Reduce node size caps and regenerate the README screenshots

README.md

@@ -3,12 +3,12 @@
 # GraphLink
 
 ![License](https://img.shields.io/badge/License-MIT-green)
-![Python](https://img.shields.io/badge/Python-3.10%2B-blue)
+![Python](https://img.shields.io/badge/Python-3.12%2B-blue)
 ![Stack](https://img.shields.io/badge/Stack-FastAPI%20%2B%20React-blue)
 ![Local First](https://img.shields.io/badge/Local--First-AI%20Workspace-orange)
 ![GitHub stars](https://img.shields.io/github/stars/dovvnloading/Graphlink?style=social)
 
-<img alt="Graphlink canvas - one question branched into two parallel lines of inquiry, each with its own web research node" src="assets/screenshots/canvas-branching.png" />
+<img alt="Graphlink canvas - one question branched into two answers inside a frame, one continuing into a Python sandbox, beside a finished build and the chart it produced" src="assets/screenshots/canvas-branching.png" />
 
 **A local-first, graph-based AI workspace for branching reasoning, tool use, and multi-provider workflows.**
 
@@ -49,6 +49,7 @@ It is built with a Python (FastAPI) backend and a Vite/React/TypeScript single-p
 - **Multiple model backends** — run locally with Ollama or direct GGUF loading via `llama-cpp-python`, or switch to API Endpoint mode for OpenAI-compatible, Anthropic Claude, or Google Gemini. The active mode is switchable in Settings.
 - **Per-branch model routing** — pin a specific model to a node or a whole branch, on top of the per-task model defaults.
 - **Plugin nodes** — attach specialist nodes for web research, code execution, drafting, and repository-aware changes (see [Plugins](#plugins)), plus a plugin SDK with per-plugin capability grants and optional MCP server integration.
+- **The workspace agent** — point it at a folder and give it a task; it reads and writes files there, runs shell commands and Python, and asks before anything that changes your machine. Bounded by a per-task turn budget, and it reads an `AGENTS.md` if the workspace has one.
 - **Repository-aware editing** — Gitlink loads a GitHub repo into structured context, previews file-level changes, and only writes after explicit approval.
 - **Knowledge base and search** — ingest documents into a local knowledge store, search it from a node, and search across every workspace at once.
 - **Workspaces and library** — organize graphs into workspaces with favorites, tags, and archiving; reopen any of them from the Library.
@@ -60,21 +61,21 @@ It is built with a Python (FastAPI) backend and a Vite/React/TypeScript single-p
 - **Diagnostics** — a token and cost counter, a command palette, and an exportable diagnostic bundle for troubleshooting.
 - **Export** — save the whole canvas as a PNG, or export individual nodes: Chat as Markdown, Code as a source file (extension inferred from language), Image as PNG.
 
-Built-in node kinds on the graph surface: **Chat**, **Code**, **Document**, **Thinking**, **HTML**, **Image**, **Conversation**, **Web Research**, **Plan** (the Builder's checklist), **Artifact**, **Gitlink**, **Code Sandbox**, **Note**, and **Chart** — plus Frames, Containers, and Navigation Pins for organizing them.
+Built-in node kinds on the graph surface: **Chat**, **Code**, **Document**, **Thinking**, **HTML**, **Image**, **Conversation**, **Web Research**, **Plan** (the Builder's checklist), **Agent** (the workspace agent's card), **Artifact**, **Gitlink**, **Code Sandbox**, **Note**, and **Chart** — plus Frames, Containers, and Navigation Pins for organizing them.
 
 ## Screenshots
 
 **A build, on the canvas.** The plan node holds the checklist, the live budget counters, and an activity log of every tool call the run made — next to the nodes it actually created.
 
-<img alt="A finished Builder run: plan node with checklist, budgets and activity log, beside the Py-Coder node and chart it produced" src="assets/screenshots/builder-run.png" />
+<img alt="A finished Builder run: plan node with its checklist, spent budgets and activity log, beside the chart it produced" src="assets/screenshots/builder-run.png" />
 
 **Launching a build.** Pick a recipe (its steps are previewed before you commit), choose how much oversight you want, and set the budgets.
 
