recursionlens 0.1.2__tar.gz → 0.2.0__tar.gz

{recursionlens-0.1.2 → recursionlens-0.2.0}/LICENSE

@@ -1,6 +1,6 @@
 MIT License
 
-Copyright (c) 2026 Ziyu Jiang and contributors
+Copyright (c) 2026 Ziyu JIANG and contributors
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal

{recursionlens-0.1.2/src/recursionlens.egg-info → recursionlens-0.2.0}/PKG-INFO

@@ -1,16 +1,16 @@
 Metadata-Version: 2.4
 Name: recursionlens
-Version: 0.1.2
-Summary: Interactive recursion tracing, subproblem DAGs, and memoization comparisons for JupyterLab.
-Author: Ziyu Jiang, Xinyue Cai, Kehan Cao, Ruoxi Lyu, Haobo Ma
+Version: 0.2.0
+Summary: Interactive recursion tracing, subproblem DAGs, memoization comparisons, and a Python Tutor-style step-through tutor for JupyterLab.
+Author: Ziyu JIANG, Xinyue CAI, Kehan CAO, Ruoxi LYU, Haobo MA
 License-Expression: MIT
 Project-URL: Homepage, https://github.com/ZiyuJiang07/recursionlens
+Project-URL: Website, https://ziyujiang07.github.io/recursionlens/
 Project-URL: Repository, https://github.com/ZiyuJiang07/recursionlens
-Project-URL: Documentation, https://github.com/ZiyuJiang07/recursionlens/blob/main/docs/USER.md
+Project-URL: Documentation, https://github.com/ZiyuJiang07/recursionlens#readme
 Project-URL: Tutorial, https://github.com/ZiyuJiang07/recursionlens/blob/main/docs/RecursionLens_Tutorial.ipynb
-Project-URL: Changelog, https://github.com/ZiyuJiang07/recursionlens/blob/main/CHANGELOG.md
 Project-URL: Issues, https://github.com/ZiyuJiang07/recursionlens/issues
-Keywords: recursion,jupyterlab,visualization,memoization,education
+Keywords: recursion,jupyterlab,visualization,memoization,education,debugging,python-tutor
 Classifier: Development Status :: 3 - Alpha
 Classifier: Intended Audience :: Education
 Classifier: Programming Language :: Python :: 3
@@ -22,21 +22,28 @@ License-File: LICENSE
 Provides-Extra: dev
 Requires-Dist: pytest>=8; extra == "dev"
 Requires-Dist: coverage>=7; extra == "dev"
+Requires-Dist: ipywidgets>=7.6; extra == "dev"
+Provides-Extra: tutor
+Requires-Dist: ipywidgets>=7.6; extra == "tutor"
 Dynamic: license-file
 
 # RecursionLens
 
+**English** | [简体中文](https://github.com/ZiyuJiang07/recursionlens/blob/main/README.zh-CN.md)
+
 **See not just what recursion returns, but where the repeated work happens.**
 
+**Try it in the browser:** https://ziyujiang07.github.io/recursionlens/ — every view and the line-by-line player run client-side with Pyodide, no installation. The examples menu collects every recursive function found in the CS1302A lectures.
+
 RecursionLens is a zero-runtime-dependency Python package for exploring recursive execution in JupyterLab. It turns a traced run into an expandable call tree, a compressed subproblem DAG, and a quantitative naive-versus-memoized comparison.
 
-> **JupyterLab tutorial, already executed:** [Open RecursionLens_Tutorial.ipynb](https://github.com/ZiyuJiang07/recursionlens/blob/main/docs/RecursionLens_Tutorial.ipynb). All eight code cells have saved output. If Jupyter hides HTML/SVG from a downloaded notebook, use its **Trust** action or click **Run All Cells** once. For a zero-setup preview, download and open the [static HTML companion](https://github.com/ZiyuJiang07/recursionlens/raw/main/docs/RecursionLens_Tutorial.html).
+> **JupyterLab tutorial, already executed:** [Open RecursionLens_Tutorial.ipynb](https://github.com/ZiyuJiang07/recursionlens/blob/main/docs/RecursionLens_Tutorial.ipynb) ([简体中文版](https://github.com/ZiyuJiang07/recursionlens/blob/main/docs/RecursionLens_Tutorial.zh-CN.ipynb)). Sixteen code cells have saved output; the final cell opens the interactive player when you run it in a live kernel. If Jupyter hides HTML/SVG from a downloaded notebook, use its **Trust** action or click **Run All Cells** once. For a zero-setup preview, download and open the [static HTML companion](https://github.com/ZiyuJiang07/recursionlens/raw/main/docs/RecursionLens_Tutorial.html) or its [Chinese version](https://github.com/ZiyuJiang07/recursionlens/raw/main/docs/RecursionLens_Tutorial.zh-CN.html).
 
