v0.4.0

Cost-optimization release. Stacking the five levers documented here can reduce AI cost by 50-75% on a long-running project.

Added

• Protocol section 7.1 — Prompt caching — recommended convention: mark the memory read as cacheable on Anthropic / OpenAI APIs. Memory files do not change between turns; cached reads pay ~10% of the original cost. Single biggest cost lever for direct API users.
• Protocol section 7.2 — Cheap-model offloading — memory writes don't require frontier intelligence. Documents which operations can be offloaded (entries, recaps, summaries) and which must stay on the main model (anti-drift, code, architecture).
• scripts/memory_assistant.py — optional companion script that routes memory operations to any OpenAI-compatible chat completions endpoint (Groq, Together, Fireworks, OpenRouter, Ollama, Anthropic, OpenAI). Subcommands: recap (deterministic by default, no LLM needed), decision-entry, drift-entry. Pure stdlib (urllib).
• scripts/compress.py — optional companion script implementing protocol section 7 automatically. Summarizes oldest entries into decisions-archive-<date>.md, replaces them with a single archive entry. --dry-run, --keep, --threshold flags. Uses the same VIBEMEM_LLM_* env vars as memory_assistant.py.
• README "Cost optimization" section — explicit 5-lever stack with sample math, code snippets, and a savings table.
• 8 new tests (33 total) — covering deterministic recap, JSON extraction from noisy LLM responses, LLM env-var requirement, compress dry-run behavior.

Changed

• CLAUDE.md documents prompt caching as a Claude-specific cost note.
• Protocol version header: 0.3.0 → 0.4.0

Notes

• memory_assistant.py and compress.py are optional companions. The protocol stays portable, files-only, no-infra by default. These scripts are for heavy users who want to optimize.
• Anti-drift (protocol section 4) must stay on the main frontier model. It is the one operation that requires real reasoning. Offloading it would defeat the protocol's most valuable feature.

v0.3.0

[0.3.0] — 2026-05-19

Added

• Real-time anti-drift (protocol section 4) — agent MUST stop and ask for confirmation before writing code that contradicts any entry in the last 50 decisions. Cannot silently override a logged decision. This is the protocol's most visible value to users.
• Memory recap (protocol section 10) — 3-line recap covering stack, in-flight item, and open drift. Triggers at every context reset: fresh session, idle > 15 min, after compaction, on explicit user request (/context, "where are we", etc.), or when memory is re-read mid-session.
• Session-end recap (protocol section 11, new) — before stopping, agent surfaces a 3-5 line summary: changed, logged, next, open question. Lets the user pick up later without scrolling.
• Mono-file mode — template/vibememory.md is a single self-contained file with the lite protocol + memory tables. Install via install.sh --mode mono. Upgrade path to full mode preserved.
• PR-comment GitHub Action — .github/workflows/memory-pr-comment.yml posts a sticky comment summarizing decisions and drifts added in each PR. Backed by scripts/pr_comment.py (3 unit tests).
• Drift detection AFTER the change moved to section 4.5 (kept distinct from real-time anti-drift in section 4).

Changed

• install.sh accepts --mode mono|full (default: full)
• README quickstart restructured around mode choice
• Protocol version header: 0.2.0 → 0.3.0