cowork-dash 0.1.8__tar.gz → 0.2.0__tar.gz

{cowork_dash-0.1.8 → cowork_dash-0.2.0}/.claude/settings.local.json

@@ -18,7 +18,9 @@
       "Bash(uv run python:*)",
       "Bash(uv run ruff check:*)",
       "Bash(uv run pytest:*)",
-      "Bash(uv add:*)"
+      "Bash(uv add:*)",
+      "Bash(source:*)",
+      "Bash(pytest:*)"
     ],
     "deny": [],
     "ask": []

{cowork_dash-0.1.8 → cowork_dash-0.2.0}/CHANGELOG.md

@@ -5,6 +5,36 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.2.0] - 2026-02-01
+
+### Added
+- **display_inline tool**: Rich inline content rendering for images, DataFrames, CSV, HTML, JSON, and PDF files
+- **think_tool integration**: Agent reasoning is now visible in chat with collapsible thinking blocks
+- **"Add to Canvas" button**: Display inline results can be added to canvas directly from chat
+- **Sandboxed bash execution**: Virtual FS mode now supports safe bash commands using bubblewrap isolation (Linux only)
+- **CSV/TSV file preview**: File modal now shows tabular data with pagination
+- **Comprehensive test suite**: Added tests for CLI, run_app API, agent loading, config, and components
+
+### Changed
+- **Redesigned UI**: Minimal professional styling with improved visual hierarchy
+- **Loading indicator**: Replaced verbose "AGENT" + "Thinking" header with clean DMC dots loader
+- **Tool call rendering**: Simplified with CSS classes for better consistency
+
+### Fixed
+- **HITL interrupt handling**: Tool calls now correctly show status after human-in-the-loop approval
+- **Duplicate tool calls**: Fixed issue where tool calls appeared twice after resuming from interrupt
+- **Breadcrumb bar dark mode**: Fixed background color in dark theme
+- **Expanded folders during execution**: Fixed folders showing "Loading..." during agent runs
+
+### Platform
+- **Virtual FS restricted to Linux**: `--virtual-fs` flag now only works on Linux (shows warning on other platforms)
+- **CLI warning**: Added helpful message when `--virtual-fs` is used on non-Linux systems
+
+## [0.1.9] - 2026-01-28
+
+### Changed
+- Updated lock file version reference
+
 ## [0.1.8] - 2026-01-27
 
 ### Fixed
@@ -141,6 +171,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Resizable split-pane interface
 - Upload/download functionality for files
 
+[0.2.0]: https://github.com/dkedar7/cowork-dash/compare/v0.1.9...v0.2.0
+[0.1.9]: https://github.com/dkedar7/cowork-dash/compare/v0.1.8...v0.1.9
 [0.1.8]: https://github.com/dkedar7/cowork-dash/compare/v0.1.7...v0.1.8
 [0.1.7]: https://github.com/dkedar7/cowork-dash/compare/v0.1.6...v0.1.7
 [0.1.6]: https://github.com/dkedar7/cowork-dash/compare/v0.1.5...v0.1.6

cowork_dash-0.2.0/Dockerfile.test

@@ -0,0 +1,23 @@
+# Dockerfile for running virtual FS tests in Linux environment
+FROM python:3.11-slim
+
+# Install bubblewrap for sandboxing
+RUN apt-get update && apt-get install -y \
+    bubblewrap \
+    git \
+    && rm -rf /var/lib/apt/lists/*
+
+# Set working directory
+WORKDIR /app
+
+# Copy all project files
+COPY pyproject.toml uv.lock README.md ./
+COPY cowork_dash/ ./cowork_dash/
+COPY tests/ ./tests/
+
+# Install uv and dependencies
+RUN pip install uv && \
+    uv sync --frozen
+
+# Default command: run sandbox tests
+CMD ["uv", "run", "pytest", "tests/test_sandbox.py", "-v"]

{cowork_dash-0.1.8 → cowork_dash-0.2.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: cowork-dash
-Version: 0.1.8
+Version: 0.2.0
 Summary: AI Agent Web Interface with Filesystem and Canvas Visualization
 Project-URL: Homepage, https://github.com/dkedar7/cowork-dash
 Project-URL: Documentation, https://github.com/dkedar7/cowork-dash/blob/main/README.md

