terminalboard 0.1.0__tar.gz

terminalboard-0.1.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 dongfangyixi
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

terminalboard-0.1.0/PKG-INFO

@@ -0,0 +1,249 @@
+Metadata-Version: 2.4
+Name: terminalboard
+Version: 0.1.0
+Summary: A pure-SSH terminal TensorBoard scalar viewer — live scalar curves in iTerm2, no browser, no X11, no port forwarding.
+Author: dongfangyixi
+License: MIT
+Project-URL: Repository, https://github.com/dongfangyixi/terminalboard
+Keywords: tensorboard,terminal,iterm2,plotext,ssh,ml
+Requires-Python: >=3.9
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: plotext>=5.2
+Provides-Extra: tb
+Requires-Dist: tensorboard>=2.9; extra == "tb"
+Provides-Extra: hq
+Requires-Dist: matplotlib>=3.5; extra == "hq"
+Provides-Extra: full
+Requires-Dist: tensorboard>=2.9; extra == "full"
+Requires-Dist: matplotlib>=3.5; extra == "full"
+Dynamic: license-file
+
+# terminalboard
+
+A **pure-SSH terminal TensorBoard scalar viewer**.
+
+Train on a remote box, SSH in, and watch your **live-updating scalar curves
+right inside the terminal** — crisp Unicode/braille text by default, or real
+iTerm2 inline images with `--hq`. No browser, no X11, no port forwarding.
+
+```
+ssh remote
+terminalboard path/to/tb_logs
+```
+
+---
+
+## Why this exists
+
+The usual TensorBoard workflow over SSH is painful: you either forward a port
+(`ssh -L 6006:...`) and open a browser, or you give up and `grep` the logs. On a
+headless training box you often can't do either cleanly. terminalboard reads the
+event files directly and draws the curves in the terminal, so a plain SSH session
+is all you need.
+
+## How it works
+
+1. **Read** the TensorBoard event files (`events.out.tfevents.*`) from a log
+   directory (scanned recursively for multiple runs) and collect the scalar series.
+2. **Render** the selected curves — by default as Unicode/braille **text** (works
+   in any terminal), or with `--hq` as a matplotlib PNG.
+3. **Display** them: braille text is printed directly; the `--hq` PNG is streamed
+   via the [iTerm2 inline-image protocol](https://iterm2.com/documentation-images.html)
+   (built in-memory, no temp file).
+4. **Watch** the log directory and re-render whenever new data lands, giving a
+   live dashboard. Repaints are **flicker-free**: the alternate screen buffer is
+   redrawn in place under synchronized output (DEC mode 2026), and an idle
+   dashboard isn't repainted at all (only changed data/views trigger a redraw).
+
+## Language: Python
+
+The viewer is written in **Python**, chosen after weighing it against a
+Next.js/TypeScript implementation:
+
+| Factor | Python ✅ | Next.js / TypeScript |
+|---|---|---|
+| Reading TB event logs | First-class. The format is TFRecord-framed protobuf; `tensorboard`/`tbparse` parse it natively, or a small self-contained parser does. | No mature TFRecord/TB-protobuf reader — you'd reimplement framing + protobuf decoding by hand. |
+| High-quality curves | matplotlib → PNG → iTerm2 inline image. | No native terminal-plotting story. |
+| Live tailing | `watchdog` / offset polling. | Doable, no advantage. |
+| Fit for purpose | It's a terminal CLI, and Python is the lingua franca of the ML/TensorBoard ecosystem. | Next.js is a web/SSR framework; its core value (React, routing, browser) is unused here. |
+| This machine | Python 3.12 already present. | Node isn't installed. |
+
+The decisive factor: TensorBoard logs are a TF-specific protobuf format with
+first-class Python tooling, and the target terminal (iTerm2) supports an
+inline-image protocol — so we can render genuine matplotlib-quality curves rather
+than ASCII art.
+
+## Two rendering backends
+