 ## Project team
 
 CS1302A Group 50:
 
-- Ziyu Jiang
+- Ziyu JIANG
 - Xinyue CAI
 - Kehan CAO
 - Ruoxi LYU
@@ -44,7 +51,9 @@ CS1302A Group 50:
 
 ## Why RecursionLens?
 
-A recursive function normally exposes only its final return value or traceback. Students still have to reconstruct the execution in their heads. RecursionLens answers three different questions:
+RecursionLens grew out of a beginner's own difficulty with recursion. Loops can be read line by line, but a non-tail-recursive call such as `fib` or a tree traversal expands into many interleaved calls: the same subproblem is entered again and again, and a value only comes back after its whole subtree has finished. The step-by-step mentor provided by our course (OPT) is excellent for ordinary programs, since it shows one execution step at a time, yet a recursive run becomes a long stream of push and pop steps with no view of the overall shape of the computation, and no indication of which subproblems repeat. The usual fallback is to draw the call tree by hand on paper. RecursionLens automates that drawing and adds the numbers a hand-drawn tree cannot show.
+
+It answers three different questions:
 
 1. **Call tree:** what actually happened, invocation by invocation?
 2. **Subproblem DAG:** how many distinct problems and dependencies exist after identical arguments are merged?
@@ -81,7 +90,7 @@ cd recursionlens
 python -m pip install -e ".[dev]"
 ```
 
-The package has no runtime dependencies.
+The package has no runtime dependencies. The interactive tutor is the only optional extra, and it needs `ipywidgets`.
 
 ## Quickstart in JupyterLab
 
@@ -127,6 +136,36 @@ assert fib(5) == 5
 assert fib.last_report.total_calls == 15
 ```
 
+## Step through a function line by line
+
+Version 0.2 adds a Python Tutor-style player that runs inside the notebook. Paste a function, choose the call, and step through it:
+
+```python
+from recursionlens import tutor
+
+tutor()   # editable code panel, an input box for the call, and step controls
+```
+
+The player highlights the current line and grows the call tree as the run proceeds: waiting calls are blue, the call being executed is coral, and returned calls stay on screen in green. The current frame's local variables sit under the tree, and changing the call (for example `fib(5)` → `fib(3)`) records a fresh run and redraws from a single node. It reports each call, line, return, and exception, which makes the order of a recursion easy to follow.
+
+The tutor runs the code you paste with normal Python semantics; it is not a sandbox.
+
+`tutor()` needs `ipywidgets`:
+
+```bash
+python -m pip install "recursionlens[tutor]"
+```
+
+The recording is plain Python, so it also works without the widget:
+
+```python
+from recursionlens import record_tutor
+
+trace = record_tutor("def fib(n): ...", "fib(5)")
+print(trace.describe())
+trace.steps[0].frames        # stack snapshot for one step
+```
+
 ## Export a report
 
 ```python
@@ -150,12 +189,11 @@ Version 0.1 targets synchronous, deterministic teaching examples such as factori
 
 ## Documentation and demo
 
-- [Executed JupyterLab tutorial](https://github.com/ZiyuJiang07/recursionlens/blob/main/docs/RecursionLens_Tutorial.ipynb) - setup, three views, comparison, export, troubleshooting, and a one-minute demo recipe
-- [Static HTML tutorial preview](https://github.com/ZiyuJiang07/recursionlens/raw/main/docs/RecursionLens_Tutorial.html) - the same tutorial with all visual reports rendered, no Jupyter required
+- [Command guide (English)](https://github.com/ZiyuJiang07/recursionlens/blob/main/docs/RecursionLens_Tutorial.ipynb) - one example per public command: decorator, report object, three views, metrics, memoization, comparison, export, and error handling
+- [命令手册（简体中文）](https://github.com/ZiyuJiang07/recursionlens/blob/main/docs/RecursionLens_Tutorial.zh-CN.ipynb) - the same guide in Simplified Chinese
+- [Static HTML preview](https://github.com/ZiyuJiang07/recursionlens/raw/main/docs/RecursionLens_Tutorial.html) - English, no Jupyter required
+- [静态 HTML 预览](https://github.com/ZiyuJiang07/recursionlens/raw/main/docs/RecursionLens_Tutorial.zh-CN.html) - 中文，不需要 Jupyter
 - [Runnable example notebook](https://github.com/ZiyuJiang07/recursionlens/blob/main/examples/demo.ipynb)
-- [User guide](https://github.com/ZiyuJiang07/recursionlens/blob/main/docs/USER.md)
-- [Developer guide](https://github.com/ZiyuJiang07/recursionlens/blob/main/docs/DEVELOPER.md)
-- [Changelog](https://github.com/ZiyuJiang07/recursionlens/blob/main/CHANGELOG.md)
 
 ## Development and testing
 
@@ -165,8 +203,10 @@ python -m build
 python -m twine check dist/*
 ```
 