{cowork_dash-0.1.8 → cowork_dash-0.2.0}/cowork_dash/agent.py

@@ -7,22 +7,24 @@ from langgraph.checkpoint.memory import InMemorySaver
 from cowork_dash.tools import (
     add_to_canvas,
     bash,
-    clear_notebook_canvas_items,
     create_cell,
     delete_cell,
     execute_all_cells,
     execute_cell,
-    get_notebook_canvas_items,
     get_script,
     get_variables,
     insert_cell,
     modify_cell,
     reset_notebook,
+    display_inline,
+    think_tool
 )
 
 SYSTEM_PROMPT = """You are a helpful AI assistant with access to a filesystem workspace and a Python code execution environment.
 You have access to a canvas, which is a markdown file located at `.canvas/canvas.md` in the workspace. This allows you to sketch out ideas, document your work, and present results to the user.
 
+ALWAYS use think_tool to reason through user requests, irrespective of complexity. Use it regularly. Human sees your thought process.
+
 ## Capabilities
 
 ### Filesystem
@@ -57,15 +59,27 @@ You have tools to write and execute Python code interactively, similar to a Jupy
 - Captures stdout, stderr, and return values
 - Shows detailed error tracebacks when code fails
 
+### Displaying Results Inline
+- `display_inline(content, title=None, display_type=None)` - Display content inline in the chat
+  - Use for: **file paths** to images, DataFrames, CSV files, HTML, JSON, PDF
+  - **For matplotlib figures**: Save to file first, then display the file path:
+    1. In a cell: `fig.savefig('chart.png', bbox_inches='tight')`
+    2. Then call: `display_inline("chart.png", title="My Chart")`
+  - Example: `display_inline("results.csv", title="Sales Data")`
+
 ### Canvas Visualization
-- `add_to_canvas(content)` - Display DataFrames, charts, images, or markdown for the user (use as a tool directly)
-- Inside notebook cells, `add_to_canvas()` is also available - call it in your code to add visualizations
-- `get_notebook_canvas_items()` - Retrieve canvas items generated by executed cells
-- `clear_notebook_canvas_items()` - Clear collected canvas items
+- `add_to_canvas(content)` - Add content to the canvas panel (persistent note-taking area)
+- Inside notebook cells, `add_to_canvas()` handles matplotlib figures directly
+- For matplotlib in cells: `add_to_canvas(fig)` works (auto-converts to image)
+- Canvas items are saved to `.canvas/` directory
+
+**When to use `display_inline` vs `add_to_canvas`:**
+- `display_inline`: For showing **saved files** (images, CSV, JSON) inline in chat
+- `add_to_canvas`: Inside notebook cells for **figure objects** (handles conversion)
 
 **Important:** When creating charts in notebook cells:
-1. Create the figure in a cell and call `add_to_canvas(fig)` within the same cell
-2. The execution result will include `canvas_items` showing what was added
+1. Create the figure and call `add_to_canvas(fig)` in the same cell
+2. The execution result will show `canvas_items` with what was added
 3. Charts are automatically saved to the `.canvas/` directory
 
 ## Workflow Guidelines
@@ -87,10 +101,17 @@ Work iteratively like a human using Jupyter:
 
 ### General
 1. ALWAYS use write_todos to track your progress and next steps
-2. ALWAYS think_tool to reason through reqests, irrespective of complexity
+2. ALWAYS use think_tool to reason through reqests, irrespective of complexity. Use it regularly. Human sees your thought process.
 3. Be proactive in exploring the filesystem when relevant
 4. Provide clear, helpful responses
 
+CRITICAL WORKFLOW REQUIREMENT:
+1. FIRST ACTION: Always call think_tool before any other tool or response
+2. Use it to reason through the user's request
+3. Then proceed with other tools/actions
+
+This applies to EVERY user message, regardless of complexity.
+
 The workspace is your sandbox - feel free to create files, organize content, and help users manage their projects."""
 
 # Get workspace root from environment variable or default to current directory
@@ -110,8 +131,8 @@ AGENT_TOOLS = [
     get_script,
     get_variables,
     reset_notebook,
-    get_notebook_canvas_items,
-    clear_notebook_canvas_items,
+    display_inline,
+    think_tool
 ]
 
 # Global agent for physical filesystem mode