+- **Default — text/braille** (`plotext`): curves drawn directly as Unicode/braille
+  characters. No image is generated, so it works over **any** SSH session, tmux, or
+  plain terminal, and redraws instantly. *(See the example below.)*
+- **`--hq` — iTerm2 image**: matplotlib rendered to an **in-memory** PNG (no temp
+  file) and streamed via the iTerm2 inline-image protocol. Pixel-perfect, but only
+  in iTerm2/WezTerm-class terminals.
+- **`--auto`**: use the image renderer in iTerm2-class terminals, else fall back to
+  text automatically.
+
+## Two parsing backends
+
+- **Default** (no flag): parse with the official `tensorboard` library
+  (`EventAccumulator`) — most robust, handles exotic summary encodings. If
+  `tensorboard` isn't installed, terminalboard falls back to `--light` automatically.
+- **`--light`**: a self-contained pure-Python TFRecord + protobuf-wire parser with
+  no heavy dependencies — tiny install, fast startup, ideal for a thin remote box.
+
+The two axes are independent: pick any parser with any renderer.
+
+## Install
+
+> ⚠️ **Not published to PyPI yet**, so `pip install terminalboard` won't work
+> until it is. Install from source for now.
+
+**From a clone** (recommended for development):
+
+```bash
+git clone https://github.com/dongfangyixi/terminalboard.git
+cd terminalboard
+pip install -e '.[full]'     # editable; full = tensorboard + matplotlib
+```
+
+**Directly from GitHub:**
+
+```bash
+pip install "git+https://github.com/dongfangyixi/terminalboard.git"
+# with extras (default parser + --hq image renderer):
+pip install "terminalboard[full] @ git+https://github.com/dongfangyixi/terminalboard.git"
+```
+
+The base install pulls only `plotext` — enough for the text renderer and the
+dependency-free `--light` parser. Extras add the heavy bits:
+
+| Extra | Adds | Enables |
+|---|---|---|
+| `[tb]` | `tensorboard` | the default (robust) parser |
+| `[hq]` | `matplotlib` | the `--hq` iTerm2 image renderer |
+| `[full]` | both | everything |
+
+<details>
+<summary>Once it's on PyPI</summary>
+
+```bash
+pip install terminalboard            # text renderer + pure-Python --light parser
+pip install 'terminalboard[tb]'      # + tensorboard (default parser)
+pip install 'terminalboard[hq]'      # + matplotlib (--hq image renderer)
+pip install 'terminalboard[full]'    # everything
+```
+
+Publishing checklist: `pip install build twine` → `python -m build` →
+`twine upload dist/*` (needs a PyPI account + the name `terminalboard` being free).
+</details>
+
+## Usage
+
+```
+terminalboard LOGDIR [options]
+
+  LOGDIR / --logdir   directory of TensorBoard event files (scanned recursively)
+  --light             use the dependency-free pure-Python parser
+  --hq / --text/--auto   image / text (default) / auto-detect renderer
+  --tags GLOB         filter tags, e.g. 'train/*loss*,val/*' (live-editable: t)
+  --experiments GLOB  filter experiments/runs (live-editable: f)
+  --smooth ALPHA      EMA smoothing weight in [0,1) (default: 0.6; 0 disables)
+  --grid RxC          panels per page (default: 2x3)
+  --interval SECONDS  live refresh interval (default: 2.0)
+  --once              render a single frame and exit
+  --list              list all scalar tags and exit
+```
+
+```bash
+terminalboard ../tb_logs                       # live text dashboard
+terminalboard ../tb_logs --tags 'train/*loss*' # filter to loss curves
+terminalboard ../tb_logs --hq --grid 2x2       # high-quality iTerm2 images
+terminalboard ../tb_logs --light --once        # one frame, no deps, no loop
+```
+
+### Interactive controls (live mode)
+
+| Key | Action |
+|---|---|
+| `q` | quit |
+| `n` / `space`, `p` | next / previous page of tags |
+| `t` | edit the **tag** filter live (Enter apply · Esc cancel · ^U clear) |