-The test suite covers trace invariants, cache behavior, signatures, exceptions, truncation, JSON, HTML escaping, DAG rendering, and comparison reports. Contributions should keep the public API, tests, and user documentation synchronized.
+The test suite covers trace invariants, cache behavior, signatures, exceptions, truncation, JSON, HTML escaping, DAG rendering, and comparison reports. Contributions should keep the public API, tests, and README synchronized.
+
+Documentation is bilingual: `README.md` / `README.zh-CN.md` and `docs/RecursionLens_Tutorial.ipynb` / `docs/RecursionLens_Tutorial.zh-CN.ipynb` must be updated together, including their generated HTML files.
 
 ## License
 
-MIT © 2026 Ziyu Jiang, Xinyue Cai, Kehan Cao, Ruoxi Lyu, Haobo Ma.
+MIT © 2026 Ziyu JIANG, Xinyue CAI, Kehan CAO, Ruoxi LYU, Haobo MA.

{recursionlens-0.1.2 → recursionlens-0.2.0}/README.md

@@ -1,16 +1,20 @@
 # RecursionLens
 
+**English** | [简体中文](https://github.com/ZiyuJiang07/recursionlens/blob/main/README.zh-CN.md)
+
 **See not just what recursion returns, but where the repeated work happens.**
 
+**Try it in the browser:** https://ziyujiang07.github.io/recursionlens/ — every view and the line-by-line player run client-side with Pyodide, no installation. The examples menu collects every recursive function found in the CS1302A lectures.
+
 RecursionLens is a zero-runtime-dependency Python package for exploring recursive execution in JupyterLab. It turns a traced run into an expandable call tree, a compressed subproblem DAG, and a quantitative naive-versus-memoized comparison.
 
-> **JupyterLab tutorial, already executed:** [Open RecursionLens_Tutorial.ipynb](https://github.com/ZiyuJiang07/recursionlens/blob/main/docs/RecursionLens_Tutorial.ipynb). All eight code cells have saved output. If Jupyter hides HTML/SVG from a downloaded notebook, use its **Trust** action or click **Run All Cells** once. For a zero-setup preview, download and open the [static HTML companion](https://github.com/ZiyuJiang07/recursionlens/raw/main/docs/RecursionLens_Tutorial.html).
+> **JupyterLab tutorial, already executed:** [Open RecursionLens_Tutorial.ipynb](https://github.com/ZiyuJiang07/recursionlens/blob/main/docs/RecursionLens_Tutorial.ipynb) ([简体中文版](https://github.com/ZiyuJiang07/recursionlens/blob/main/docs/RecursionLens_Tutorial.zh-CN.ipynb)). Sixteen code cells have saved output; the final cell opens the interactive player when you run it in a live kernel. If Jupyter hides HTML/SVG from a downloaded notebook, use its **Trust** action or click **Run All Cells** once. For a zero-setup preview, download and open the [static HTML companion](https://github.com/ZiyuJiang07/recursionlens/raw/main/docs/RecursionLens_Tutorial.html) or its [Chinese version](https://github.com/ZiyuJiang07/recursionlens/raw/main/docs/RecursionLens_Tutorial.zh-CN.html).
 
 ## Project team
 
 CS1302A Group 50:
 
-- Ziyu Jiang
+- Ziyu JIANG
 - Xinyue CAI
 - Kehan CAO
 - Ruoxi LYU
@@ -18,7 +22,9 @@ CS1302A Group 50:
 
