might run with manual setup, but the included helper scripts, service installation, and process-management assumptions are aimed at Raspberry Pi OS / Linux / systemd.

Feature overview

Local dashboard

FastAPI backend.
Plain CSS and vanilla JavaScript frontend.
Mobile-first responsive layout.
Themeable UI.
Collapsible dashboard sections.
Saved dashboard accordion state per printer.
Compact build/version chips in the header.
Configurable top navigation visibility for Portal, Multi-View, Upload, Control, Kiosk, AI Training, and Logs. File Manager and Filament Manager are locked off in this public test build.
/health and /api/version diagnostics.

Printer discovery and pairing

UDP Centauri discovery using Elegoo method 7000.
Verified-printer scan filtering so unrelated LAN devices are not shown as printer candidates.
Manual printer add when discovery is blocked.
Alphanumeric printer PIN/access-code fields.
No prefilled default PIN.
Printer serial/SN, access code, MQTT host/port, type, and command permissions stored per printer.
Dummy/simulator printers can be added from Settings for testing multi-printer switching, camera/status rendering, and AI badge behavior without owning a second printer.

Dashboard controls

Optional dashboard actions include:

Light toggle.
Pause print.
Resume print.
Cancel print.
Camera wake/enable.
Speed preset selection.
Manual camera analysis.

Command buttons are controlled by per-printer safety settings. Dangerous commands remain gated to reduce accidental activation from touch devices.

Camera relay

Keeps one upstream MJPEG camera connection to the printer.
Serves dashboard clients from local relay/fanout endpoints.
Provides cached latest frame for Portal AI and feedback capture.
Helps prevent multiple browser tabs and AI checks from dogpiling the printer camera endpoint.

Portal AI

Telemetry/rule-based print health checks.
Optional Ollama vision analysis.
Local image heuristics for dark frames, contrast, fine-edge/stringing-like changes, and stale/frozen-looking frames.
Active-print-only monitoring so idle printers do not waste cycles or create meaningless warnings.
Optional guarded auto-pause can pause on high-risk failures after a configurable warning countdown. Auto-pause is off by default and never cancels a print.

Feedback-aware AI review

Looks Good / Looks Bad / False Alarm buttons.
Fresh camera frame capture on feedback click, with cached-frame fallback.
Feedback interpreted into true positive / false positive / false negative / true negative.
Same-print suppression for repeated low/severity false alarms.
Feedback and frame data saved locally for later review/tuning.

Stock portal bridge

Bundled stock Elegoo-style portal page.
Local MQTT-over-WebSocket bridge.
Fullscreen portal route.
Portal camera rewrite shim that tries to route embedded camera views through cc2-dash's camera relay.

Experimental file and filament tools

The File Manager and Filament Manager are still present in the codebase, but they are disabled and locked off for this public test build. This keeps the community release focused on the stable dashboard, portal, camera relay, Upload page, Control page, Failure Detection, kiosk, settings, logs, and AI training workflows while the most firmware-sensitive file/history/CANVAS tools get more validation.

The release gate is configured in:

cc2_dash/config.py

Current gate switch:

COMMUNITY_RELEASE_EXPERIMENTAL_LOCKS = True

Current locked feature keys:

EXPERIMENTAL_FEATURE_LOCKS = {
    "file_manager_enabled": ...,
    "filament_manager_enabled": ...,
}

When the master lock is enabled, cc2-dash forces those feature flags to false during config load/save, disables their Settings toggles, and blocks direct page/API access with a disabled-feature page or 403 response. Control is intentionally not listed in EXPERIMENTAL_FEATURE_LOCKS, so it remains available while still being protected by command permissions, offline checks, and active-print lockouts.

When the file/filament lock is removed in a future build, these retained tools include:

File Manager support for stock-style printer files, USB files, print history, and video records where firmware supports it.
Timelapse export/download helpers, subject to firmware behavior.
CANVAS/MMS filament slot display and command helpers.
Idle-only filament load/unload/edit controls.

Themes

Built-in themes include:

Octo Dark Blue.
Amber Terminal.
Mainsail-ish Dark.
Carbon Glass.
Toxic Green Lab.
Blood Red Terminal.
Elegoo Dark.
Klipper Blue.
OLED Mono.
Cyberpunk Magenta.
High Contrast.

Theme preview cards are available in first-run setup and Settings.

What cc2-dash does not do

Important boundaries:

It does not make the printer safe to leave unattended.
It does not replace the stock Elegoo portal.
It does not guarantee failure detection.
It does not cancel, resume, load/unload filament, jog axes, set heaters, or override stock controls from AI decisions. Optional auto-pause is off by default and only sends pause after an explicit opt-in countdown.
It does not harden your LAN or provide production-grade authentication.
It does not fix firmware features that are broken in the stock portal itself.

Use it as a local dashboard, helper, and experiment platform.

Install from GitHub on Raspberry Pi OS

These instructions assume Raspberry Pi OS, Debian, Ubuntu, or a similar apt-based Linux system.

1. Update the Pi and install basic tools

sudo apt update
sudo apt install -y git python3 python3-venv python3-pip

2. Clone the repository

Clone the current project repository:

git clone https://github.com/merberg-ai/cc2-dash.git
cd cc2-dash

3. Make helper scripts executable

chmod +x install.sh run.sh uninstall.sh

4. Install Python dependencies

./install.sh

The installer will:

Check for Python 3.
Create .venv/ if needed.
Upgrade pip, setuptools, and wheel.
Install packages from requirements.txt.
Create the local data/ folder.

Dependencies currently include:

fastapi
httpx
jinja2
paho-mqtt
pydantic
python-multipart
requests
uvicorn[standard]
Pillow

Run manually

Start the app:

./run.sh

Open from another device on the same LAN:

http://<pi-ip>:8088/

Example:

http://192.168.1.50:8088/

To run on a different port:

CC2_PORT=8090 ./run.sh

Install as a systemd service

For always-on use, install cc2-dash as a background service:

./install.sh --service --port=8088

Useful commands:

sudo systemctl status cc2-dash --no-pager
sudo systemctl restart cc2-dash
sudo systemctl stop cc2-dash
sudo journalctl -u cc2-dash -f

The service runs:

python -m uvicorn cc2_dash.main:app --host 0.0.0.0 --port 8088

Update from GitHub

From the project folder:

git pull
./install.sh

If installed as a service:

sudo systemctl restart cc2-dash

If running manually, stop the old process and restart:

./run.sh

Your local runtime data lives in data/. Do not delete it unless you want to reset printers, logs, AI feedback, learned AI profiles, and settings.

If you previously installed the older cc2-dash-lite.service, the installer now attempts to stop, disable, and remove that legacy service so it does not compete with cc2-dash.service for the same port.

First-run setup

When no valid printer is configured, cc2-dash opens the setup wizard.

Setup flow:

Step	Purpose
1	Find printers using verified Centauri discovery
2	Add a printer manually if discovery fails
3	Pick theme and font preferences
4	Configure LAN access allowlist
5	Configure Portal AI and optional Ollama vision
6	Review and launch dashboard

The scanner only shows devices that answer the expected Centauri discovery probe. Routers, phones, smart plugs, and random LAN web servers should stay hidden.

Printer settings saved during setup:

Printer display name
Printer host/IP
Printer serial/SN
Printer PIN/access code
MQTT port, usually 1883
Default printer selection
Command permission flags

Using the dashboard

Primary pages:

Page	Description
Dash	Main printer view with status, camera, quick actions, Portal AI, and connection info.
Multi-View	Multi-printer overview with one snapshot/status/progress/temperature/AI card per configured printer.
Portal	Stock Elegoo portal bridge/fallback.
Kiosk	Camera-first fullscreen display for tablets or spare monitors.
Upload	Stage `.gcode` files in cc2-dash, review parsed metadata/thumbnails, then upload or upload-and-print.
Control	Stock-portal-style jog/home, light, fans, speed, bed/extruder temperature controls, and read-only chamber temperature readout with safety lockouts.
Files	Experimental file/history/timelapse helper page. Locked off in this public test build.
Filament	Experimental CANVAS/MMS filament manager. Locked off in this public test build.
Settings	Printer Manager, themes, menu visibility, quick actions, access, camera relay, AI settings.
Logs	Filterable runtime log viewer.

The Portal, Multi-View, Upload, Control, Kiosk, AI Training, and Logs navigation items can be shown or hidden in:

Settings → Menu / Features

The Files and Filament rows remain visible in Settings as locked release-gated features while COMMUNITY_RELEASE_EXPERIMENTAL_LOCKS is enabled.

Active printer switching

The header now includes a compact Printer selector whenever more than one visible printer is configured. If only one printer is available, the selector is hidden to keep the header clean. This changes the printer being viewed on the current page without changing the saved default printer. Page links preserve the active printer with a URL like:

/?printer=<printer_id>
/files?printer=<printer_id>
/control?printer=<printer_id>

The selected/viewed printer is remembered in the browser for convenience. The saved default printer remains the fallback used when no printer is selected, when a URL has no printer= query string, or when a bookmarked printer no longer exists.

Multi-View

Open:

Multi-View

The Multi-View page is a read-only overview for multiple configured printers. Each card shows:

current camera snapshot
printer state
active file
progress bar
hotend / bed / chamber temperatures
cached Failure Detection / AI badge
quick links to Dashboard, Control, and Files for that printer

Cards are large tap targets on mobile. Tapping the card opens that printer's full dashboard using ?printer=<printer_id>. Multi-View uses a fast cached status path for AI badges, so it does not intentionally run expensive Ollama vision checks for every card just because the overview page is open. Background monitoring still owns actual AI checks.

