Aura

Aura v1.2.1

Focused Actions

Aura can now operate on a specific file region instead of broad chat context, making edits faster and more predictable.

• File viewer: Double-click a workspace file to open it in Aura's right-side code viewer, which supports text selection and highlighting.
• Context menu: Right-click selected text to access actions: Ask Aura, Explain, Fix, Refactor, Simplify, Add logging, Add type hints, and Write tests.
• Keyboard shortcut: Ctrl+Shift+A opens "Ask Aura about selection."
• Structured prompts: Focused prompts automatically include workspace-relative file path, selected text, line range, surrounding context, and action instructions.
• Safety flow unchanged: Edit actions route through the Planner/Worker loop with diff approval. Read-Only mode turns edit actions into suggestion-only.
• Centralized helpers in aura/focused_actions.py for prompt construction and context gathering. Tests added for prompt structure and context behavior.

Persistent Project Memory

Three new subsystems eliminate redundant work and retain context across sessions.

BM25 Codebase Index Caching

• BM25Scorer serializes its full state (term frequencies, document lengths, average length, parameters) to ~/.config/Aura/bm25_cache/<sha256>.json.
• On restart, the index restores instantly; only files with changed mtime are re-tokenized.
• Cache writes are atomic (write .tmp, then rename) to prevent corruption.

AST-Based Repo Map with Prompt Cache Optimization

• aura/repo_map.py generates a structural overview of the workspace (classes, functions, signatures) from Python ASTs.
• The map is injected into all system prompts via {TIER1_CONTEXT}, achieving >95% prompt cache hit rates on the prefix.
• Recomputed only when workspace files change.

SQLite+FTS5 Project Memory

• aura/memory_db.py provides a full-text searchable database at .aura/memory.db.
• Two new Planner tools: search_project_memory and save_to_project_memory.
• Completed Worker dispatch records are automatically persisted and become searchable.

Self-Updater

Aura can now update itself from its own Git repository.

• Backend in aura/updater.py: detects Aura's repo from the package path, runs git fetch, compares HEAD...@{u}, checks git status --porcelain, and pulls with git pull --ff-only (no merge conflicts, no surprise divergences).
• UI in aura/gui/update_dialog.py: modal with Check for Updates, Pull Latest, status fields, and git output/errors.
• Toolbar integration: Update action next to Settings in the top toolbar, wired through main_window.py.
• Tests in tests/test_updater.py for the updater logic.

Other

• Tier1 context (repo map + project rules) computed on startup and refreshed on workspace change, mode switch, or custom prompt change.
• 8 unit tests added for the memory database.
• Focused action tests for prompt and context handling.
• Updater tests for detection and fast-forward pull logic.