# GRIME3 0.1.0 RC8.4 — Conda Quick Start

GRIME3 RC8.4 is distributed through the versioned CoffeeSig static Conda
channel at:

https://coffeesig.com/gaugecam/conda/rc8.4

The versioned path keeps release-candidate software isolated from other
GRIME3 Conda publication paths.

The GRIME3 package is `noarch: python`; Conda resolves platform-specific
runtime dependencies.

## Recommended RC8.4 installation

Create a dedicated environment from the CoffeeSig RC8.4 channel:

```bash
conda create -n grime3-rc8.4 \
  --override-channels \
  -c conda-forge \
  -c https://coffeesig.com/gaugecam/conda/rc8.4 \
  python=3.11 \
  grime3=0.1.0
```

Activate it:

```bash
conda activate grime3-rc8.4
```

Verify it:

```bash
grime3 --version
grime3 --help
grime3 measure --help
grime3 web --help
```

The CoffeeSig RC8.4 channel supplies GRIME3. Normal runtime dependencies are
resolved from conda-forge.

## Direct artifact installation

The exact `.conda` artifact is also distributed with RC8.4 for archival,
reproducibility, and controlled package validation.

For normal installation, install GRIME3 by package name from the RC8.4
Conda channel as shown above. This allows Conda to resolve all required
runtime dependencies and create the platform-specific command launcher.

Installing the `.conda` file directly by filesystem path is not the
recommended end-user installation method. In validation on Windows, an
explicit artifact-path installation did not populate the required runtime
dependencies in an otherwise bare environment.

The validated Windows RC8.4 installation path creates:

```text
Scripts\grime3.exe
Scripts\grime3-script.py
```

and the `grime3` command is directly available after environment activation.

## Runtime dependency policy

RC8.4 Conda uses:

```text
python >=3.10,<3.14
opencv 4.13.0 headless_*
ffmpeg * lgpl_*
```

along with NumPy, Pillow, FastAPI, Pydantic, Uvicorn, and Click.

Qt, PyTorch, torchvision, torchaudio, and SAM2 are not required by the RC8.4
Conda runtime.

The Conda OpenCV version intentionally differs from the exact PyPI headless
OpenCV wheel used for pip/PyInstaller-based RC8.4 distributions.

## Start the browser

```bash
grime3 web --open-browser
```

The normal local address is:

```text
http://127.0.0.1:8000/
```

## Measure one image

```bash
grime3 measure image \
  --image "/data/images/image001.jpg" \
  --site-config "/data/site/site_config.json" \
  --output-dir "/data/results"
```

## Measure one folder

```bash
grime3 measure folder \
  --images "/data/images" \
  --site-config "/data/site/site_config.json" \
  --output-dir "/data/results"
```

## Measure a folder of folders

```bash
grime3 measure folders \
  --images-root "/data/campaign" \
  --site-config "/data/site/site_config.json" \
  --output-dir "/data/results"
```

The browser and CLI use the same canonical Site Configuration and scientific
measurement implementation.

## Verify headless OpenCV

```bash
python - <<'PY'
import cv2

print(cv2.__version__)

for line in cv2.getBuildInformation().splitlines():
    if line.strip().startswith("GUI:"):
        print(line.strip())
PY
```

The GUI line should report `NONE`.

## Updating

For RC8.4, install newer RC8.4 builds from the same versioned channel when provided.

A later production package can use a stable CoffeeSig Conda channel path
without requiring users to select an RC-specific path.

## Remove GRIME3

```bash
conda activate grime3-rc8.4
conda remove grime3
```

## Remove the complete environment

```bash
conda deactivate
conda env remove -n grime3-rc8.4
```

## Troubleshooting

### `grime3: command not found`

```bash
conda env list
conda activate grime3-rc8.4
grime3 --version
```

### GRIME3 cannot be found by Conda

For RC8.4, make sure the versioned release-candidate channel is present:

```text
https://coffeesig.com/gaugecam/conda/rc8.4
```

The RC8.4 package is intentionally isolated from other Conda publication paths.

### Unexpected GUI OpenCV

The official recipe requires:

```text
opencv 4.13.0 headless_*
ffmpeg * lgpl_*
```

### Browser does not open automatically

Leave `grime3 web` running and open:

```text
http://127.0.0.1:8000/
```

manually.

### Data moved to another computer

Use the current filesystem paths for images, Site Configuration, and output.
Persistent Site and Configuration identities do not depend on the path where
the JSON file happens to be stored.

## Provenance

Preserve the Site Configuration, JSON, CSV, run manifest, and build identity
with scientific results.