Klipper / Moonraker manual add

cc2-dash now has a first-pass Klipper/Moonraker printer type. Klipper systems are manual-add only; the existing Carbon 2 scan/discovery flow is unchanged and still only shows verified Centauri Carbon printers.

Go to:

Settings → Printer Manager → Klipper / Moonraker printer

Enter the Moonraker host/IP and port. The port does not need to be the default 7125; use whatever your Moonraker instance actually listens on. Optional fields let you provide a Moonraker API key/token and manual camera stream/snapshot URLs. If camera URLs are omitted, cc2-dash attempts to read Moonraker's configured webcam list.

The current basic Klipper phase supports:

Dashboard, Kiosk, and Multi-View status cards.
Camera snapshot/stream through the cc2-dash camera endpoints.
Hotend, bed, and chamber/temperature-sensor display when Moonraker reports those objects.
Progress/current file from Klipper print_stats / virtual_sdcard.
Pause, resume, and cancel through Moonraker only when explicitly enabled per printer.
A type-aware Open Device Portal link that opens the configured Moonraker/device URL directly instead of the embedded Elegoo portal.
A Klipper-safe Control page panel that exposes print-job controls only and disables Carbon 2-only motion/fan/heater/light controls.
AI monitoring using the same normalized status/camera path as CC2 printers.

Out of scope for this Klipper basic phase: Klipper file manager, upload/print, macros, fans, movement, bed mesh, console, power devices, and emergency stop. Those should be added as Klipper-native features later instead of being mixed into the Centauri Carbon command path.

Printer Manager

Open:

Settings → Printer Manager

Available actions:

Scan for verified Centauri printers.
Add a printer manually.
Edit printer name, IP/host, serial, access code, and MQTT port.
Enable or disable printer entries.
Choose the default printer.
Enable/disable normal commands.
Enable/disable dangerous commands.
Remove old printer entries.
Add a Dummy / simulator printer for safe UI testing. Dummy printers provide fake status, temperatures, progress, a generated camera stream, sample files/history/timelapse rows, and selectable AI badge states. They never connect to MQTT and never send real printer commands.

Command permissions are intentionally separate from pairing. You can monitor a printer while keeping control buttons locked down.

The default printer is now treated as a startup/fallback printer only. Opening a page with ?printer=<printer_id> or choosing a printer from the header switcher changes the currently viewed printer without rewriting the default printer setting. This is the first phase of the multi-printer UI foundation.

Dummy / simulator printers

Settings → Printer Manager includes an Add Dummy Printer tool. This creates a safe fake printer entry with configurable scenarios:

Printing
Idle
Paused
Time-lapse generating
Error
Offline

Dummy printers are useful when testing multi-printer UI behavior with only one real CC2. For public builds or non-testing installs, they can be disabled from cc2_dash/config.py by setting DUMMY_PRINTERS_ENABLED = False; saved dummy config entries are then hidden/blocked without deleting them. They show up in the header printer switcher, dashboard, control/kiosk camera views, File Manager reads, and AI status badges. The generated camera stream is local to cc2-dash, and command/file operations return no-op dummy responses for UI testing only. No MQTT client is started and no real printer command is sent.

Camera Relay / stream protection

The CC2 camera can get cranky when too many things connect to it directly. The Camera Relay reduces that load.

How it works:

cc2-dash opens one upstream MJPEG connection to the printer camera.
The latest frame is cached in memory.
Browser clients receive a local MJPEG stream from the dashboard server.
Portal AI and feedback capture use the cached/latest frame instead of opening extra direct camera connections.
The stock portal shim tries to rewrite embedded camera URLs through the relay.

Useful endpoints:

GET  /api/printers/<printer_id>/camera/stream
GET  /api/printers/<printer_id>/camera/snapshot.jpg
GET  /api/printers/<printer_id>/camera/latest.jpg
GET  /api/printers/<printer_id>/camera/status
GET  /api/camera/status
POST /api/printers/<printer_id>/camera/restart

Recommended default: relay enabled, start-on-boot enabled, portal rewrites enabled, direct fallback disabled unless debugging.

Kiosk mode

Kiosk mode opens a minimal camera-first page intended for:

Wall tablet.
Spare phone.
Browser tab on a shop monitor.
Quick glance print display.

It can show:

Camera relay state.
Active printer.
Active file.
Print state: IDLE or PRINTING.
Progress bar and percent.
Estimated time remaining.
Portal AI badge.

Settings live under:

Settings → Kiosk Mode

The Kiosk nav item can be shown/hidden under:

Settings → Menu / Features

Portal AI and Ollama vision

Portal AI combines printer telemetry, local rules, optional camera heuristics, and optional Ollama vision output into an advisory status.

Current behavior:

Runs only during active print jobs.
Stands by when the printer is idle.
Tracks stale status, printer error states, pause/error/fail states, stuck progress, temp sanity, filament status hints, and camera/vision issues.
Uses Ollama vision only when enabled and when an active print is detected.
Does not cancel jobs automatically. Optional high-risk auto-pause is available only when explicitly enabled.

Multi-printer watchdog behavior:

The background watchdog keeps separate cached AI/vision state per printer_id.
Auto-pause countdowns, cancellations, cooldowns, and sent-pause timestamps are isolated per printer.
The scheduler staggers due printer checks so several active printers do not all hit camera/Ollama at the same instant.
The currently viewed printer can be softly prioritized, but the saved default printer is not changed.
Global AI warning banners can surface high-risk warnings from another printer while you are viewing a different printer/page.

New multi-printer settings under Settings → Portal AI:

Control	Purpose
Multi-printer AI scheduler	Enables staggered due-time scheduling across configured printers.
Max AI checks per tick	Number of due printers to process in one scheduler tick. Keep at `1` for Pi/Ollama stability.
Printer check stagger	Seconds between due printer checks when more than one printer needs attention.
Prioritize viewed printer	Softly checks the printer currently open in the browser before other equally due printers.
Global AI warning banner	Shows high-risk/cross-printer warnings without triggering new camera/model checks.

Common checks:

Printer connected/reachable.
MQTT status freshness.
Error/fail/emergency/stopped states.
Paused state warning.
Stuck progress timer.
Hotend/bed target sanity during active print.
Filament sensor hints.
Printer exception-code decoding, such as 1252 — Extruder unload failure / unload timeout, when the firmware reports known exception IDs.
Camera availability hints.
Dark/low-contrast frame checks.
Fine-edge/stringing-like frame checks.
Vision model classification.

Ollama settings live under:

Settings → Portal AI

Typical Ollama URL:

http://192.168.1.24:11434

Related controls:

Control	Purpose
Load Models	Fetch installed Ollama models from `/api/tags`.
Test	Check that the selected model is reachable.
Pull	Request model download through Ollama.
Analyze Camera Now	Manually trigger a one-shot vision check during an active print.
Treat benign uncertainty as OK	Downgrade uncertain/no-evidence responses instead of warning loudly.

AI feedback and false-alarm suppression

Feedback buttons:

Looks Good
Looks Bad
False Alarm
Report Missed Failure — captures a frozen camera frame and lets you draw a mobile-friendly box around the failed area.

Feedback records are saved to:

data/ai_feedback.jsonl
data/ai_feedback_frames/<printer_id>/
data/ai_feedback_suppressions.json
data/ai_learning.sqlite3

When feedback is clicked, cc2-dash tries to capture a fresh frame. If that fails, it falls back to the latest cached frame. After the fast click is saved, an optional reason-chip panel can tag why the feedback was given, such as normal supports, purge tower, spaghetti/stringing, detached print, low light but visible, or a custom note.

For missed localized failures, Report Missed Failure opens a frozen snapshot with a touch-safe SVG overlay. Draw one box around the specific failed area, choose the failure type, and save. The backend stores normalized ROI coordinates plus a full-frame image, tight ROI crop, and padded context crop under data/ai_feedback_frames/<printer_id>/. This is review/training evidence only in v1.2.60; it does not change auto-pause or cancel permissions yet.

Feedback is interpreted against what Portal AI believed at the time:

AI state	User feedback	Interpreted outcome
Warning	Looks Bad	True positive
Warning	Looks Good / False Alarm	False positive
OK	Looks Bad	False negative
OK	Looks Good	True negative

False-positive feedback can create a temporary suppression for similar low/severity warnings on the same active print. This helps stop repeated “same thing again” warnings without changing your manual heuristic thresholds.

Review endpoints:

GET /api/ai/feedback/recent
GET /api/ai/feedback/stats
GET /api/ai/feedback/suppressions
GET /api/ai/learning/samples
GET /api/ai/learning/samples/<sample_id>/frame
GET /api/ai/learning/samples/<sample_id>/roi-frame
POST /api/ai/learning/import-jsonl
GET /api/printers/<printer_id>/ai/learning/samples
POST /api/printers/<printer_id>/ai/feedback/frame
POST /api/printers/<printer_id>/ai/feedback/reason

Manual threshold values remain manual. Feedback suppression does not silently rewrite your dark-frame or fine-edge thresholds.

Persistent AI learning

cc2-dash now includes a lightweight SQLite-backed learning foundation for Portal AI feedback. The goal is long-term tuning while keeping resource use reasonable on Raspberry Pi-class hardware.

Files used:

data/ai_feedback.jsonl
data/ai_feedback_frames/<printer_id>/
data/ai_learning.sqlite3

How it works in this version:

Feedback is still written to the human-readable JSONL audit log.
The same feedback is mirrored into data/ai_learning.sqlite3 as structured samples.
Samples are grouped per printer.
Rebuild endpoints calculate per-printer learning profiles, outcome counts, normal baselines, and suggested threshold modifiers.
Settings → Portal AI shows the AI Feedback Learning controls and profile cards.
Manual threshold settings are not overwritten.
Default mode is suggest_only, so learned modifiers are calculated and shown but not applied to live detection unless you explicitly switch to auto_adjust_safe.
In auto_adjust_safe, bounded learned modifiers are applied only to the live in-memory vision check thresholds; your manual settings are still not overwritten.
Failure Detection auto-pause is opt-in. When enabled, high-risk active-print failures can arm a countdown and send PAUSE_PRINT if not cancelled. Cancel print, resume, load/unload filament, and other job-control actions remain manual.