 ## Why RecursionLens?
 
-A recursive function normally exposes only its final return value or traceback. Students still have to reconstruct the execution in their heads. RecursionLens answers three different questions:
+RecursionLens grew out of a beginner's own difficulty with recursion. Loops can be read line by line, but a non-tail-recursive call such as `fib` or a tree traversal expands into many interleaved calls: the same subproblem is entered again and again, and a value only comes back after its whole subtree has finished. The step-by-step mentor provided by our course (OPT) is excellent for ordinary programs, since it shows one execution step at a time, yet a recursive run becomes a long stream of push and pop steps with no view of the overall shape of the computation, and no indication of which subproblems repeat. The usual fallback is to draw the call tree by hand on paper. RecursionLens automates that drawing and adds the numbers a hand-drawn tree cannot show.
+
+It answers three different questions:
 
 1. **Call tree:** what actually happened, invocation by invocation?
 2. **Subproblem DAG:** how many distinct problems and dependencies exist after identical arguments are merged?
@@ -55,7 +61,7 @@ cd recursionlens
 python -m pip install -e ".[dev]"
 ```
 
-The package has no runtime dependencies.
+The package has no runtime dependencies. The interactive tutor is the only optional extra, and it needs `ipywidgets`.
 
 ## Quickstart in JupyterLab
 
@@ -101,6 +107,36 @@ assert fib(5) == 5
 assert fib.last_report.total_calls == 15
 ```
 
+## Step through a function line by line
+
+Version 0.2 adds a Python Tutor-style player that runs inside the notebook. Paste a function, choose the call, and step through it:
+
+```python
+from recursionlens import tutor
+
+tutor()   # editable code panel, an input box for the call, and step controls
+```
+
+The player highlights the current line and grows the call tree as the run proceeds: waiting calls are blue, the call being executed is coral, and returned calls stay on screen in green. The current frame's local variables sit under the tree, and changing the call (for example `fib(5)` → `fib(3)`) records a fresh run and redraws from a single node. It reports each call, line, return, and exception, which makes the order of a recursion easy to follow.
+
+The tutor runs the code you paste with normal Python semantics; it is not a sandbox.
+
+`tutor()` needs `ipywidgets`:
+
+```bash
+python -m pip install "recursionlens[tutor]"
+```
+
+The recording is plain Python, so it also works without the widget:
+
+```python
+from recursionlens import record_tutor
+
+trace = record_tutor("def fib(n): ...", "fib(5)")
+print(trace.describe())
+trace.steps[0].frames        # stack snapshot for one step
+```
+
 ## Export a report
 
 ```python
@@ -124,12 +160,11 @@ Version 0.1 targets synchronous, deterministic teaching examples such as factori
 
 ## Documentation and demo
 
-- [Executed JupyterLab tutorial](https://github.com/ZiyuJiang07/recursionlens/blob/main/docs/RecursionLens_Tutorial.ipynb) - setup, three views, comparison, export, troubleshooting, and a one-minute demo recipe
-- [Static HTML tutorial preview](https://github.com/ZiyuJiang07/recursionlens/raw/main/docs/RecursionLens_Tutorial.html) - the same tutorial with all visual reports rendered, no Jupyter required
+- [Command guide (English)](https://github.com/ZiyuJiang07/recursionlens/blob/main/docs/RecursionLens_Tutorial.ipynb) - one example per public command: decorator, report object, three views, metrics, memoization, comparison, export, and error handling
+- [命令手册（简体中文）](https://github.com/ZiyuJiang07/recursionlens/blob/main/docs/RecursionLens_Tutorial.zh-CN.ipynb) - the same guide in Simplified Chinese
+- [Static HTML preview](https://github.com/ZiyuJiang07/recursionlens/raw/main/docs/RecursionLens_Tutorial.html) - English, no Jupyter required
+- [静态 HTML 预览](https://github.com/ZiyuJiang07/recursionlens/raw/main/docs/RecursionLens_Tutorial.zh-CN.html) - 中文，不需要 Jupyter
 - [Runnable example notebook](https://github.com/ZiyuJiang07/recursionlens/blob/main/examples/demo.ipynb)
-- [User guide](https://github.com/ZiyuJiang07/recursionlens/blob/main/docs/USER.md)
-- [Developer guide](https://github.com/ZiyuJiang07/recursionlens/blob/main/docs/DEVELOPER.md)
-- [Changelog](https://github.com/ZiyuJiang07/recursionlens/blob/main/CHANGELOG.md)
 
 ## Development and testing
 
@@ -139,8 +174,10 @@ python -m build
 python -m twine check dist/*
 ```
 
-The test suite covers trace invariants, cache behavior, signatures, exceptions, truncation, JSON, HTML escaping, DAG rendering, and comparison reports. Contributions should keep the public API, tests, and user documentation synchronized.
+The test suite covers trace invariants, cache behavior, signatures, exceptions, truncation, JSON, HTML escaping, DAG rendering, and comparison reports. Contributions should keep the public API, tests, and README synchronized.
+
+Documentation is bilingual: `README.md` / `README.zh-CN.md` and `docs/RecursionLens_Tutorial.ipynb` / `docs/RecursionLens_Tutorial.zh-CN.ipynb` must be updated together, including their generated HTML files.
 