+| `f` | edit the **experiment/run** filter live |
+| `r` | refresh now |
+| `+` / `-` | more / less smoothing |
+| `0` | disable smoothing |
+| `z` / `Z` | zoom out / in — panels per page: `1·2·4·6·9·12·16·24·36` |
+
+Filters match per comma-separated token: a plain word is a case-insensitive
+**substring** (`loss` → `train/loss`, `val/loss`), while `* ? [` make it a glob
+(`train/*loss*`). The plots re-filter as you type. Tag and experiment filters
+combine — a tag only shows if a currently-visible experiment has it.
+
+### Multiple experiments
+
+When a logdir holds several runs, their curves are **overlaid in each panel**,
+each experiment in its own color, with a legend above the grid. Colors are
+**stable** — an experiment keeps its color no matter which others you filter in
+or out — so you can always tell which curve is which. Use `f` (or
+`--experiments`) to focus on a subset.
+
+In the filter prompt: **←/→** move the cursor, **↑/↓** recall previous patterns,
+**Home/End** (or `^A`/`^E`) jump, `^U` clears. If a pattern matches nothing the
+current plots are **kept** (no jarring re-layout) and a red warning is shown
+until you fix or cancel it.
+
+### Example (text renderer)
+
+```
+                              train/text_token_accuracy
+    ┌──────────────────────────────────────────────────────────────────────────┐
+0.97┤                                                   ⡠⣄⣀⣀⡠⠖⠦⠤⠤⠖⠒⠒⠒⠒⠉⠙⠒⠒⠉⠉⠉⠉⠉│
+    │                                              ⣠⠒⠒⠒⠞                       │
+    │                                          ⡤⠲⠴⠤⠇                           │
+0.82┤                                         ⢰⠁                               │
+    │                                     ⢠⠒⠲⠤⠎                                │
+0.67┤                                 ⣀⣀⣀⣠⠃                                    │
+    │                           ⢀⠔⠒⠒⠲⠇                                         │
+0.52┤          ⣀⣀⣀⣀⣀⣀⣀⣀⡠⠤⠤⠤⠤⠞⠉⠉⠉⠛                                              │
+    │  ⡴⠲⠒⠉⠉⠉⠉⠉⠁                                                               │
+    └┬─────────────────┬──────────────────┬─────────────────┬─────────────────┬┘
+    10               1510               3010              4510             6010
+```
+
+## Roadmap
+
+- [x] **Reader — `--light`**: pure-Python TFRecord + protobuf-wire parser
+      (Event → Summary → Value; both `simple_value` and tensor-encoded scalars).
+- [x] **Reader — default**: `tensorboard` `EventAccumulator` backend with a shared
+      `ScalarSeries` data model and recursive multi-run logdir scan.
+- [x] **Render — text**: `plotext` braille grid, the default (no image).
+- [x] **Render — `--hq`**: matplotlib grid → in-memory PNG → iTerm2 inline image.
+- [x] **Live loop + CLI**: flicker-free repaints, keyboard navigation; argparse front end.
+- [x] **Zoom** (`z`/`Z`): 1·2·4·6·9·12·16·24·36 panels per page.
+- [x] **Interactive filters** (`t`/`f`): live tag & experiment filtering with a
+      line editor (cursor, history, no-match warning).
+- [x] **Multi-experiment overlay** with stable per-run colors and a legend.
+- [ ] Publish to PyPI.
+- [ ] Sixel fallback for non-iTerm2 terminals; config file; per-tag y-axis options.
+
+## Status
+
+Working. The default text dashboard, `--hq` iTerm2 images, the `--light` parser,
+multi-experiment overlay, zoom, and live interactive tag/experiment filtering are
+all functional. Test event logs are kept in the **parent working folder** (e.g.
+`../tb_logs/`), deliberately outside this repository — they're real training data
+and don't belong in a public repo.
+
+## Development
+
+```bash
+python3 -m venv .venv
+.venv/bin/pip install -e '.[full]'
+.venv/bin/terminalboard ../tb_logs --once
+```
+
+## License
+
+[MIT](LICENSE).

