Feature cache cleaning

[qchapp]

This pull request introduces several improvements for development workflow, deployment, and project documentation. It adds a full devcontainer setup for VS Code, a GitHub Actions workflow for building and publishing Docker images, and updates documentation to reflect these changes. It also includes a new changelog and minor configuration and dependency updates.

Development Environment & Tooling

• Added a .devcontainer setup (.devcontainer/Dockerfile, .devcontainer/devcontainer.json) for reproducible VS Code development environments using a prebuilt Python 3.12 image, user setup, and convenient extensions. [1] [2]
• Introduced a justfile with common development and serving commands for streamlined local workflows.

Deployment & CI/CD

• Added a GitHub Actions workflow (.github/workflows/publish_image_in_GHCR.yaml) to automatically build, tag, and publish Docker images to GHCR on pushes and PRs, including release note extraction from the changelog and cleanup of PR images.
• Updated the documentation and added instructions for building and running the Docker image, including environment variable usage.

Documentation & Project Organization

• Added comprehensive project architecture and workflow documentation in .github/copilot-instructions.md, covering system design, data flow, error handling, and conventions.
• Introduced a CHANGELOG.md following Keep a Changelog conventions, and updated the project version in pyproject.toml. [1] [2]
• Updated .env.dist and README.md to clarify environment variable usage and reflect changes to model configuration. [1] [2]

Dependencies

• Added pydantic-ai to the project dependencies in pyproject.toml.

These changes collectively improve the developer experience, streamline deployment, and clarify project usage and structure.

[Copilot]

Pull Request Overview

This pull request introduces an experimental Pydantic AI agent architecture with tool-based retrieval, along with Docker deployment infrastructure, developer environment improvements, and enhanced temp file cleanup. The agent provides an alternative execution pathway (feature flag USE_AGENT=1) that uses structured tool calls instead of the original VLM pipeline for tool selection.

Key Changes:

• Adds Pydantic AI agent with tool-based retrieval (search, rerank, repo_info, resolve_demo_link)
• Introduces Docker/devcontainer infrastructure for reproducible development and deployment
• Implements systematic temp file cleanup for preview and DICOM extraction directories
• Simplifies environment variable naming (consolidates model configuration)

Reviewed Changes

Copilot reviewed 29 out of 29 changed files in this pull request and generated 9 comments.

Show a summary per file

File	Description
src/ai_agent/agent/agent.py	New agent orchestration with tool registration and execution flow
src/ai_agent/agent/tools/*.py	Tool implementations for search, rerank, repo_info, and Gradio space interaction
src/ai_agent/agent/utils.py	Agent state management and per-tool quota enforcement
src/ai_agent/api/pipeline.py	Adds agent-facing APIs (retrieve_no_rerank, rerank_only, get_doc)
src/ai_agent/ui/app.py	Integrates agent pathway with USE_AGENT flag and port fallback logic
src/ai_agent/utils/*.py	Enhanced temp file cleanup, improved runnable link selection
tools/image/Dockerfile	Production Docker image with uv package manager
.devcontainer/*	VS Code devcontainer setup for consistent development
.github/workflows/publish_image_in_GHCR.yaml	CI/CD workflow for automated Docker publishing
pyproject.toml	Version bump and pydantic-ai dependency

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

ENTRYPOINT exec form does not invoke a shell, so this will attempt to execute 'source' as a binary rather than sourcing the venv. Use CMD with shell form or switch to: ENTRYPOINT [\"/bin/bash\", \"-c\", \"source .venv/bin/activate && python -m ai_agent.app\"]

Suggested change

	ENTRYPOINT ["source .venv/bin/activate && python -m ai_agent.app"]
	ENTRYPOINT ["/bin/bash", "-c", "source .venv/bin/activate && python -m ai_agent.app"]

[Copilot]

The conditional if handle_file will always be True since handle_file is imported at the module level. If the intention is to guard against import failures, wrap the import in a try/except and set a default. Otherwise, remove the conditional.

Suggested change

	payload.append(handle_file(image_path) if handle_file else image_path)
	payload.append(handle_file(image_path))

[qchapp]

No need to modify that because gradio space tool will be changed soon

[Copilot]

Creating a new empty RAGImagingPipeline instance will not have access to the loaded catalog docs. Use get_pipeline() from .tools.utils instead to access the shared pipeline instance with the full catalog.

[Copilot]

The error message check 'Cannot find empty port' may not match the actual OSError message for port-in-use conditions. Verify the actual error message or use errno.EADDRINUSE for more reliable detection.

[Copilot]

The grep pattern 'version =' (with space before equals) won't match the actual format in pyproject.toml which uses 'version=' (no space). Change to grep 'version=' pyproject.toml or use a more flexible pattern.

Suggested change

	VERSION=$(grep 'version =' pyproject.toml | sed -E 's/version = "([^"]+)"/\1/')
	VERSION=$(grep -E '^\s*version\s*=' pyproject.toml | sed -E 's/^\s*version\s*=\s*["'\'']([^"'\''"]+)["'\'']/\1/')