 ## License
 
-MIT © 2026 Ziyu Jiang, Xinyue Cai, Kehan Cao, Ruoxi Lyu, Haobo Ma.
+MIT © 2026 Ziyu JIANG, Xinyue CAI, Kehan CAO, Ruoxi LYU, Haobo MA.

{recursionlens-0.1.2 → recursionlens-0.2.0}/pyproject.toml

@@ -4,20 +4,20 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "recursionlens"
-version = "0.1.2"
-description = "Interactive recursion tracing, subproblem DAGs, and memoization comparisons for JupyterLab."
+version = "0.2.0"
+description = "Interactive recursion tracing, subproblem DAGs, memoization comparisons, and a Python Tutor-style step-through tutor for JupyterLab."
 readme = "README.md"
 requires-python = ">=3.9"
 license = "MIT"
 license-files = ["LICENSE"]
 authors = [
-  { name = "Ziyu Jiang" },
-  { name = "Xinyue Cai" },
-  { name = "Kehan Cao" },
-  { name = "Ruoxi Lyu" },
-  { name = "Haobo Ma" },
+  { name = "Ziyu JIANG" },
+  { name = "Xinyue CAI" },
+  { name = "Kehan CAO" },
+  { name = "Ruoxi LYU" },
+  { name = "Haobo MA" },
 ]
-keywords = ["recursion", "jupyterlab", "visualization", "memoization", "education"]
+keywords = ["recursion", "jupyterlab", "visualization", "memoization", "education", "debugging", "python-tutor"]
 classifiers = [
   "Development Status :: 3 - Alpha",
   "Intended Audience :: Education",
@@ -28,14 +28,15 @@ classifiers = [
 
 [project.urls]
 Homepage = "https://github.com/ZiyuJiang07/recursionlens"
+Website = "https://ziyujiang07.github.io/recursionlens/"
 Repository = "https://github.com/ZiyuJiang07/recursionlens"
-Documentation = "https://github.com/ZiyuJiang07/recursionlens/blob/main/docs/USER.md"
+Documentation = "https://github.com/ZiyuJiang07/recursionlens#readme"
 Tutorial = "https://github.com/ZiyuJiang07/recursionlens/blob/main/docs/RecursionLens_Tutorial.ipynb"
-Changelog = "https://github.com/ZiyuJiang07/recursionlens/blob/main/CHANGELOG.md"
 Issues = "https://github.com/ZiyuJiang07/recursionlens/issues"
 
 [project.optional-dependencies]
-dev = ["pytest>=8", "coverage>=7"]
+dev = ["pytest>=8", "coverage>=7", "ipywidgets>=7.6"]
+tutor = ["ipywidgets>=7.6"]
 
 [tool.setuptools.packages.find]
 where = ["src"]

{recursionlens-0.1.2 → recursionlens-0.2.0}/src/recursionlens/__init__.py

@@ -3,14 +3,19 @@
 from ._model import CallNode, DAGNode
 from ._report import ComparisonReport, TraceReport
 from ._trace import trace
+from ._tutor import TutorStep, TutorTrace, record_tutor, tutor
 
-__version__ = "0.1.2"
+__version__ = "0.2.0"
 
 __all__ = [
     "CallNode",
     "ComparisonReport",
     "DAGNode",
     "TraceReport",
+    "TutorStep",
+    "TutorTrace",
+    "record_tutor",
     "trace",
+    "tutor",
     "__version__",
 ]