Learning modes under portal_ai config:

Mode	Behavior
`off`	Store feedback but ignore learning suggestions.
`suggest_only`	Calculate suggested modifiers and expose them through the API; live detection uses manual settings.
`auto_adjust_safe`	Apply small bounded modifiers to live thresholds. Manual settings remain unchanged.

Current bounds/defaults:

Setting	Default
`ai_learning_min_samples`	`8`
`ai_learning_min_false_positives`	`4`
`ai_learning_min_false_negatives`	`2`
`ai_learning_max_dark_luma_adjustment`	`8`
`ai_learning_max_edge_density_adjustment`	`0.05`
`ai_learning_max_required_bad_checks_adjustment`	`1`

The learning database uses Python's built-in sqlite3 module with WAL mode, normal sync, a short busy timeout, and no image blobs. Images stay on disk; SQLite stores paths and metrics only.

Settings → Portal AI now includes AI Feedback Learning controls for:

enabling/disabling persistent learning;
switching between off, suggest_only, and auto_adjust_safe;
tuning minimum sample requirements and clamp bounds;
choosing which modifier types may be suggested;
rebuilding profiles;
resetting learned tuning without deleting JSONL feedback;
viewing per-printer sample counts, outcomes, baselines, reasons, and manual/suggested/applied/effective thresholds;
reviewing recent feedback samples with filters for printer, label, outcome, metrics, reasons, and captured feedback frames;
importing/backfilling older data/ai_feedback.jsonl audit rows into SQLite with duplicate skipping and optional profile rebuild.

Important

Learned effective thresholds are wired into live vision checks only when auto_adjust_safe is explicitly selected. off and suggest_only remain advisory/no-op for live scoring. Feedback reason chips enrich future samples but still do not let AI control print jobs automatically.

AI Training review page

The /ai-training page is a lightweight local review console for Portal AI feedback samples. It is meant to make the SQLite learner inspectable without needing to SSH into the Pi or manually query the database.

Current AI Training tools include:

Filter feedback samples by printer, label, and interpreted outcome.
Review captured feedback frame thumbnails when available.
Edit/relabel a sample's feedback label, interpreted outcome, and reason note.
Delete bad SQLite training samples while keeping the JSONL audit log and captured frame files intact.
Export a lightweight ZIP dataset containing public sample metadata, raw JSONL rows, and optionally captured frame files.
Export a full restore-ready learning backup ZIP containing the SQLite learner database, JSONL audit log, profiles/events metadata, and feedback frame/ROI crop files.
Preview an uploaded learning backup before importing it.
Merge a backup into the current learner, skipping obvious duplicate SQLite samples.
Replace the current learning library from a backup after explicit confirmation. Replace mode creates a local pre-import backup ZIP under data/ai_import_backups/ before overwriting the current learner.

Import modes:

Mode	Behavior	Best use
Merge	Adds imported samples/frames to the current learner and rebuilds profiles. Existing matching samples are skipped where possible.	Combining old and current learning data
Replace	Restores the uploaded backup as the active learning library. This overwrites the current SQLite learner, JSONL audit log, and feedback frame library after creating a local pre-import backup.	Fresh install restore or full rollback

The full backup export endpoint is:

GET  /api/ai/learning/backup/export
POST /api/ai/learning/backup/import

The page does not train an Ollama model, upload data, or send commands to the printer. It only reviews, imports, exports, and cleans local feedback/training records used by cc2-dash's lightweight heuristic learner.

Upload page

The Upload page is enabled in this build and is meant for a cautious, review-before-send workflow.

Current upload flow:

Choose a local .gcode file.
cc2-dash stages the file under data/staged_gcode_uploads/.
The page shows upload progress while the browser sends the file to cc2-dash.
cc2-dash parses common slicer metadata, estimates motion bounds where possible, calculates hashes, and extracts embedded thumbnail images when available.
After review, you can send the file to the printer, or send it and request print start.

Useful staged-upload endpoints:

GET  /upload
POST /api/uploads/stage
GET  /api/uploads/<upload_id>
GET  /api/uploads/<upload_id>/thumbnail
POST /api/printers/<printer_id>/uploads/<upload_id>/send

Compatibility/direct-upload endpoint retained for simpler clients:

POST /api/printers/<printer_id>/files/upload

Caution

Upload and especially upload-and-print are real printer actions. Keep command permissions restricted, verify the staged file details, and compare with the stock portal/slicer workflow if your firmware behaves differently.

File Manager

The File Manager is experimental and locked off in this public test build. The source code and API handlers remain in place, but the menu toggle is disabled and direct access returns a disabled-feature page until the release gate is removed.

Sections:

Section	Purpose
Printer Files	Stock-style local printer file list.
USB Drive	Stock-style USB/u-disk file list with folder navigation.
Print History	Print history records where firmware reports them.
Video List	Timelapse/video records derived from stock history/video metadata. Videos with status `1` must be exported/generated first; Download is enabled after export completes/status becomes ready.

Each File Manager section has mobile-friendly search, filter, and sort controls. Printer Files and USB Drive can filter by folders, G-code, images, videos, or other files. Print History can filter timelapse/completed/failed rows. Video List can filter ready, needs-export, generating, or failed rows. Sorting is client-side after the list loads, so changing search/sort does not re-query the printer.

The Info buttons are themed from the active cc2-dash theme and open a responsive modal instead of a browser alert. The modal shows a friendly file/history summary, an optional G-code thumbnail preview when firmware provides one, and an expandable raw printer response for troubleshooting.

Bulk deletion is available from the selection toolbar in each File Manager section. On mobile, tap the large Select control on each row, then use Delete selected. Select all visible respects the active search/filter results, making it safer to delete a targeted group without selecting the whole printer list. Printer and USB file deletes are sent one at a time with visible progress, so cleaning up a pile of files no longer requires one confirmation per file. Local Printer Files deletion uses the stock portal shape for method 1047: file_path is sent as a selected-file array even for one file, and cc2-dash normalizes accidental /local/... paths back to plain filenames before sending. Print History and Video List rows can also be selected and deleted together through the stock history delete command.

Stock command IDs used include:

1036  Get history task
1037  Get history task detail
1038  Delete history
1044  Get file list
1045  Get file thumbnail
1046  Get file detail
1047  Delete file
1051  Get/export timelapse video list

Timelapse export/download flow:

1. Open Files → Video List.
2. Rows marked needs export/status 1 show Download disabled.
3. Tap Export. cc2-dash starts a backend export job and the UI shows Time-lapse video generating… instead of holding the browser request open.
4. The backend keeps polling the printer Video List until that row is actually reported as generated/download-ready, instead of trusting the initial export-command acknowledgement.
5. The frontend polls the lightweight cc2-dash job status until the backend confirms readiness or times out.
6. Once ready, refresh/download uses cc2-dash's proxied `/download` route so the printer PIN/internal URL is not exposed to the browser.

API helpers:

POST /api/printers/<printer_id>/timelapse/export
GET  /api/printers/<printer_id>/timelapse/export/<job_id>
GET  /api/printers/<printer_id>/timelapse/download?file_name=<printer-file-token>

Caution

The stock firmware may not reliably generate/export timelapse videos even when the stock portal shows the UI. cc2-dash now starts timelapse export as a backend job, shows Time-lapse video generating…, polls the printer Video List for confirmed generated/download-ready status, and only enables Download once a generated file is available. It still cannot fix firmware-side export failures.

Filament Manager / CANVAS controls

The Filament Manager is experimental and locked off in this public test build. The source code and API handlers remain in place, but the menu toggle is disabled and direct access returns a disabled-feature page until CANVAS/MMS behavior is tested more broadly.

Current CANVAS/MMS features:

Read CANVAS status.
Display filament slot cards.
Display color swatches.
Display filament metadata where firmware reports it.
Display filament sensor state with improved normalization.
Slot layout order: 1, 4, 2, 3.
Load/feed selected slot.
Unload selected slot.
Edit selected filament profile.
Toggle Auto Filament Refill.
Refresh from printer after edit/load/unload/refill changes.
Lock load/unload/edit controls unless the printer is idle.

Stock command IDs used include:

2001  Load/feed filament
2002  Unload filament
2003  Edit CANVAS filament info
2004  Auto Filament Refill
2005  Get CANVAS status
1055  Set mono filament info
1061  Get mono filament info

Warning

Filament load/unload physically moves filament. Keep the printer supervised while testing. The UI blocks these actions during active prints, but firmware behavior still needs real-world validation.

Control page

The Control page is enabled in this build. It is still treated as command-sensitive and firmware-specific, but it is no longer behind the public-release master lock.

Current control features:

Compact live camera relay panel.
Current printer connection/status summary.
Live X/Y/Z position display.
Stock-style X/Y jog wheel and Z jog rail.
Home all axes, home X/Y, and individual X/Y/Z home buttons.
Jog step buttons for 0.1mm, 1mm, 10mm, and 30mm.
Print speed presets: Silent 50%, Balanced 100%, Sport 130%, and Ludicrous 160%.
Model, Assistance/Auxiliary, and Case/Box fan controls using stock portal field names.
Fan UI displays 0–100%, while outgoing method 1030 values are converted to the stock portal's 0–255 PWM-style scale.
Light toggle using the same stock-style power payload as the dashboard light control.
Current/target extruder and bed temperature display.
Chamber temperature readout when firmware telemetry provides it.
Set extruder target temperature, bed target temperature, and turn either heater off.
Compact themed temperature inputs populated from the current target/set values.
Live status refresh that avoids overlapping refreshes and avoids overwriting fan/temp inputs while you are typing.