 <img alt="The Builder launch dialog showing a recipe with its steps previewed, co-pilot/autopilot oversight modes, and budget presets" src="assets/screenshots/builder-launcher.png" />
 
-**Code and charts inline.** Code Sandbox runs Python in a per-node virtualenv with declared dependencies; any node's content can become a chart.
+**Code and charts inline.** The Virtual Environment Runner keeps the requirements, the code and the terminal output on one node; any node's content can become a chart.
 
-<img alt="A chat question feeding a Py-Coder node with code and output, feeding a bar chart" src="assets/screenshots/code-and-charts.png" />
+<img alt="A Virtual Environment Runner node with its requirements, code and terminal output, beside the chart built from its results" src="assets/screenshots/code-and-charts.png" />
 
 ## The Builder
 
@@ -107,7 +108,7 @@ Attach these specialist nodes to a branch from the plugin picker:
 
 ### Requirements
 
-- Python 3.10 or newer. Windows is the primary development target today.
+- Python 3.12 or newer. Windows is the primary development target today.
 - Node.js 22 or newer (`web_ui/.nvmrc` pins 24, the exact version this project is developed against), needed only to build the frontend once.
 - Internet access is optional, and only needed for API Endpoint mode, GitHub-backed plugins, and web research.
 
@@ -222,7 +223,7 @@ Contributions are welcome. See [CONTRIBUTING.md](CONTRIBUTING.md) for setup, dev
 
 | Symptom | Things to check |
 | --- | --- |
-| App does not start | Dependencies installed from `requirements.txt`; the frontend is built (`web_ui/dist/app/index.html` exists — run `cd web_ui && npm run build` if not); launched with `python graphlink_desktop.py` from the repo root; Python 3.10+. |
+| App does not start | Dependencies installed from `requirements.txt`; the frontend is built (`web_ui/dist/app/index.html` exists — run `cd web_ui && npm run build` if not); launched with `python graphlink_desktop.py` from the repo root; Python 3.12+. |
 | Ollama features fail | Ollama installed and running; the selected model has been pulled and exists locally. |
 | Llama.cpp features fail | `llama-cpp-python` installed; the configured path points to a real `.gguf`; try a `chat_format` override or lower runtime settings. Use Ollama or API mode for image/audio. |
 | API mode fails | API key present; base URL correct for OpenAI-compatible mode; the selected models exist on the endpoint. |

backend/domain/graph.py

@@ -2320,8 +2320,8 @@ def _node_wire(self, n: SceneNode) -> dict[str, Any]:
             "chartType": n.state.chart_type if isinstance(n.state, ChartState) else "",
             "chartData": dict(n.state.chart_data) if isinstance(n.state, ChartState) else {},
             "chartError": n.state.chart_error if isinstance(n.state, ChartState) else "",
-            "chartWidth": n.state.chart_width if isinstance(n.state, ChartState) else 680.0,
-            "chartHeight": n.state.chart_height if isinstance(n.state, ChartState) else 500.0,
+            "chartWidth": n.state.chart_width if isinstance(n.state, ChartState) else 480.0,
+            "chartHeight": n.state.chart_height if isinstance(n.state, ChartState) else 340.0,
             "chartAspectLocked": n.state.chart_aspect_locked if isinstance(n.state, ChartState) else True,
             "chartSourceNodeId": n.state.chart_source_node_id if isinstance(n.state, ChartState) else "",
             # R6.3: HTML splitter + chat scroll gaps.

backend/domain/node_states.py

@@ -263,8 +263,8 @@ class ChartState(NodeState):
     chart_type: str = ""
     chart_data: dict = field(default_factory=dict)
     chart_error: str = ""
-    chart_width: float = 680.0
-    chart_height: float = 500.0
+    chart_width: float = 480.0
+    chart_height: float = 340.0
     chart_aspect_locked: bool = True
     chart_source_node_id: str = ""
 

backend/tests/test_canvas.py

@@ -5745,7 +5745,7 @@ def test_add_chart_node_creates_a_chart_connected_to_its_parent():
     assert chart.state.chart_type == "bar"
     assert chart.state.chart_data == _CHART_DATA
     assert chart.state.chart_source_node_id == parent.id
-    assert chart.state.chart_width == 680.0 and chart.state.chart_height == 500.0
+    assert chart.state.chart_width == 480.0 and chart.state.chart_height == 340.0
     assert chart.state.chart_aspect_locked is True
     assert any(e.source == parent.id and e.target == chart.id for e in doc.edges.values())
 