terminalboard-0.1.0/README.md

@@ -0,0 +1,228 @@
+# terminalboard
+
+A **pure-SSH terminal TensorBoard scalar viewer**.
+
+Train on a remote box, SSH in, and watch your **live-updating scalar curves
+right inside the terminal** — crisp Unicode/braille text by default, or real
+iTerm2 inline images with `--hq`. No browser, no X11, no port forwarding.
+
+```
+ssh remote
+terminalboard path/to/tb_logs
+```
+
+---
+
+## Why this exists
+
+The usual TensorBoard workflow over SSH is painful: you either forward a port
+(`ssh -L 6006:...`) and open a browser, or you give up and `grep` the logs. On a
+headless training box you often can't do either cleanly. terminalboard reads the
+event files directly and draws the curves in the terminal, so a plain SSH session
+is all you need.
+
+## How it works
+
+1. **Read** the TensorBoard event files (`events.out.tfevents.*`) from a log
+   directory (scanned recursively for multiple runs) and collect the scalar series.
+2. **Render** the selected curves — by default as Unicode/braille **text** (works
+   in any terminal), or with `--hq` as a matplotlib PNG.
+3. **Display** them: braille text is printed directly; the `--hq` PNG is streamed
+   via the [iTerm2 inline-image protocol](https://iterm2.com/documentation-images.html)
+   (built in-memory, no temp file).
+4. **Watch** the log directory and re-render whenever new data lands, giving a
+   live dashboard. Repaints are **flicker-free**: the alternate screen buffer is
+   redrawn in place under synchronized output (DEC mode 2026), and an idle
+   dashboard isn't repainted at all (only changed data/views trigger a redraw).
+
+## Language: Python
+
+The viewer is written in **Python**, chosen after weighing it against a
+Next.js/TypeScript implementation:
+
+| Factor | Python ✅ | Next.js / TypeScript |
+|---|---|---|
+| Reading TB event logs | First-class. The format is TFRecord-framed protobuf; `tensorboard`/`tbparse` parse it natively, or a small self-contained parser does. | No mature TFRecord/TB-protobuf reader — you'd reimplement framing + protobuf decoding by hand. |
+| High-quality curves | matplotlib → PNG → iTerm2 inline image. | No native terminal-plotting story. |
+| Live tailing | `watchdog` / offset polling. | Doable, no advantage. |
+| Fit for purpose | It's a terminal CLI, and Python is the lingua franca of the ML/TensorBoard ecosystem. | Next.js is a web/SSR framework; its core value (React, routing, browser) is unused here. |
+| This machine | Python 3.12 already present. | Node isn't installed. |
+
+The decisive factor: TensorBoard logs are a TF-specific protobuf format with
+first-class Python tooling, and the target terminal (iTerm2) supports an
+inline-image protocol — so we can render genuine matplotlib-quality curves rather
+than ASCII art.
+
+## Two rendering backends
+
+- **Default — text/braille** (`plotext`): curves drawn directly as Unicode/braille
+  characters. No image is generated, so it works over **any** SSH session, tmux, or
+  plain terminal, and redraws instantly. *(See the example below.)*
+- **`--hq` — iTerm2 image**: matplotlib rendered to an **in-memory** PNG (no temp
+  file) and streamed via the iTerm2 inline-image protocol. Pixel-perfect, but only
+  in iTerm2/WezTerm-class terminals.
+- **`--auto`**: use the image renderer in iTerm2-class terminals, else fall back to
+  text automatically.
+
+## Two parsing backends
+
+- **Default** (no flag): parse with the official `tensorboard` library
+  (`EventAccumulator`) — most robust, handles exotic summary encodings. If
+  `tensorboard` isn't installed, terminalboard falls back to `--light` automatically.
+- **`--light`**: a self-contained pure-Python TFRecord + protobuf-wire parser with
+  no heavy dependencies — tiny install, fast startup, ideal for a thin remote box.
+
+The two axes are independent: pick any parser with any renderer.
+
+## Install
+
+> ⚠️ **Not published to PyPI yet**, so `pip install terminalboard` won't work
+> until it is. Install from source for now.
+
+**From a clone** (recommended for development):
+
+```bash
+git clone https://github.com/dongfangyixi/terminalboard.git
+cd terminalboard
+pip install -e '.[full]'     # editable; full = tensorboard + matplotlib
+```
+
+**Directly from GitHub:**
+
+```bash
+pip install "git+https://github.com/dongfangyixi/terminalboard.git"
+# with extras (default parser + --hq image renderer):
+pip install "terminalboard[full] @ git+https://github.com/dongfangyixi/terminalboard.git"
+```
+
+The base install pulls only `plotext` — enough for the text renderer and the
+dependency-free `--light` parser. Extras add the heavy bits:
+
+| Extra | Adds | Enables |
+|---|---|---|
+| `[tb]` | `tensorboard` | the default (robust) parser |
+| `[hq]` | `matplotlib` | the `--hq` iTerm2 image renderer |
+| `[full]` | both | everything |
+
+<details>
+<summary>Once it's on PyPI</summary>
+
+```bash
+pip install terminalboard            # text renderer + pure-Python --light parser
+pip install 'terminalboard[tb]'      # + tensorboard (default parser)
+pip install 'terminalboard[hq]'      # + matplotlib (--hq image renderer)
+pip install 'terminalboard[full]'    # everything
+```
+