Stock command IDs used include:

1026  Home axes, payload {homed_axes}
1027  Move/jog axes, payload {axes, distance}
1028  Set temperature, payload {extruder} or {heater_bed}
1029  Light toggle, payload {power}
1030  Set fan speed, payload {fan}, {aux_fan}, or {box_fan} using 0-255 values
1031  Set print speed mode, payload {mode}

Speed mode mapping:

Mode	UI label
`0`	Silent 50%
`1`	Balanced 100%
`2`	Sport 130%
`3`	Ludicrous / Frenzy 160%

Temperature limits in the UI/backend:

Heater	Range
Extruder	`0–350°C`
Bed	`0–110°C`
Chamber	Read-only telemetry

Safety behavior:

All Control page commands are blocked while a print job is active.
All Control page commands are blocked while the printer is offline/stale/connecting.
Fan, speed, light, and temperature controls require Commands enabled for the printer.
Jog and home controls require both Commands enabled and Dangerous commands enabled.
The page can be hidden from the top navigation in Settings, but it is not part of the File/Filament master release lock.

Caution

Movement, homing, fan, heater, and speed controls are real printer commands. Keep eyes on the machine. A browser UI should not be your only safety plan.

Stock Elegoo portal bridge

Routes:

/portal-fullscreen
/portal
/elegoo/octo_portal.html

Local MQTT WebSocket bridge:

/ws/mqtt/<printer_id>

The bridge shuttles browser WebSocket MQTT frames to the printer's MQTT port, usually 1883.

The stock portal remains the fallback/reference view. If a cc2-dash feature is experimental or firmware-specific, compare behavior against the stock portal.

Themes and appearance

Themes live in:

cc2_dash/themes.py

Current built-in themes:

Theme	Style
Octo Dark Blue	Clean dark blue dashboard
Amber Terminal	Warm terminal / CRT-ish style
Mainsail-ish Dark	Familiar printer-dashboard dark UI
Carbon Glass	Dark translucent glass panels
Toxic Green Lab	Green terminal/lab console vibe
Blood Red Terminal	Red horror-terminal look
Elegoo Dark	Closer to stock portal colors
Klipper Blue	Blue printer-dashboard theme
OLED Mono	Minimal black/white high readability
Cyberpunk Magenta	Neon magenta/cyan chaos, in a good way
High Contrast	Accessibility-focused contrast

Theme preview cards are available in:

Setup wizard → UI step
Settings → Theme + Fonts

Font stacks are CSS-based. No font files are bundled.

Logs and diagnostics

Logs page:

/logs

Persisted log file:

data/logs/system.jsonl

Common log sources:

system
app
setup
settings
scanner
command
portal_ai
vision
filament
files
control

Useful diagnostics:

GET /health
GET /api/version
GET /api/status
GET /api/ai/monitor
GET /api/camera/status

Safety gates and command behavior

Per-printer command permissions are configured in:

Settings → Printer Manager

There are two important permission layers:

Permission	Meaning
Commands enabled	Allows normal printer command actions.
Dangerous commands enabled	Allows riskier actions such as cancel/delete/start-style operations.

Current command mapping summary:

Feature	Method / behavior
File listing/history	`1036`, `1037`, `1044`, `1046`, `1051`
File/history delete	`1038`, `1047`
Filament Manager	`2001`, `2002`, `2003`, `2004`, `2005`, `1055`, `1061`
Control page jog/home	`1026`, `1027` gated by dangerous-command permission
Control page fans/speed/light/temp	`1030`, `1031`, `1029`, `1028` gated by command permission
Control page fan value scaling	UI percent is converted to 0–255 stock portal fan values
Control page temperature	`1028` with `{extruder}` or `{heater_bed}`; chamber is read-only telemetry
Light toggle	`1029`
Upload staged G-code	Stage locally, then HTTP upload to printer; upload-and-print can request print start after transfer
Pause print	`1021`
Resume print	`1023`
Cancel print	`1022`
Camera wake/enable	`1042` / `1054`
Speed preset	`1031` stock mode payload on Control page; dashboard legacy speed action remains available
Analyze Camera Now	Server-side advisory vision check only

Speed preset modes:

Mode	Label
`0`	Silent
`1`	Balanced
`2`	Sport
`3`	Ludicrous / Frenzy

Again: Failure Detection auto-pause is opt-in and pause-only. It never sends cancel print automatically.

Configuration and data paths

Default runtime data folder:

./data/

Important files:

data/config.json
data/logs/system.jsonl
data/vision/<printer_id>/latest.jpg
data/tmp_uploads/
data/staged_gcode_uploads/
data/staged_gcode_uploads/meta/
data/staged_gcode_uploads/thumbnails/
data/ai_feedback.jsonl
data/ai_feedback_frames/<printer_id>/
data/ai_feedback_suppressions.json
data/ai_learning.sqlite3

Useful environment variables:

export CC2_DATA_DIR=/path/to/data
export CC2_CONFIG=/path/to/config.json
export CC2_PORT=8088

Default access allowlist:

192.168.1.0/24
localhost

Configure this during setup or later in Settings. Keep it restricted to trusted LAN ranges.

Useful API endpoints

Multi-printer / AI monitoring:

GET  /api/multi-view/status
GET  /api/ai/monitor
GET  /api/ai/global-alerts

General:

GET /health
GET /api/version
GET /api/status

Camera:

GET  /api/printers/<printer_id>/camera/stream
GET  /api/printers/<printer_id>/camera/latest.jpg
GET  /api/printers/<printer_id>/camera/snapshot.jpg
GET  /api/printers/<printer_id>/camera/status
POST /api/printers/<printer_id>/camera/restart

AI / vision:

GET  /api/ai/monitor
GET  /api/printers/<printer_id>/ai/status
POST /api/printers/<printer_id>/ai/check-now
POST /api/printers/<printer_id>/ai/feedback
POST /api/printers/<printer_id>/ai/feedback/frame
POST /api/printers/<printer_id>/ai/feedback/reason
GET  /api/ai/feedback/frame
GET  /api/ai/feedback/recent
GET  /api/ai/feedback/stats
GET  /api/ai/feedback/suppressions
GET  /api/ai/learning/status
POST /api/ai/learning/rebuild
POST /api/ai/learning/reset
POST /api/ai/learning/import-jsonl
GET  /api/printers/<printer_id>/ai/learning
POST /api/printers/<printer_id>/ai/learning/rebuild
POST /api/printers/<printer_id>/ai/learning/reset
GET  /api/printers/<printer_id>/ai/learning/samples
GET  /api/ai/learning/samples/<sample_id>/roi-frame
GET  /api/printers/<printer_id>/vision/status
POST /api/printers/<printer_id>/vision/check-now
GET  /api/printers/<printer_id>/vision/latest.jpg
GET  /api/vision/models
POST /api/vision/pull

Upload:

GET  /upload
POST /api/uploads/stage
GET  /api/uploads/<upload_id>
GET  /api/uploads/<upload_id>/thumbnail
POST /api/printers/<printer_id>/uploads/<upload_id>/send
POST /api/printers/<printer_id>/files/upload

Control:

GET  /control
GET  /api/printers/<printer_id>/control/status
POST /api/printers/<printer_id>/control/fan
POST /api/printers/<printer_id>/control/temperature
POST /api/printers/<printer_id>/control/speed
POST /api/printers/<printer_id>/control/move
POST /api/printers/<printer_id>/control/home
POST /api/printers/<printer_id>/control/light

Stock portal bridge:

GET /portal
GET /portal-fullscreen
GET /elegoo/octo_portal.html
WS  /ws/mqtt/<printer_id>

Troubleshooting

The dashboard will not start

Run:

./install.sh
./run.sh

If using systemd:

sudo systemctl status cc2-dash --no-pager
sudo journalctl -u cc2-dash -f

Browser cannot reach the dashboard

Check:

Pi IP address.
Port, default 8088.
Firewall/router rules.
cc2-dash access allowlist.
Whether the service is running.

Scan does not find the printer

Try:

Confirm the printer is powered on.
Confirm the Pi and printer are on the same LAN/subnet.
Try direct printer IP in setup/manual add.
Confirm printer serial/SN and access code.
Check Logs → scanner.

The scan UI only shows verified Centauri responses. A generic open web port does not count.

Stock portal opens but does not control the printer

Check:

Printer serial/SN.
PIN/access code.
MQTT port, usually 1883.
Printer Manager command toggles.
Browser console.
Logs → command.

Camera stream is flaky

Try:

Enable Camera Relay.
Restart the camera relay from Settings or API.
Close other direct camera viewers.
Avoid opening printer :8080 directly in multiple tabs.
Check /api/camera/status.

Ollama model list does not load

Check:

Ollama is running.
The URL includes protocol and port, for example http://192.168.1.24:11434.
The Pi can reach the Ollama host.
The selected vision model is installed.

Failure Detection does nothing while idle

That is expected. Current behavior is active-print-only monitoring. The loop still wakes lightly to check status so it can resume when a print starts, but it avoids heavy AI/vision work while idle.

Control page buttons fail with error 1003

Error 1003 means the printer rejected the command parameters. cc2-dash now uses the stock portal payload shapes for Control commands, but firmware differences can still happen. Check:

Printer is online and not actively printing.
Printer Manager → Commands enabled.
Printer Manager → Dangerous commands enabled for jog/home only.
The stock portal can perform the same command.
Logs → command for the raw method/payload/response.