{recursionlens-0.1.2 → recursionlens-0.2.0}/src/recursionlens/_render.py

@@ -9,7 +9,7 @@ from ._model import DAGNode
 
 _CSS = """
 .rl-wrap{--rl-blue:#185fa5;--rl-blue-bg:#e6f1fb;--rl-coral:#d85a30;--rl-coral-bg:#faece7;--rl-green:#0f6e56;--rl-green-bg:#e1f5ee;--rl-text:#242424;--rl-muted:#66645f;--rl-border:#dedcd5;color:var(--rl-text);background:#fff;border:1px solid var(--rl-border);border-radius:12px;padding:18px;font-family:-apple-system,BlinkMacSystemFont,"Segoe UI",sans-serif;line-height:1.5;box-sizing:border-box;margin:8px 0;max-width:100%;overflow:auto}
-.rl-wrap *{box-sizing:border-box}.rl-title{font-size:18px;font-weight:600;margin:0 0 4px}.rl-subtitle{color:var(--rl-muted);font-size:13px;margin:0 0 16px}.rl-metrics{display:grid;grid-template-columns:repeat(auto-fit,minmax(105px,1fr));gap:8px;margin:12px 0 18px}.rl-metric{background:#f6f5f1;border-radius:8px;padding:10px}.rl-metric-label{font-size:11px;color:var(--rl-muted)}.rl-metric-value{font-size:21px;font-weight:600;margin-top:2px}.rl-section{font-size:14px;font-weight:600;margin:18px 0 8px}.rl-tree,.rl-tree ul{list-style:none;padding-left:20px}.rl-tree{padding-left:0}.rl-tree details{margin:4px 0}.rl-tree summary{cursor:pointer;padding:5px 8px;border-radius:6px;font-family:ui-monospace,SFMono-Regular,Menlo,monospace;font-size:12px}.rl-tree summary:hover{background:var(--rl-blue-bg)}.rl-result{color:var(--rl-muted);font-size:11px;margin-left:6px}.rl-badge{display:inline-block;border-radius:4px;padding:1px 5px;margin-left:5px;font-family:-apple-system,BlinkMacSystemFont,"Segoe UI",sans-serif;font-size:10px;font-weight:500}.rl-repeat{color:#993c1d;background:var(--rl-coral-bg)}.rl-cached{color:var(--rl-green);background:var(--rl-green-bg)}.rl-error{color:#a32d2d;background:#fcebeb}.rl-warning{color:#854f0b;background:#faeeda;border-radius:6px;padding:8px;margin:8px 0;font-size:12px}.rl-dag{max-width:100%;overflow:auto;background:#fafaf8;border-radius:8px;padding:8px}.rl-legend{font-size:11px;color:var(--rl-muted);margin:6px 0}.rl-table{border-collapse:collapse;width:100%;font-size:12px;margin:8px 0 16px}.rl-table th,.rl-table td{text-align:right;border-bottom:1px solid var(--rl-border);padding:8px}.rl-table th:first-child,.rl-table td:first-child{text-align:left}.rl-table th{color:var(--rl-muted);font-weight:500}.rl-delta-up{color:#a32d2d}.rl-delta-down{color:var(--rl-green)}.rl-columns{display:grid;grid-template-columns:repeat(auto-fit,minmax(300px,1fr));gap:12px}.rl-note{font-size:12px;color:var(--rl-muted)}.rl-expanded{max-width:100%;overflow:auto;background:#fafaf8;border-radius:8px;padding:8px}.rl-expanded svg{display:block;margin:auto}
+.rl-wrap *{box-sizing:border-box}.rl-title{font-size:18px;font-weight:600;margin:0 0 4px}.rl-subtitle{color:var(--rl-muted);font-size:13px;margin:0 0 16px}.rl-metrics{display:grid;grid-template-columns:repeat(auto-fit,minmax(105px,1fr));gap:8px;margin:12px 0 18px}.rl-metric{background:#f6f5f1;border-radius:8px;padding:10px}.rl-metric-label{font-size:11px;color:var(--rl-muted)}.rl-metric-value{font-size:21px;font-weight:600;margin-top:2px}.rl-section{font-size:14px;font-weight:600;margin:18px 0 8px}.rl-tree,.rl-tree ul{list-style:none;padding-left:18px}.rl-tree{padding-left:0}.rl-tree ul{border-left:1px solid #eae7de;margin-left:8px}.rl-tree details{margin:2px 0}.rl-tree summary{cursor:pointer;padding:6px 9px;border-radius:7px;font-family:ui-monospace,SFMono-Regular,Menlo,monospace;font-size:12.5px;color:#2b2a26}.rl-tree summary:hover{background:#f4f2ec}.rl-result{color:var(--rl-muted);font-size:11px;margin-left:6px}.rl-badge{display:inline-block;border-radius:4px;padding:1px 5px;margin-left:5px;font-family:-apple-system,BlinkMacSystemFont,"Segoe UI",sans-serif;font-size:10px;font-weight:500}.rl-repeat{color:#993c1d;background:var(--rl-coral-bg)}.rl-cached{color:var(--rl-green);background:var(--rl-green-bg)}.rl-error{color:#a32d2d;background:#fcebeb}.rl-warning{color:#854f0b;background:#faeeda;border-radius:6px;padding:8px;margin:8px 0;font-size:12px}.rl-dag{max-width:100%;overflow:auto;background:#fafaf8;border-radius:8px;padding:8px}.rl-legend{font-size:11px;color:var(--rl-muted);margin:6px 0}.rl-table{border-collapse:collapse;width:100%;font-size:12px;margin:8px 0 16px}.rl-table th,.rl-table td{text-align:right;border-bottom:1px solid var(--rl-border);padding:8px}.rl-table th:first-child,.rl-table td:first-child{text-align:left}.rl-table th{color:var(--rl-muted);font-weight:500}.rl-delta-up{color:#a32d2d}.rl-delta-down{color:var(--rl-green)}.rl-columns{display:grid;grid-template-columns:repeat(auto-fit,minmax(300px,1fr));gap:12px}.rl-note{font-size:12px;color:var(--rl-muted)}.rl-expanded{max-width:100%;overflow:auto;background:#fbfaf6;border:1px solid #efede5;border-radius:10px;padding:12px}.rl-expanded svg{display:block;margin:auto}
 """
 