+Publishing checklist: `pip install build twine` → `python -m build` →
+`twine upload dist/*` (needs a PyPI account + the name `terminalboard` being free).
+</details>
+
+## Usage
+
+```
+terminalboard LOGDIR [options]
+
+  LOGDIR / --logdir   directory of TensorBoard event files (scanned recursively)
+  --light             use the dependency-free pure-Python parser
+  --hq / --text/--auto   image / text (default) / auto-detect renderer
+  --tags GLOB         filter tags, e.g. 'train/*loss*,val/*' (live-editable: t)
+  --experiments GLOB  filter experiments/runs (live-editable: f)
+  --smooth ALPHA      EMA smoothing weight in [0,1) (default: 0.6; 0 disables)
+  --grid RxC          panels per page (default: 2x3)
+  --interval SECONDS  live refresh interval (default: 2.0)
+  --once              render a single frame and exit
+  --list              list all scalar tags and exit
+```
+
+```bash
+terminalboard ../tb_logs                       # live text dashboard
+terminalboard ../tb_logs --tags 'train/*loss*' # filter to loss curves
+terminalboard ../tb_logs --hq --grid 2x2       # high-quality iTerm2 images
+terminalboard ../tb_logs --light --once        # one frame, no deps, no loop
+```
+
+### Interactive controls (live mode)
+
+| Key | Action |
+|---|---|
+| `q` | quit |
+| `n` / `space`, `p` | next / previous page of tags |
+| `t` | edit the **tag** filter live (Enter apply · Esc cancel · ^U clear) |
+| `f` | edit the **experiment/run** filter live |
+| `r` | refresh now |
+| `+` / `-` | more / less smoothing |
+| `0` | disable smoothing |
+| `z` / `Z` | zoom out / in — panels per page: `1·2·4·6·9·12·16·24·36` |
+
+Filters match per comma-separated token: a plain word is a case-insensitive
+**substring** (`loss` → `train/loss`, `val/loss`), while `* ? [` make it a glob
+(`train/*loss*`). The plots re-filter as you type. Tag and experiment filters
+combine — a tag only shows if a currently-visible experiment has it.
+
+### Multiple experiments
+
+When a logdir holds several runs, their curves are **overlaid in each panel**,
+each experiment in its own color, with a legend above the grid. Colors are
+**stable** — an experiment keeps its color no matter which others you filter in
+or out — so you can always tell which curve is which. Use `f` (or
+`--experiments`) to focus on a subset.
+
+In the filter prompt: **←/→** move the cursor, **↑/↓** recall previous patterns,
+**Home/End** (or `^A`/`^E`) jump, `^U` clears. If a pattern matches nothing the
+current plots are **kept** (no jarring re-layout) and a red warning is shown
+until you fix or cancel it.
+
+### Example (text renderer)
+
+```
+                              train/text_token_accuracy
+    ┌──────────────────────────────────────────────────────────────────────────┐
+0.97┤                                                   ⡠⣄⣀⣀⡠⠖⠦⠤⠤⠖⠒⠒⠒⠒⠉⠙⠒⠒⠉⠉⠉⠉⠉│
+    │                                              ⣠⠒⠒⠒⠞                       │
+    │                                          ⡤⠲⠴⠤⠇                           │
+0.82┤                                         ⢰⠁                               │
+    │                                     ⢠⠒⠲⠤⠎                                │
+0.67┤                                 ⣀⣀⣀⣠⠃                                    │
+    │                           ⢀⠔⠒⠒⠲⠇                                         │
+0.52┤          ⣀⣀⣀⣀⣀⣀⣀⣀⡠⠤⠤⠤⠤⠞⠉⠉⠉⠛                                              │
+    │  ⡴⠲⠒⠉⠉⠉⠉⠉⠁                                                               │
+    └┬─────────────────┬──────────────────┬─────────────────┬─────────────────┬┘
+    10               1510               3010              4510             6010
+```
+
+## Roadmap
+
+- [x] **Reader — `--light`**: pure-Python TFRecord + protobuf-wire parser
+      (Event → Summary → Value; both `simple_value` and tensor-encoded scalars).
+- [x] **Reader — default**: `tensorboard` `EventAccumulator` backend with a shared
+      `ScalarSeries` data model and recursive multi-run logdir scan.
+- [x] **Render — text**: `plotext` braille grid, the default (no image).
+- [x] **Render — `--hq`**: matplotlib grid → in-memory PNG → iTerm2 inline image.
+- [x] **Live loop + CLI**: flicker-free repaints, keyboard navigation; argparse front end.
+- [x] **Zoom** (`z`/`Z`): 1·2·4·6·9·12·16·24·36 panels per page.
+- [x] **Interactive filters** (`t`/`f`): live tag & experiment filtering with a
+      line editor (cursor, history, no-match warning).
+- [x] **Multi-experiment overlay** with stable per-run colors and a legend.
+- [ ] Publish to PyPI.
+- [ ] Sixel fallback for non-iTerm2 terminals; config file; per-tag y-axis options.
+
+## Status
+
+Working. The default text dashboard, `--hq` iTerm2 images, the `--light` parser,
+multi-experiment overlay, zoom, and live interactive tag/experiment filtering are
+all functional. Test event logs are kept in the **parent working folder** (e.g.
+`../tb_logs/`), deliberately outside this repository — they're real training data
+and don't belong in a public repo.
+
+## Development
+
+```bash
+python3 -m venv .venv
+.venv/bin/pip install -e '.[full]'
+.venv/bin/terminalboard ../tb_logs --once
+```
+
+## License
+
+[MIT](LICENSE).