Fan percentages look wrong

The CC2 stock portal shows fans as percentages, but the command payload uses 0–255 style values. cc2-dash converts both directions. If a fan still looks wrong, compare with the stock portal and check whether the firmware reports the fan under fan, aux_fan, box_fan, ModelFan, AuxiliaryFan, or BoxFan.

Temperature inputs do not match the printer

The Control page prefers the firmware-reported target/set temperature. If target temperature is missing from telemetry, it may fall back to blank or current values depending on what the printer reports. Compare against the stock portal and check Logs → command after sending a new target.

Printer Files delete fails with error 1003

Error 1003 means the printer rejected the delete command parameters. For local Printer Files, cc2-dash now matches the stock Elegoo portal by sending method 1047 with file_path as an array, even for a single selected file. It also strips accidental local path prefixes before sending. If this still fails, compare the same file delete in the stock portal and check Logs → files/command for the raw payload/response.

File Manager video download/export fails

For timelapse rows, use Export first. cc2-dash starts the export in the backend and shows Time-lapse video generating… while the printer creates the MP4. Download stays disabled until the printer Video List reports the video as generated/download-ready. The initial export command can return before the MP4 is done, so cc2-dash keeps showing generation status until the list confirms readiness.

If export eventually errors or times out, refresh Video List and compare with the stock Elegoo portal. The backend waits up to about 30 minutes for the printer to mark the video as generated, and the phone UI polls for up to about 35 minutes. The printer firmware may still finish generation after cc2-dash polling stops, but cc2-dash cannot force firmware-side timelapse creation if the stock portal also fails.

Filament sensor says unknown

The app normalizes several known stock/raw sensor paths, but firmware may report different shapes depending on mode, CANVAS state, or printer firmware. Check Logs → filament and compare against the stock portal.

Known limitations

Private/hobby LAN use only; not production-hardened.
Windows is untested.
AI/vision monitoring is advisory only.
AI does not automatically cancel jobs. Optional auto-pause is opt-in, countdown-gated, and pause-only.
Vision checks can produce false positives and false negatives.
Camera quality, lighting, glare, focus, and angle matter a lot.
Firmware response shapes may vary by version.
Timelapse/video export may not work even in the stock portal.
File Manager and Filament Manager remain locked and firmware-sensitive. Control and Upload are enabled but still firmware-sensitive.
No frontend build is required, but this also means the UI is intentionally simple and dependency-light.

Uninstall

Remove the service while keeping local data:

./uninstall.sh

Remove service plus .venv and data/:

./uninstall.sh --purge

Caution

--purge deletes local configuration, logs, vision frames, and AI feedback data.

Project layout

cc2-dash/
├── cc2_dash/
│   ├── __init__.py
│   ├── ai.py
│   ├── build_info.py
│   ├── camera_proxy.py
│   ├── config.py
│   ├── feedback_learning.py
│   ├── logger.py
│   ├── print_state.py
│   ├── printer_client.py
│   ├── scanner.py
│   ├── themes.py
│   ├── vision.py
│   ├── cc2/
│   │   ├── client.py
│   │   ├── commands.py
│   │   ├── discovery.py
│   │   ├── runtime.py
│   │   └── state.py
│   └── elegoo_web/
│       ├── cc2dash-camera-shim.js
│       ├── cc2dash-shim.js
│       └── octo_portal.html
├── static/
│   ├── app.css
│   └── app.js
├── templates/
│   ├── base.html
│   ├── control.html
│   ├── feature_disabled.html
│   ├── filaments.html
│   ├── files.html
│   ├── index.html
│   ├── kiosk.html
│   ├── logs.html
│   ├── multi_view.html
│   ├── portal.html
│   ├── settings.html
│   ├── setup.html
│   └── upload.html
├── install.sh
├── run.sh
├── uninstall.sh
├── requirements.txt
└── README.md

Release notes

v1.2.76 Klipper device portal controls

Made portal links type-aware for Klipper/Moonraker printers. The Dashboard button now says Open Device Portal and opens the configured Moonraker/device URL directly in a new tab/window instead of using the Elegoo portal wording.
Updated the top navigation portal label for Klipper printers to Device while keeping Carbon 2 printers on the stock Portal flow.
Added a Klipper-safe Control page job panel with pause, resume, and cancel buttons routed through Moonraker and still gated by each printer's command permission toggles.
Disabled/hardened Carbon 2-only controls on the Control page when the selected printer is Klipper, so fan, heater, light, movement, homing, and speed controls do not send unsupported CC2 commands.
Kept Carbon 2 scan/discovery, MQTT runtime, stock portal, and existing controls untouched.

v1.2.75 Klipper / Moonraker basic manual add

Added a first-pass klipper / moonraker printer type.
Kept Klipper discovery intentionally manual-only; Carbon 2 scan/discovery is unchanged.
Added Settings → Printer Manager controls for Moonraker host/port, optional API key, optional camera stream/snapshot URLs, and command permission toggles.
Added Moonraker status normalization for print state, current file, progress, hotend, bed, and chamber/temperature sensors.
Added Klipper camera snapshot/stream support through the existing cc2-dash camera endpoints.
Added gated Klipper pause/resume/cancel via Moonraker when explicitly allowed in the printer config.
Routed Klipper printers through Dashboard, Kiosk, Multi-View, and AI monitoring without changing the Carbon 2 scan or MQTT runtime path.

v1.2.74 Multi-printer AI scheduler

Reworked the background Failure Detection watchdog into a multi-printer scheduler with per-printer due times.
Added staggered background checks so multiple printers do not all hit camera/Ollama on the same tick.
Added Settings → Portal AI controls for multi-printer scheduler enablement, max checks per tick, stagger seconds, viewed-printer priority, and the global AI warning banner.
Kept auto-pause state isolated by printer_id, including countdowns, cancellations, cooldowns, and sent-pause tracking.
Auto-pause countdowns can now pull a printer forward in the scheduler so due warnings are rechecked near their deadline.
Added /api/ai/global-alerts for lightweight cached cross-printer warning polling.
Added a themed global warning banner that can point you to another printer with a high-risk AI/auto-pause warning.
Expanded /api/ai/monitor with per-printer scheduler state and recently viewed-printer hints for diagnostics.

v1.2.73 Multi-View dashboard

Added a Multi-View page showing every visible configured printer as a mobile-friendly snapshot/status card.
Each printer card shows camera snapshot, state, active file, progress, hotend/bed/chamber temperatures, and cached Failure Detection/AI badge.
Cards link to full Dashboard, Control, and Files views using the selected printer context.
Added /api/multi-view/status for lightweight multi-printer overview refreshes.
Hid the global printer selector when only one visible printer is configured.
Added DUMMY_PRINTERS_ENABLED in cc2_dash/config.py so public builds can hide/block dummy printers without deleting saved dummy entries.
Added a Multi-View menu visibility toggle in Settings and updated README docs.

v1.2.72 Dummy printer simulator

Added a safe Dummy / simulator printer type for testing multi-printer UI flows without needing a second physical CC2.
Settings → Printer Manager can now add dummy printers and edit scenario, progress, temperature, and AI badge states.
Dummy printers appear in the global header printer switcher and render through the normal Dashboard/Kiosk/Control status paths.
Added generated local dummy camera snapshot/MJPEG stream, fake telemetry, sample File Manager rows, sample history/timelapse rows, and no-op command responses for UI testing.
Dummy printers never start an MQTT client and never send real printer commands.

v1.2.71 Multi-printer view context foundation

Added a global header Printer selector that appears when configured printers exist.
Split the idea of the saved default printer from the currently viewed printer. The default printer remains a fallback, while pages can view a specific printer using ?printer=<printer_id>.
Updated main navigation and footer links to preserve the selected printer across Dash, Upload, Files, Filament, Control, Settings, AI Training, and Logs.
Dashboard refresh and quick actions now target the selected/viewed printer instead of silently falling back to the saved default.
Remembered the last viewed printer in browser localStorage so mobile browsers return to the same printer unless a URL explicitly chooses another one.
Added a light printer type/printer_type config field for future dummy/simulator printer work without changing existing CC2 behavior.

v1.2.70 Chamber readout-only polish

Changed dashboard chamber temperature display from current / off to a single read-only temperature value.
Kept Hotend and Bed as current / target because those are controllable heaters, while Chamber is telemetry-only on the CC2 firmware.
Updated the vision context wording so chamber temperature is treated as a read-only sensor rather than a controllable target/setpoint.

v1.2.69 Chamber temperature status

Added chamber temperature normalization from CC2 telemetry aliases such as ztemperature_sensor.temperature and chamber.temperature.
Dashboard Print Status now shows Chamber beside Hotend and Bed when telemetry is available.
Control page temperature section now includes a read-only Chamber card, matching the stock portal's chamber temperature visibility without exposing unsupported chamber heater controls.
Status/API payloads now include chamber_current and optional chamber_target when firmware ever exposes it, and the vision prompt context includes chamber temperature for better print-environment awareness.

v1.2.68 AI Training backup and restore

Added a full AI Training backup export separate from the existing lightweight dataset export.
Backup ZIPs include a manifest, SQLite learner database, JSONL audit log, public/raw sample exports, learning profiles/events metadata, and feedback frame/ROI crop files when selected.
Added AI Training import UI with mobile-friendly backup file picker, Preview Import, Merge, and Replace flows.
Import preview shows backup schema, sample/profile/event counts, frame count, included data types, and warnings before anything is changed.
Merge mode imports JSONL sample rows and frame files into the existing learner while skipping obvious duplicates.
Replace mode requires explicit confirmation, creates a local pre-import backup under data/ai_import_backups/, then restores the uploaded backup as the active learning library.
Profiles are rebuilt after import so restored samples can immediately contribute to the lightweight heuristic learner.