 
@@ -105,6 +105,24 @@ def render_tree(report) -> str:
     )
 
 
+def _mono_width(text: str, size: float = 12.0) -> float:
+    """Approximate the advance width of monospace text at a given size."""
+
+    return len(text) * size * 0.6
+
+
+def _node_style(node) -> Tuple[str, str, str, str, str]:
+    """Return fill, stroke, dot colour, label colour, and dash pattern."""
+
+    if node.cached:
+        return "#eaf6f0", "#9ecbb9", "#0f6e56", "#0b5c48", ""
+    if node.status == "error":
+        return "#fdecec", "#e3a6a6", "#a32d2d", "#a32d2d", ""
+    if node.repeated:
+        return "#fdf1ec", "#e4b3a2", "#d85a30", "#993c1d", ' stroke-dasharray="4 3"'
+    return "#ffffff", "#d8d5cc", "#185fa5", "#242424", ""
+
+
 def render_expanded_tree(report, limit: int = 120) -> str:
     """Render every selected invocation as a compact, fully visible SVG tree."""
 
@@ -126,8 +144,17 @@ def render_expanded_tree(report, limit: int = 120) -> str:
     for root in report.roots:
         collect(root)
 
+    labels = {id(node): compact_call(node.signature) for node in selected}
+    widths = {
+        id(node): max(62.0, min(210.0, _mono_width(labels[id(node)]) + 30))
+        for node in selected
+    }
+
+    node_height = 38.0
+    level_gap = 92.0
+    sibling_gap = 26.0
     positions: Dict[int, Tuple[float, float]] = {}
-    leaf_cursor = [0]
+    cursor = [0.0]
     maximum_depth = 0
 
     def place(node, depth: int) -> None:
@@ -135,32 +162,31 @@ def render_expanded_tree(report, limit: int = 120) -> str:
         maximum_depth = max(maximum_depth, depth)
         children = child_lists[id(node)]
         if not children:
-            x = 80 + leaf_cursor[0] * 135
-            leaf_cursor[0] += 1
+            x = cursor[0] + widths[id(node)] / 2
+            cursor[0] = x + widths[id(node)] / 2 + sibling_gap
         else:
             for child in children:
                 place(child, depth + 1)
             first_x = positions[id(children[0])][0]
             last_x = positions[id(children[-1])][0]
             x = (first_x + last_x) / 2
-        positions[id(node)] = (x, 24 + depth * 72)
+        positions[id(node)] = (x, 30 + depth * level_gap)
 
     for root in report.roots:
         if id(root) in child_lists:
             place(root, 0)
 
