Dhee 7.1.0

Production release for Dhee as the world memory layer and context compiler for AI agents.\n\nHighlights:\n- SQLite-first Narrative Scene Intelligence v1: Series -> Season -> Episode -> Scene -> SceneCard -> MemoryItem.\n- New scene intelligence MCP tools: dhee_scene_start, dhee_scene_event, dhee_scene_end, dhee_scene_context, dhee_narrative_prior.\n- SceneCards as canonical prompt-safe retrieval objects with privacy, secret, supersession, contradiction, and proof-gate filters.\n- NVIDIA/Nemotron embedder and reranker path for high-quality SceneCard retrieval when configured.\n- Auditable Episode, Season, and Series rollups using Gemma 4 31B and Kimi k2.6.\n- README refreshed in a simple, direct product style: world memory, context compiler, repo brain, and bigger-story anticipation.\n\nValidation:\n- 43 focused tests passed, 1 skipped.\n- python -m build passed.\n- twine check passed.\n- Published to PyPI: https://pypi.org/project/dhee/7.1.0/

Dhee 7.0.0 - Production Developer Brain

Dhee 7.0.0 marks the production-ready Developer Brain release for AI coding agents. It frames Dhee as a local-first memory, repo cognition, handoff, routing, task contract, and proof layer for Codex, Claude Code, Cursor, Cline, Gemini CLI, and MCP clients.

Highlights:

• Mandatory memory-quality contract separating canonical facts, passive evidence, test fixtures, operational events, and repo handoff state.
• NVIDIA-compatible default semantic stack with Nemotron embeddings, NVIDIA reranker, and zvec/dhee-accel vector profile.
• Persistent repo brain with symbol, dependency, route, component, test, ownership, failure, and impact maps.
• Shorter README with three Excalidraw-style diagrams explaining the coding-agent flow and repo-impact reasoning.
• Release metadata bumped to 7.0.0 and Production/Stable.

Verification:

• pytest: 1675 passed, 11 skipped
• twine check: wheel and sdist passed
• dhee release check: release allowed

v6.2.0: Production launch hardening

Production launch hardening

• Marked Dhee package metadata as Production/Stable for the public launch.
• Hardened the one-command installer with handoff-bus verification, GitHub repair fallback, non-interactive repo wiring, and a clear dhee ui next step.
• Fixed release packaging so the wheel includes the bundled handoff bus and the shipped Dhee UI assets.
• Verified clean curl install from GitHub main, repo initialization, dhee status, context check, and UI boot.
• Verified package artifacts with twine check and the full local test suite: 1,472 passed, 16 skipped.

v3.3.0: Native Claude Code Hooks

Native Claude Code Hooks — Self-Evolving Sessions Without Markdown

One command makes every Claude Code session learn from its own execution:

dhee install

No CLAUDE.md bloat. No SKILL.md files. No static routing tables. Vector memory with strength-based decay + token-budgeted XML context injection (~630 tokens for rich context, constant regardless of memory volume).

What's new

6 lifecycle hooks wired into Claude Code's native hook system:

• SessionStart — injects full context (last session, insights, performance trends, relevant memories)
• UserPromptSubmit — surfaces per-turn memories relevant to current prompt
• PostToolUse — captures tool outcomes with automatic secret filtering
• PreCompact — checkpoints state to survive context compaction
• Stop / SessionEnd — records outcomes, what worked, what failed

CLI commands:

• dhee task "fix the auth bug" — starts Claude Code with hooks pre-configured
• dhee install / dhee uninstall-hooks — manual hook management

Privacy filter — 10 regex patterns strip API keys, tokens, passwords, and secrets before any tool output enters memory.

XML renderer — priority-ordered sections, budget-enforced. High-priority context (session state, performance) always included. Low-priority sections (episodes, policies) drop gracefully when budget is tight.

Why not CLAUDE.md / RESOLVER.md / SKILL.md?

Markdown files are static. After accumulation they rot — stale patterns at equal weight to current ones, no retrieval ranking, no forgetting. Dhee uses vector memory with Ebbinghaus decay. Relevant memories surface. Irrelevant ones fade. The context budget stays constant whether you have 50 memories or 50,000.

Stats

• 42 new tests (renderer, privacy, installer, dispatch handlers)
• 878 total tests passing, 0 regressions
• ~630 tokens for fully-populated context
• 1500 token budget cap (configurable)