v1.2.67 File Manager themed info modal

Replaced the old File Manager alert()/console-only Info behavior with a themed, mobile-friendly modal.
The Printer Files, USB Drive, and Print History Info buttons now use active cc2-dash theme colors.
File info modals show a friendly summary, optional G-code thumbnail preview when the printer exposes one, and an expandable raw printer response for troubleshooting.
Added tap-backdrop, close-button, and Escape-key modal dismissal.
Updated README File Manager documentation and release notes.

v1.2.66 File Manager search, filter, and sort

Added mobile-friendly search, type/status filters, and sort controls to File Manager sections.
Printer Files and USB Drive can now be filtered by folders, G-code, image, video, or other files and sorted by name, date, size, or type.
Print History can now be searched, filtered by timelapse/completed/failed rows, and sorted by date, name, or size.
Video List can now be searched, filtered by ready/needs-export/generating/failed states, and sorted by date, name, size, or status.
Select all visible now respects the active search/filter results, making bulk delete safer and less annoying on mobile.

v1.2.65 Printer Files delete payload fix

Fixed Printer Files deletion returning 1003 — Invalid parameter on firmware builds that require the stock portal's array-shaped delete payload.
Method 1047 now sends file_path as a selected-file array even when deleting one local printer file, matching the stock Elegoo portal behavior.
Local printer-file delete paths are normalized before sending so /local/foo.gcode, local/foo.gcode, and foo.gcode all become the stock-style foo.gcode.
Kept a legacy string-payload retry only as a fallback for firmware variants that reject the stock array shape.
README troubleshooting/docs updated for Printer Files deletion behavior.

v1.2.64 File Manager multi-select polish

The File Manager now has touch-friendly multi-select toolbars for Printer Files, USB Drive, Print History, and Video List.
Printer/USB file rows include large mobile-safe selection controls plus Select all visible, Clear, and Delete selected actions.
Bulk file deletion sends delete commands one at a time with visible progress, so mobile browsers are not stuck doing repetitive one-file confirms.
Print History and Video List rows can also be selected and deleted together through the existing history delete command.
The timelapse Export button now uses the configured cc2-dash theme colors instead of looking like an uninvited stock portal gremlin.

v1.2.63 timelapse export confirmation

Fixed timelapse export jobs reporting complete too early on longer videos by treating the initial firmware export response as an acknowledgement only.
Backend export jobs now keep polling the printer Video List until the selected row reports generated/download-ready before exposing the Download URL.
Extended timelapse export waiting to better handle longer videos: backend timeout is about 30 minutes; mobile UI polling is about 35 minutes.
Added a friendly mapping for sub-status 3020 so the dashboard shows Time-lapse video generating instead of Sub 3020.
Updated File Manager docs and troubleshooting notes for confirmed export readiness.

v1.2.62 async timelapse export

Reworked File Manager → Video List timelapse export so the browser no longer waits on the long-running firmware export request.
Added backend timelapse export jobs with polling status: generating, ready/complete, or error.
Updated the Video List UI so status 1/not-yet-generated rows require Export before Download; Download is disabled until the generated video is ready.
Added visible Time-lapse video generating… status with row-level spinner/state so the phone browser can sit on a lightweight polling loop instead of timing out.
Kept downloads proxied through cc2-dash after export so the stock printer /download URL/PIN stays behind the dashboard.
Updated File Manager documentation and troubleshooting notes for the export-first workflow.

v1.2.61 ROI feedback mobile/cache fix

Adds cache-busting query strings to /static/app.css and /static/app.js so phone browsers pick up new dashboard JavaScript/CSS after upgrades.
Themes the Report Missed Failure button with a fallback for browsers that do not support newer CSS color-mix rules.
Prevents the Report Missed Failure button from also firing the normal generic feedback handler.
Makes the ROI feedback modal opening handler delegated and mobile-safe so it still works if dashboard markup is refreshed later.

v1.2.60 ROI missed-failure feedback

Added a Report Missed Failure dashboard workflow for localized failures the detector missed, especially detached small parts, fallen prime towers, and air-printing zones.
Added a mobile-safe ROI annotation modal using a frozen still frame plus SVG/pointer-event drawing, so mouse, touch, and stylus all use the same code path.
Added POST /api/printers/<printer_id>/ai/feedback/frame to capture a still feedback frame before the user draws an ROI box.
Extended POST /api/printers/<printer_id>/ai/feedback with optional ROI annotation metadata while keeping existing Looks Good / Looks Bad / False Alarm feedback compatible.
Saved normalized ROI coordinates, tight ROI crops, padded context crops, and annotation metadata into JSONL/SQLite raw training data and AI Training review/export flows.
Added safe image-serving endpoints for feedback frames and ROI crops. ROI evidence is stored for learning/review only in this release and does not alter auto-pause behavior.

v1.2.59 failure-detection pause gate

Added a shared print-state helper module so dashboard status, Portal AI, vision checks, and auto-pause permission use the same preparation/active-print/pause-safety classification instead of drifting apart across files.
Added a single Failure Detection pause-permission gate that returns explicit allowed actions, veto reasons, evidence, failure family, and a hard cancel_allowed: false decision.
Hardened auto-pause so a countdown only arms when the gate allows pause, and the backend performs a fresh status + forced vision recheck before sending PAUSE_PRINT.
Hardened the dashboard Pause now action so it requires the active auto-pause token and also passes the same fresh recheck/gate before sending a pause command.
Made camera/view-quality and telemetry-only warnings inspect/warn states rather than auto-pause triggers, keeping auto-pause reserved for pause-grade print failure evidence.

v1.2.58 control, upload, error-code, and release-lock polish

Updated the top cc2-dash banner link so it opens the public GitHub repository instead of linking back to the app host.
Added printer exception-code decoding so known firmware exception IDs display with human-readable meanings, such as 1252 — Extruder unload failure / unload timeout, instead of only raw bracketed codes.
Reworked Control page command payloads to match the bundled stock Elegoo portal more closely: {homed_axes} for home, {axes, distance} for jog, {power} for light, stock fan keys, and stock speed modes.
Fixed fan percentage handling by converting the UI's 0–100% values to the stock portal's 0–255 fan command scale, and converting telemetry back for display.
Added Control page bed/extruder current/target temperature display plus set/off controls using method 1028 with {extruder} or {heater_bed}.
Polished Control page refresh behavior so telemetry updates do not overlap or stomp fan/temperature inputs while the user is editing them.
Re-enabled the community-release master lock for File Manager and Filament Manager while leaving Control enabled and protected by runtime safety gates.
Documented the enabled Upload page workflow, staged G-code review, thumbnail/metadata extraction, and upload/upload-and-print endpoints.
Updated README status, safety notes, API endpoint lists, troubleshooting, known limitations, and release-gate explanation to match the current code.

v1.2.57 community release prep

Locked the experimental File Manager and Filament Manager off for public test builds while keeping their source, templates, routes, and command code in place for later re-enabling. Control was restored in v1.2.58 with runtime safety gates.
Added a single release-gate switch in cc2_dash/config.py: COMMUNITY_RELEASE_EXPERIMENTAL_LOCKS. Set it to False to restore normal Settings toggles and direct access for listed release-locked features when they are ready.
Server-side config migration and save handling now force locked experimental feature flags to false, including when users edit raw JSON.
Settings now shows locked experimental rows with disabled controls instead of presenting them as public-ready options.
Direct experimental pages now show a clean disabled-feature notice, and command-heavy experimental APIs return a clear 403 while locked. Dashboard G-code thumbnail image support remains available.
Polished README wording to make the public-test status, safety boundaries, and experimental feature status clearer.

v1.2.56 offline state detection

Adds backend connection-health classification separate from printer job state: online, stale, offline, connecting, and registration/auth error.
Uses MQTT registration plus fresh heartbeat/PONG timing as the primary source of truth so stale cached telemetry no longer makes the dashboard keep saying Printing or Idle when the printer is disconnected.
Status payloads now include connection_state, offline, stale, connection_reason, and connection_health fields.
Dashboard, Kiosk, Control page, browser title, and Failure Detection now show Offline / Connection Stale / Connecting instead of stale job state text.
Failure Detection and auto-pause are paused while printer telemetry is offline/stale, and the Control page locks all command controls until the printer is online again.

v1.2.55 control camera relay fix

Fixed the Control page camera panel so it explicitly uses the cc2-dash camera relay stream endpoint from Control status instead of relying only on the initial template image state.
Added Control-page camera load/failure handlers that hide the loading overlay when the relay starts producing frames and retry the relay stream if the browser drops the MJPEG connection.
Removed the extra descriptive blurbs from the Control page hero and camera card to keep the page tighter on mobile.

v1.2.54 control page print lock, camera, and light toggle polish

The Control page now locks all printer command controls while a print job is active, including movement, homing, speed, fans, and the Control-page light switch. The Refresh/status path remains available.
Added a compact live camera relay panel at the top of the Control page so movement/control checks can be done without jumping back to the dashboard.
Added backend active-print protection for Control page fan, speed, move, home, and Control-page light commands.
Changed Dashboard → Quick Actions → Light from a push button into a themed pill toggle with an on/off icon and live state sync from printer telemetry.
Exposed light_on in dashboard status payloads so the dashboard and Control page can share the same light state.

