Highlights

• Heartbeat bootstrap assets for one-link installs: examples/HEARTBEAT.md.template + scripts/setup.sh auto-creates HEARTBEAT.md.
• Added --help/-h usage output for scripts/install-ai.sh, scripts/setup.sh, and scripts/uninstall.sh.
• Added sanitized workspace templates under examples/workspace/ (SOUL/AGENTS/USER/HEARTBEAT/TOOLS/MEMORY) for AI-first adoption.

Notes

• Recommended heartbeat cadence in docs/examples is now 5m.
• README.zh-CN.md now marks v0.4.2 as released.

Highlights

• Added explicit --retrieval-model support to scripts/setup.sh and scripts/install-ai.sh
• memory-retrieval-watchdog-v1 now installs with an explicit stable default model (glm5)
• Clarified OPS_TARGET compatibility with direct chat / group / supergroup and the need for post-install route probes
• Strengthened recommendation to pass an absolute --qmd-path in cron / isolated environments
• Clarified retrieval order: memory/tasks -> memory_search -> memory_get

Why this patch matters

This patch aligns the OSS memory architecture with current OpenClaw runtime behavior and avoids lightweight watchdog timeouts caused by unstable model fallback chains.

Added

• Memory context budget + dynamic profile injection V1:
  • examples/memory/context-profiles.json
  • scripts/memory_context_budget_guard.py
  • scripts/memory_context_pack.py
• Durable-memory conflict checker:
  • scripts/memory_conflict_check.py
• Retrieval health watchdog:
  • scripts/memory_retrieval_watchdog.py
• Setup now installs two additional cron jobs:
  • memory-retrieval-watchdog-v1 (every 30 minutes)
  • memory-qmd-nightly-maintain (daily 03:20)

Changed

• scripts/setup.sh now bootstraps context profile + 4 new scripts into workspace.
• scripts/uninstall.sh now removes all 5 memory-* cron jobs created by setup.
• scripts/validate.sh now validates Python syntax for new scripts and JSON for context-profiles.
• README (EN/ZH), architecture/operations/ai-agent-prompt docs updated for V1 memory upgrades.
• examples/AGENTS-memory-section.md adds budget/profile/conflict-check guidance.

openclaw-memory-final v0.3.0

Highlights

• One-link install is now truly usable: install-ai.sh auto-bootstraps baseline memory files/scripts.
• Product-style bilingual architecture diagrams with explicit multi-agent memory handoff.
• QMD prerequisite and qmd_not_found recovery guide added to README + AI prompt + troubleshooting docs.

Added

• examples/CURRENT_STATE.md.template
• examples/memory-INDEX.md.template
• scripts/mem-log.sh
• scripts/memory-reflect.sh

Changed

• README (EN/ZH): one-link flow + post-install verification + QMD prerequisite guidance.
• docs/architecture.md: short-term workspace + multi-agent memory model and updated diagrams.
• docs/operations.md: baseline usage with installed workspace scripts.
• docs/ai-agent-prompt.md: QMD missing auto-recovery steps.
• docs/troubleshooting-gateway.md: expanded QMD missing/invalid-path diagnosis.
• scripts/setup.sh: auto-initializes baseline files/scripts and reports them in JSON output.
• scripts/validate.sh: validates new helper scripts.

Full changelog: https://github.com/codesfly/openclaw-memory-final/blob/main/CHANGELOG.md

v0.2.3

Fixed

• Fixed scripts/setup.sh JSON parsing bug (NameError: true) caused by heredoc + here-string conflict.
• Fixed scripts/uninstall.sh same parsing issue that could break cron id lookup/removal.

Impact

• setup.sh --print-json now returns stable JSON.
• setup.sh --force-recreate no longer risks duplicate memory-* jobs from false-negative existence checks.
• uninstall.sh now removes memory jobs deterministically.

v0.2.2

• Added one-command installer: bash scripts/install-skills-pack.sh
• Added deterministic success marker for agents: SKILLS_INSTALL_OK + JSON payload
• Updated README (EN/ZH), AI prompt doc, and skills README to prefer scripted install

Quick use

bash scripts/install-skills-pack.sh

v0.2.1

• Added AI-friendly workspace skills pack under examples/skills/
• Added combined archive openclaw-skills-pack-v2026-02-25.tar.gz
• Updated README (EN/ZH) with optional skills-pack install flow
• Updated docs/ai-agent-prompt.md with optional skill bootstrap steps

Included skills

• memory-task-card
• cron-doctor
• long-task-async
• github-release-flow
• heartbeat-ops-check
• trading-stack-autorepair

v0.2.0

• Added sub-agent task memory index (memory/tasks/YYYY-MM-DD.md) for result-only task cards
• Added retrieval order guidance: task cards -> semantic memory -> raw session history
• setup.sh now creates memory/tasks and reports taskMemoryDir in JSON output
• Updated docs and prompts for task-memory workflow