@@ -5833,8 +5833,8 @@ def test_scene_payload_exposes_all_chart_fields():
     assert row["chartType"] == "bar"
     assert row["chartData"] == _CHART_DATA
     assert row["chartError"] == "degraded"
-    assert row["chartWidth"] == 680.0
-    assert row["chartHeight"] == 500.0
+    assert row["chartWidth"] == 480.0
+    assert row["chartHeight"] == 340.0
     assert row["chartAspectLocked"] is True
     assert row["chartSourceNodeId"] == parent.id
 

contracts/graphlink_scene_payload.py

@@ -552,8 +552,8 @@ class SceneNodeRow:
     chartType: str = ""
     chartData: ChartDataRow = field(default_factory=ChartDataRow)
     chartError: str = ""
-    chartWidth: float = 680.0
-    chartHeight: float = 500.0
+    chartWidth: float = 480.0
+    chartHeight: float = 340.0
     chartAspectLocked: bool = True
     chartSourceNodeId: str = ""
     # R6.3: the Source/Preview splitter position (html) and chat scroll

pyproject.toml

@@ -323,8 +323,10 @@ ignore_missing_imports = true
 #   - backend/tests/perf/measure_baselines.py ... `python -m ...` perf measurement CLI (ADR-019 stage 19.1)
 #   - backend/tests/perf/check_baseline.py ..... nightly regression checker, its stdout IS the CI job log (ADR-019 stage 19.3)
 #   - tools/build_app_icon.py ........... one-off dev icon builder
+#   - tools/seed_demo_graph.py .......... builds the README screenshots' fixture; its stdout confirms what it wrote
 "contracts/codegen.py" = ["T201"]
 "backend/evals/__main__.py" = ["T201"]
 "backend/tests/perf/measure_baselines.py" = ["T201"]
 "backend/tests/perf/check_baseline.py" = ["T201"]
 "tools/build_app_icon.py" = ["T201"]
+"tools/seed_demo_graph.py" = ["T201"]

tests/test_node_state_migration.py

@@ -432,8 +432,8 @@ def test_scene_payload_key_set_is_unchanged_by_the_migration():
     "chartType": "",
     "chartData": {},
     "chartError": "",
-    "chartWidth": 680.0,
-    "chartHeight": 500.0,
+    "chartWidth": 480.0,
+    "chartHeight": 340.0,
     "chartAspectLocked": True,
     "chartSourceNodeId": "",
     "isLocked": True,

tools/capture_backend.py

@@ -0,0 +1,51 @@
+"""Boots the real backend for tools/capture_screenshots.mjs.
+
+A sibling of tests_e2e/run_backend.py with two differences, both so the
+screenshots are reproducible rather than dependent on whoever is running
+them:
+
+  * the port and data directory are arguments, so the capture script owns
+    the temp dir it seeded a chats.db into (tools/seed_demo_graph.py) and
+    can hand the same one here;
+  * nothing else. This is create_app() exactly as shipped, serving the real
+    built SPA out of web_ui/dist/app - a screenshot of a mocked app would be
+    worth nothing.
+
+auth_token=None disables the /api and /ws capability token, the same
+"expected only in tests" posture run_backend.py documents: there is no
+desktop shell here to mint one, and this binds to loopback only.
+
+    python tools/capture_backend.py <port> <data-dir>
+"""
+
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+
+REPO_ROOT = Path(__file__).resolve().parent.parent
+sys.path.insert(0, str(REPO_ROOT))
+
+
+def main() -> None:
+    if len(sys.argv) != 3:
+        raise SystemExit("usage: python tools/capture_backend.py <port> <data-dir>")
+    port = int(sys.argv[1])
+    data_dir = Path(sys.argv[2])
+    data_dir.mkdir(parents=True, exist_ok=True)
+
+    import uvicorn
+
+    from backend.app import create_app
+
+    app = create_app(
+        spa_dir=REPO_ROOT / "web_ui" / "dist" / "app",
+        settings_state_file=data_dir / "session.dat",
+        chat_db_path=data_dir / "chats.db",
+        auth_token=None,
+    )
+    uvicorn.run(app, host="127.0.0.1", port=port, log_level="warning")
+
+
+if __name__ == "__main__":
+    main()