Huawei Cloud Agent Integration (OpenViking Long-Term Memory)

Integrate / unbind OpenViking long-term memory for coding agents running in bwrap sandboxes (/root/job-envs/sandboxes/). Every agent uses its native mechanism (MCP, HTTP memory provider, TS extension hooks, settings.json), so integration survives agent upgrades. Writes are template-level persistent: config is injected into /root/template/<agent>/start.sh, so sandbox stop + start preserves the integration.

Supported Agents

Agent	Mechanism
CodeArts CLI	@openviking/opencode-plugin → .codeartsdoer/
OpenCode	@openviking/opencode-plugin (npm mirror → GitHub fallback)
OpenClaw	clawhub:@openviking/openclaw-plugin + contextEngine slot (disables session-memory hook, denies memory_search)
Hermes	Built-in memory.provider: openviking (HTTP REST, no MCP)
WorkSwarm	Dual-channel: native provider + MCP (15 tools) + code-mode patch
KimiCode	hooks + MCP (session lifecycle) — config.toml [[hooks]] → ov-kimi-hook.sh + MCP via mcp.json
DeepSeek Harness	@openviking/dsh-memory-plugin bundle (on-demand from GitHub)
Prime Agent	@openviking/pi-coding-agent-extension (on-demand from GitHub)

Per-agent config paths and details: references/agent-configs.md.

Prerequisites

• OpenViking server healthy: curl -s http://127.0.0.1:1933/health → healthy.
• Target sandbox exists under /root/job-envs/sandboxes/.
• Host tools: curl, python3, bash; npm for OpenCode/OpenClaw; pip install --upgrade "mcp>=2.0" for verify_mcp.sh.
• No Huawei Cloud IAM required — operates on local bwrap sandboxes only.

参数 (Parameters)

Flag	Required	Meaning
--agent <name>	Yes (unless --all)	Target agent
--all	Yes (unless --agent)	Operate on all 8 agents
--endpoint <url>	No	OpenViking URL (default http://127.0.0.1:1933)
--api-key <key>	No	For auth-mode servers. Never echo in chat. Must be consumed by an agent config or verify_mcp.sh; an unconsumed --api-key is reported as a warning.
--dry-run	No	Show changes without applying
--yes / -y	No	Skip confirmation (automation only)
--json	No	status.sh: machine-readable output

核心命令

功能	命令
查看状态	scripts/status.sh（--json / --agent <name>）
验证 MCP	scripts/verify_mcp.sh
集成	scripts/integrate.sh --agent <name> [--endpoint URL] [--dry-run] [--yes]（或 --all）
解绑	scripts/unbind.sh --agent <name> [--dry-run] [--yes]（或 --all）

Workflow

SKILL_DIR=/root/.agents/skills/huawei-cloud-openviking-agent-integration
$SKILL_DIR/scripts/status.sh          # per-agent: template+live / template only / live only / none
$SKILL_DIR/scripts/verify_mcp.sh      # full MCP handshake (initialize → tools/list → health)
$SKILL_DIR/scripts/integrate.sh --agent <name> [--yes]
$SKILL_DIR/scripts/unbind.sh  --agent <name> [--yes]

1. Status first — never claim integrated before status.sh confirms (do not fabricate state).
2. Verify MCP — verify_mcp.sh before integration.
3. Integrate / Unbind — each requires explicit confirm (or --yes); --dry-run previews safely.
4. States: template + live = active; template only = activates on restart; live only = lost on restart (warn user).

Authorization & Safety

• integrate.sh / unbind.sh require explicit confirmation; never edit agent configs directly — go through the skill scripts.
• Every config modification creates .bak.<timestamp> (keep 5) for rollback.
• --api-key values must never appear in output.

能力边界（Cannot Do）

本技能只做"OpenViking 记忆 ↔ 编码 agent 的集成/解绑/状态/验证"。以下操作明确不做，遇到请改走对应路径或明确报错：

• 不直接读写 agent 配置文件、sandbox 内业务数据或用户代码——一律通过 scripts/integrate.sh / unbind.sh 等入口执行。
• 不删除、不修改用户自装的第三方包/目录/数据。卸载只清理 OpenViking 自身产物（@openviking/* 插件、注入的配置段、OV 专属目录）；node_modules 整目录删除等越界行为是缺陷，不执行。
• 服务器不可达/健康检查失败时不报告成功：integrate.sh、unbind.sh、verify_mcp.sh 均先做健康门禁，失败即失败（exit 非 0），绝不假成功。
• --dry-run 只预览改动，不写任何持久文件；预览后续真实执行仍需确认。
• --api-key 必须被某个 agent 配置或验证流程实际消费；若传入却无处消费，脚本会告警（而非静默忽略）。
• 不做 embedding 模型切换、向量库重建、服务器自身配置变更——那是 huawei-cloud-openviking-embedding-switch 等技能的范围。
• 不支持在集成/解绑过程中创建或修改 OpenViking 账号、用户、权限体系。
• 不承诺"查看 → 已集成"：一律以 status.sh 实测结果为准，禁止凭模板猜测 live 状态。

References

Document	When to read
agent-configs.md	Per-agent config paths/formats, MCP tools, server info
notes.md	Guardrails, env vars, sandbox restart, plugin sources, troubleshooting, adding a new agent