-    tree_width = 160 + max(0, leaf_cursor[0] - 1) * 135
-    width = max(600, tree_width)
-    horizontal_shift = (width - tree_width) / 2
-    if horizontal_shift:
+    content_width = max(0.0, cursor[0] - sibling_gap)
+    width = max(640.0, content_width + 64)
+    shift = (width - content_width) / 2
+    if shift:
         positions = {
-            node_id: (x + horizontal_shift, y)
-            for node_id, (x, y) in positions.items()
+            node_id: (x + shift, y) for node_id, (x, y) in positions.items()
         }
-    height = max(110, 24 + maximum_depth * 72 + 56)
+    height = max(140.0, 30 + maximum_depth * level_gap + node_height / 2 + 28)
     parts = [
         f'<svg role="img" aria-label="Fully expanded recursion tree for {esc(report.function_name)}" '
-        f'viewBox="0 0 {width} {height}" width="{width}" height="{height}" '
+        f'viewBox="0 0 {width:.0f} {height:.0f}" width="{width:.0f}" height="{height:.0f}" '
         'xmlns="http://www.w3.org/2000/svg">',
         f"<title>{esc(report.function_name)} expanded call tree</title>",
     ]
@@ -169,39 +195,65 @@ def render_expanded_tree(report, limit: int = 120) -> str:
         parent_x, parent_y = positions[id(parent)]
         for child in child_lists[id(parent)]:
             child_x, child_y = positions[id(child)]
+            start_y = parent_y + node_height / 2
+            end_y = child_y - node_height / 2
+            middle = (start_y + end_y) / 2
             parts.append(
-                f'<path class="rl-expanded-edge" d="M{parent_x},{parent_y + 28} '
-                f'L{child_x},{child_y + 6}" fill="none" stroke="#a9a7a0" stroke-width="1"/>'
+                f'<path class="rl-expanded-edge" d="M{parent_x:.1f},{start_y:.1f} '
+                f'C{parent_x:.1f},{middle:.1f} {child_x:.1f},{middle:.1f} '
+                f'{child_x:.1f},{end_y:.1f}" fill="none" stroke="#c6c3ba" '
+                'stroke-width="1.2"/>'
             )
 
     for node in selected:
         x, y = positions[id(node)]
+        fill, stroke, dot, ink, dash = _node_style(node)
         note = [node.signature]
         if node.cached:
             note.append("cache hit")
-            color = "#0f6e56"
-            decoration = ""
         elif node.status == "error":
             note.append(node.error or "error")
-            color = "#a32d2d"
-            decoration = ""
         elif node.repeated:
             note.append("repeated")
-            color = "#993c1d"
-            decoration = ' text-decoration="underline"'
-        else:
-            color = "#242424"
-            decoration = ""
         if node.result is not None:
             note.append(f"returned {node.result}")
-        parts.extend(
-            [
-                f'<g class="rl-expanded-node"><title>{esc(" · ".join(note))}</title>',
-                f'<text x="{x}" y="{y + 17}" text-anchor="middle" '
-                f'font-family="ui-monospace,SFMono-Regular,Menlo,monospace" font-size="12" '
-                f'fill="{color}"{decoration}>{esc(compact_call(node.signature))}</text></g>',
-            ]
+
+        if node.status == "error":
+            value = "error"
+        elif node.result is not None:
+            value = f"→ {node.result}"
+        else:
+            value = ""
+
+        left = x - widths[id(node)] / 2
+        if value:
+            label_y = y - 2
+            value_y = y + 12
+        else:
+            label_y = y + 4.3
+            value_y = None
+
+        parts.append(f'<g class="rl-expanded-node"><title>{esc(" · ".join(note))}</title>')
+        parts.append(
+            f'<rect x="{left:.1f}" y="{y - node_height / 2:.1f}" '
+            f'width="{widths[id(node)]:.1f}" height="{node_height:.1f}" rx="9" '
+            f'fill="{fill}" stroke="{stroke}" stroke-width="1"{dash}/>'
+        )
+        parts.append(
+            f'<circle cx="{left + 11:.1f}" cy="{y:.1f}" r="3" fill="{dot}"/>'
         )
+        parts.append(
+            f'<text x="{left + 20:.1f}" y="{label_y:.1f}" '
+            f'font-family="ui-monospace,SFMono-Regular,Menlo,monospace" font-size="12" '
+            f'fill="{ink}">{esc(labels[id(node)])}</text>'
+        )
+        if value_y is not None:
+            parts.append(
+                f'<text x="{left + 20:.1f}" y="{value_y:.1f}" '
+                'font-family="-apple-system,BlinkMacSystemFont,Segoe UI,sans-serif" '
+                f'font-size="10" fill="#7a776e">{esc(short(value, 22))}</text>'
+            )
+        parts.append("</g>")
     parts.append("</svg>")
 
     omitted = report.recorded_calls - len(selected)
@@ -213,7 +265,7 @@ def render_expanded_tree(report, limit: int = 120) -> str:
     return (
         '<h3 class="rl-section">Expanded call tree</h3>'
         f'<div class="rl-legend">All {len(selected)} calls are visible. '
-        "Underlined coral = repeated; green = cache hit; red = error.</div>"
+        "White = executed · dashed coral = repeated · green = cache hit · red = error.</div>"
         f'<div class="rl-expanded">{"".join(parts)}</div>{notice}'
     )