v1.2.53 stock-style control page

Added a new stock-Elegoo-inspired Control page with XY/Z jog controls, homing buttons, step-size buttons, print-speed presets, fan controls for Model / Assistance / Case, and a light toggle.
Added /control plus /api/printers/{printer_id}/control/* endpoints for status, fan, speed, move, and home actions.
Added a configurable Control top-nav item in Settings → Menu / Features. It is hidden by default, like the experimental Files/Filament pages.
Reused the existing command safety gates: fan/speed/light require Commands enabled; jog/home require both Commands enabled and Dangerous commands enabled.
Extended CC2 state normalization for stock position and fan-speed fields, including CurrentFanSpeed.ModelFan, AuxiliaryFan, and BoxFan.

v1.2.52 themed toggle cleanup

Replaced remaining native checkbox styling with the same cc2-dash themed pill switch style where it fits the UI, including generated dashboard/action/printer settings rows, setup wizard options, AI Training export options, and advanced JSON override.
Themed checkbox-style controls now follow the active theme color instead of browser/default checkbox rendering.
Removed the bulky recent feedback sample browser from Settings > Failure Detection because the AI Training page already owns sample review/export workflows.
Settings still keep AI learning status, rebuild/import controls, and auto-pause configuration, but no longer duplicate the training sample list.

v1.2.51 themed failure toggle fix

Fixed the Failure Detection enable/disable switch so tapping it inside the collapsed dashboard card header reliably toggles detection instead of being eaten by the accordion summary click behavior.
The same switch behavior is now shared by the dashboard card and Settings toggle, so enabling/disabling Failure Detection stays synchronized immediately.
Restyled the dashboard and Settings switch controls with a fully custom themed pill toggle using the active cc2-dash theme color instead of browser/default checkbox styling.
The Auto-pause on high-risk failure switch now uses the same themed control style.
No changes to auto-pause safety behavior, pause command behavior, or failure scoring thresholds in this release.

v1.2.50 failure detection auto-pause

Renamed the dashboard AI Info card to Failure Detection.
Added a modern pill-style enable/disable toggle on the Failure Detection card header, visible even while collapsed.
Added live Failure Detection enable/disable behavior from the dashboard and Settings.
Added opt-in automatic pause-on-failure behavior with configurable risk threshold, countdown seconds, cooldown minutes, and high-level-only requirement.
Added a themed high-risk warning modal with red warning text, failure summary, countdown timer, Cancel Pause, Pause Now, and post-cancel feedback options.
Backend watchdog now owns pending auto-pause countdowns so the pause can still fire even if the browser view is not the only thing ticking.
Tuned failure scoring: telemetry rule disabling is honored more cleanly, startup temperature/filament grace windows reduce false alarms, and paused states no longer trigger progress-stall failures.
Auto-pause remains conservative: it only considers active, connected, non-prep, non-paused print states. Cancel print is still not automated.

v1.2.49 dashboard title and idle AI summary

Updated the collapsed Failure Detection summary pill so idle/standby telemetry shows Idle instead of the generic Looks Good label.
Added live browser tab title updates on the dashboard using the current printer status, progress percentage, and time remaining when available.
Connection trouble now also reflects in the page title so a stale/offline tab is easier to spot.
No changes to Portal AI scoring or print-prep guards in that release.

v1.2.48 collapsed status summary

Updated the collapsed Print Status card to show the printer telemetry status label, such as Bed Preheating, Extruder Preheating, Homing, Printing, or Idle, instead of always showing PRINTING during active/prep states.
The collapsed summary now keeps the progress bar visible for active print-preparation phases and uses warning-colored styling for preparation states, green for real printing, muted for idle, and red for error/offline states.
No changes to Portal AI scoring or print-prep guards in that release.

v1.2.47 print-prep AI guard

Added explicit print-preparation state detection for bed preheating, extruder preheating, homing, auto-leveling, self-checking, and initializing.
Portal AI now reports these phases as Preparing instead of Failure Likely, and the dashboard summary pill shows Preparing instead of a scary false alarm.
Vision/Ollama failure checks are paused during normal start-of-job preparation, preventing dark/empty-bed camera frames from causing false alarms.
Temperature-gap, filament-out, and progress-stall rules are also paused during preparation and resume when actual printing starts.
Added print_phase, status_code, and sub_status_code fields to status payloads for better debugging and future UI work.
No printer-control/autopause behavior changed in that release; auto-pause arrives later in v1.2.50.

v1.2.46 configurable navigation

Added Menu / Features toggles for every top navigation item except Dash and Settings.
Portal, Files, Filament, Kiosk, AI Training, and Logs can now be shown or hidden from Settings without disabling their underlying routes.
Dash and Settings remain pinned so the dashboard and configuration page cannot be hidden accidentally.
Kiosk top-nav Portal/Logs links now respect the same navigation visibility settings.
Footer Console link now respects the Logs visibility setting.
Updated config defaults/migration to preserve existing File Manager and Filament Manager hidden-by-default behavior.
No AI scoring, learning, printer command, or safety behavior changes.

v1.2.45 AI Training review page

Added /ai-training, a lightweight review/export page for Portal AI feedback samples.
Added sample review tools for relabeling feedback label, interpreted outcome, and reason/note.
Added SQLite sample deletion from the review set while keeping JSONL audit rows and frame files.
Added filtered dataset ZIP export with public metadata, raw JSONL rows, and optional captured frames.
Added navigation link for AI Training.
No changes to printer commands or AI advisory-only safety behavior.

v1.2.44 JSONL feedback import

Added explicit JSONL feedback import/backfill from data/ai_feedback.jsonl into the SQLite learning database.
Added POST /api/ai/learning/import-jsonl with duplicate skipping, malformed-line counts, reason-update replay, and optional profile rebuild.
Added Settings → Portal AI → Import old JSONL feedback controls with import summary output.
Import is manual/on-demand only and does not run on every startup.
Keeps JSONL as the human-readable audit log and stores images on disk only.
No changes to printer commands, advisory-only AI safety behavior, or automatic print control.

v1.2.43 dashboard learning badge

Adds a compact AI learning badge/details panel to the dashboard Portal AI card.
Shows whether learning is off, suggesting, or auto-adjusting safely.
Surfaces sample count, confidence, and manual/suggested/applied/effective thresholds when threshold data is available from the live vision path.
Clarifies whether learned modifiers are being applied live or only suggested.
Changes the feedback custom “Save note” button to a theme-safe success/green style instead of danger red.
Keeps manual thresholds unchanged and Portal AI advisory-only.

v1.2.42 feedback UI polish

Fixed optional Portal AI feedback training reason controls so they remain readable across themes instead of appearing as white-on-white buttons.
Reason chips, skip, and custom note save controls now use a dark red danger-style treatment with white text.
Printer Manager in Settings now loads collapsed by default, matching the other Settings panels.
No AI scoring, learning, printer command, or safety behavior changes.

v1.2.41 recent feedback samples

Added a lightweight Recent AI Feedback Samples review panel under Settings → Portal AI.
Added filters for printer, interpreted outcome, feedback label, and sample count.
Shows timestamp, printer, label, optional reason/note, outcome, file/stage/progress, risk/severity/confidence, local heuristic metrics, triggered flags, and captured feedback frame thumbnails/links when available.
Added global GET /api/ai/learning/samples with pagination/filter query parameters.
Enhanced GET /api/printers/<printer_id>/ai/learning/samples with the same lightweight public sample shape and filters.
Added GET /api/ai/learning/samples/<sample_id>/frame to safely serve captured feedback frames from data/ai_feedback_frames/.
Does not store image blobs in SQLite and does not change live AI scoring or printer-control behavior.

v1.2.40 feedback reason chips

Added optional reason chips after fast Portal AI feedback clicks.
Feedback buttons still save immediately; reason selection is a second, optional training-quality step.
Looks Bad reasons include spaghetti/stringing, detached print, blob/nozzle buildup, first-layer issue, layer shift, filament issue, camera bad/unclear, and custom notes.
False Alarm reasons include normal supports, purge tower, infill pattern, reflection/glare, low light but visible, multicolor purge mess, camera angle, and custom notes.
Looks Good reasons include normal print, normal idle, normal purge/supports, and custom notes.
Reason updates are appended to data/ai_feedback.jsonl and attached to the matching SQLite feedback sample when available.
This improves future learning/review data without changing live AI scoring, thresholds, or printer-control safety behavior.

v1.2.39 active-file layer total lookup

Improved dashboard layer progress when the live printer status only reports the current layer.
cc2-dash now mirrors the stock Elegoo portal behavior more closely by looking up the active G-code file metadata for total layers when live telemetry omits it.
Layer progress now prefers current/total from file metadata, falling back to current/? only when the file lookup cannot find a total layer count.
File metadata lookups are cached so the dashboard poll loop does not hammer the printer.
No printer-control, Portal AI, or learning behavior changes.

v1.2.38 layer progress unknown-total polish

Updated dashboard layer display so printers that report current layer but not total layers now show current/? instead of only current.
Added layer_total_missing to print metrics/status payloads so the UI/API can clearly distinguish "known current layer, unknown total layer" from a simple single-value layer display.
No live AI, printer command, or learning behavior changes in this patch.

v1.2.37 dashboard/settings polish

Settings panels now load collapsed by default so the Portal AI / learning controls are not buried under a forever-scroll wall.
Removed the unreliable Filament Used field from the dashboard Print Status card.
Updated the dashboard Layer display to prefer current/total layer values such as 18/250.
Removed the small API Reachable status field from the dashboard Print Status card to reduce noise.
Live AI learning behavior is unchanged from v1.2.36.

v1.2.36 safe auto thresholds

Wired persistent AI learning effective thresholds into the live vision monitor.
In off and suggest_only modes, live vision continues to use the manual threshold values exactly as before.
In auto_adjust_safe mode, live vision uses bounded effective values for dark luma, fine-edge density, and required bad checks.
Manual settings are still never overwritten; learned values are applied only to the in-memory vision check configuration.
Vision API results now include learning_thresholds and learning_applied so the dashboard/logs can explain when bounded modifiers were used.
Added dashboard Vision metadata showing learning mode and applied modifiers during live checks.
Added low-noise warning fallback: if the learning database/config lookup fails, vision monitoring falls back to manual thresholds instead of failing the check.
Failure Detection auto-pause is now available in v1.2.50 as an opt-in pause-only guard; cancel/resume/control actions remain manual.