terminalboard-0.1.0/pyproject.toml

@@ -0,0 +1,37 @@
+[build-system]
+requires = ["setuptools>=61"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "terminalboard"
+version = "0.1.0"
+description = "A pure-SSH terminal TensorBoard scalar viewer — live scalar curves in iTerm2, no browser, no X11, no port forwarding."
+readme = "README.md"
+requires-python = ">=3.9"
+license = { text = "MIT" }
+authors = [{ name = "dongfangyixi" }]
+keywords = ["tensorboard", "terminal", "iterm2", "plotext", "ssh", "ml"]
+
+# Base install is intentionally light: the text renderer (plotext) and the
+# pure-Python --light parser work with just this.
+dependencies = [
+    "plotext>=5.2",
+]
+
+[project.optional-dependencies]
+# Default parser backend (EventAccumulator).
+tb = ["tensorboard>=2.9"]
+# High-quality --hq image renderer (matplotlib -> iTerm2 inline image).
+hq = ["matplotlib>=3.5"]
+# Everything.
+full = ["tensorboard>=2.9", "matplotlib>=3.5"]
+
+[project.scripts]
+terminalboard = "terminalboard.cli:main"
+
+[project.urls]
+Repository = "https://github.com/dongfangyixi/terminalboard"
+
+[tool.setuptools.packages.find]
+where = ["."]
+include = ["terminalboard*"]

terminalboard-0.1.0/setup.cfg

@@ -0,0 +1,4 @@
+[egg_info]
+tag_build = 
+tag_date = 0
+

terminalboard-0.1.0/terminalboard/__init__.py

@@ -0,0 +1,9 @@
+"""terminalboard — a pure-SSH terminal TensorBoard scalar viewer.
+
+Reads TensorBoard event logs and renders live-updated scalar curves directly
+inside the terminal — by default as Unicode/braille text (works over any SSH
+session), or as high-quality matplotlib images via the iTerm2 inline-image
+protocol with ``--hq``. No browser, no X11, no port forwarding.
+"""
+
+__version__ = "0.1.0"

terminalboard-0.1.0/terminalboard/__main__.py

@@ -0,0 +1,4 @@
+from .cli import main
+
+if __name__ == "__main__":
+    raise SystemExit(main())