v1.2.35 learning settings UI

Added a Settings → Portal AI → AI Feedback Learning section.
Added controls for persistent learning enablement, learning mode, minimum sample counts, maximum learned adjustment bounds, modifier-type toggles, and rebuild-on-feedback behavior.
Added per-printer learning profile cards showing sample counts, true/false positive/negative outcomes, manual/suggested/applied/effective thresholds, normal baselines, confidence, and explanation reasons.
Added Settings buttons to refresh learning status, rebuild all profiles, and reset learned tuning while keeping feedback samples and JSONL audit logs.
Kept live AI scoring unchanged in that version; suggest-only mode remains the default. Auto-pause arrives later in v1.2.50 as an opt-in pause-only guard.

v1.2.34 persistent AI learning foundation

Added cc2_dash/ai_learning_db.py for lightweight SQLite setup, schema creation, feedback sample inserts, profile storage, event logging, reset helpers, and health checks.
Added cc2_dash/ai_learning.py for structured sample extraction, outcome counts, per-printer profile rebuilds, normal baseline calculations, suggested modifier calculations, and effective-threshold summaries.
Added data/ai_learning.sqlite3 as a sidecar learning database. JSONL feedback logging remains intact.
Feedback clicks now mirror structured samples into SQLite while preserving the existing JSONL audit trail and same-print false-alarm suppression behavior.
Added learning status/rebuild/reset/sample APIs for global and per-printer use.
Added /health AI learning database status.
Added portal_ai config defaults for persistent learning, defaulting to suggest_only.
Kept Portal AI advisory-only; no automatic pause/cancel/control behavior was added.
Folded in project cleanup: README clone URL now points to https://github.com/merberg-ai/cc2-dash.git, internal runtime class renamed to Cc2PrinterRuntime, old cc2-dash-lite.service cleanup added, and backend 123456 access-code fallbacks removed.

v1.2.33 dashboard metrics and G-code thumbnails

Dashboard Print Status now attempts to populate Filament Used from additional stock/firmware field names including totalFilamentUsed, material weight, and filament length aliases. If firmware does not publish a usable value, the UI still shows - rather than inventing one.
Expanded Print Status now shows layer progress when available, such as 120/450.
Added optional dashboard G-code thumbnail preview for the active file. The preview only appears when the printer returns a usable thumbnail image.
Clicking the thumbnail opens a larger themed glass modal with a close button.
Added Settings → Dashboard Layout → G-code thumbnail preview to show/hide the thumbnail section.

v1.2.32 crt themes

Added two new built-in retro monitor themes: Retro CRT Blue-Gray and Green Phosphor CRT.
Both themes use the built-in Retro CRT font stack with scanline/glow styling for an old-monitor feel.
Theme preview cards in Settings and first-run setup now include the two new CRT-style themes.

v1.2.31 theme expansion

Added six built-in themes: Toxic Green Lab, Blood Red Terminal, Elegoo Dark, Klipper Blue, OLED Mono, and Cyberpunk Magenta.
Added clickable theme preview cards to Settings and first-run setup.
Existing theme/font override behavior is preserved.

v1.2.30 filament polish

Reordered CANVAS slot display to 1, 4, 2, 3.
Added refresh-after-action behavior for edit/load/unload/Auto Refill.
Locked load/feed, unload, and edit controls to idle-only behavior.
Added backend rejection while printing or during filament/extruder operation states.
Improved Auto Filament Refill behavior using the stock payload.
Improved filament sensor normalization.
Made firmware command failures louder instead of reporting fake success.

v1.2.29 filament CANVAS controls

Added stock-shaped CANVAS status, load/feed, unload, edit, and Auto Refill controls.
Added filament color swatches and richer slot metadata.
Added mono-filament helper methods where firmware exposes them.

v1.2.28 collapsed print state + filament hidden

Collapsed Print Status header shows IDLE or PRINTING with compact progress.
Filament nav item defaults hidden and can be re-enabled in Settings.

v1.2.27 idle AI standby

Normalized raw idle status code Sub 0 to Idle.
Added active-print detection.
Portal AI/watchdog/vision monitoring now stands by when idle.

v1.2.26 file manager hidden

File Manager nav item defaults hidden because firmware timelapse/export behavior appears inconsistent.
Existing stock-style file manager work remains available for later testing.

v1.2.25 timelapse download proxy

Added cc2-dash timelapse download proxy through the printer stock /download handler.
Converted export-returned video paths/tokens into dashboard download links.

v1.2.24 stock-style file manager

Reworked File Manager around stock command shapes.
Added Printer Files, USB Drive, Print History, and Video List sections.

v1.2.23 feedback learning

Added fresh-frame feedback capture.
Added true/false positive/negative interpretation.
Added current-print false-alarm suppression for similar low/severity warnings.
Improved feedback stats and suppression API.

v1.2.22 alphanumeric access codes

Updated setup/settings PIN fields to allow letters and numbers.
Removed prefilled default PIN.
Backend rejects blank access codes.

v1.2.21 setup copy cleanup

Simplified first-run setup copy.
Reduced first card to progress-only header treatment.

v1.2.20 kiosk camera warm-up

Kiosk uses faster cached status.
Improved camera placeholder/stream loading behavior.

v1.2.19 kiosk mode

Added hideable Kiosk nav item and camera-first fullscreen page.

v1.2.18 AI header status

Added compact Portal AI status pill to the collapsed AI/Failure Detection header.

v1.2.17 collapsed progress

Added compact progress bar to collapsed Print Status header.

v1.2.16 dashboard section split

Split Camera, Print Status, AI/Failure Detection, Quick Actions, and Connection into clearer collapsible sections.

v1.2.15 dashboard accordion polish

Added saved dashboard accordion state.

v1.2.14 mobile header/settings cleanup

Improved mobile header build chips.
Reworked Settings into collapsible panels with global Save All / Cancel controls.

v1.2.13 vision sanity + service cleanup

Improved benign-uncertainty handling for Ollama vision.
Improved install/uninstall systemd cleanup.

v1.2.12 portal navigation fix

Portal nav opens fullscreen stock portal in a new tab instead of nesting wrappers.

v1.2.11 camera relay

Added MJPEG relay/fanout, cached latest-frame endpoints, and portal camera rewrite shim.

v1.2.10 documentation/source cleanup

Reworked documentation and removed informal placeholder references.

v1.2.1 build metadata

Added version/commit/branch build metadata in header, /api/version, and /health.

v1.2.0 background watchdog

Added background Portal AI monitoring loop and /api/ai/monitor.

v1.0.0 stable baseline

First-run setup, stock portal bridge, mobile dashboard shell, themes, feature toggles, install scripts, and early file hooks.

Development checks

No frontend build step is required for normal use.

After edits, useful checks are:

python -m compileall cc2_dash
python - <<'PY'
from jinja2 import Environment, FileSystemLoader
from pathlib import Path

env = Environment(loader=FileSystemLoader('templates'))
for path in Path('templates').glob('*.html'):
    env.get_template(path.name)
print('templates ok')
PY
node --check static/app.js

The Node check is optional and only verifies JavaScript syntax if Node is installed.

Name		Name	Last commit message	Last commit date
cc2_dash		cc2_dash
frontend		frontend
static		static
templates		templates
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md
install.sh		install.sh
requirements.txt		requirements.txt
run.sh		run.sh
uninstall.sh		uninstall.sh

Latest commit

History

Folders and files

Repository files navigation

cc2-dash

Table of contents

Current status

Tested hardware and platform notes

Tested

Expected to work better

Windows

Feature overview

Local dashboard

Printer discovery and pairing

Dashboard controls

Camera relay

Portal AI

Feedback-aware AI review

Stock portal bridge

Experimental file and filament tools

Themes

What cc2-dash does not do

Install from GitHub on Raspberry Pi OS

1. Update the Pi and install basic tools

2. Clone the repository

3. Make helper scripts executable

4. Install Python dependencies

Run manually

Install as a systemd service

Update from GitHub

First-run setup

Using the dashboard

Active printer switching

Multi-View

Klipper / Moonraker manual add

Printer Manager

Dummy / simulator printers

Camera Relay / stream protection

Kiosk mode

Portal AI and Ollama vision

AI feedback and false-alarm suppression

Persistent AI learning

AI Training review page

Upload page

File Manager

Filament Manager / CANVAS controls

Control page

Stock Elegoo portal bridge

Themes and appearance

Logs and diagnostics

Safety gates and command behavior

Configuration and data paths

Useful API endpoints

Troubleshooting

The dashboard will not start

Browser cannot reach the dashboard

Scan does not find the printer

Stock portal opens but does not control the printer

Camera stream is flaky

Ollama model list does not load

Failure Detection does nothing while idle

Control page buttons fail with error 1003

Fan percentages look wrong

Temperature inputs do not match the printer

Printer Files delete fails with error 1003

File Manager video download/export fails

Filament sensor says unknown

Known limitations

Uninstall

Project layout

Release notes

v1.2.76 Klipper device portal controls

v1.2.75 Klipper / Moonraker basic manual add

v1.2.74 Multi-printer AI scheduler

v1.2.73 Multi-View dashboard

v1.2.72 Dummy printer simulator

v1.2.71 Multi-printer view context foundation

v1.2.70 Chamber readout-only polish

v1.2.69 Chamber temperature status

v1.2.68 AI Training backup and restore