NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
Go modules, #1596 by repository stars
Latest stable
v0.20.0
Last release 18 days ago
04 Nov 2026
Ships on a steady schedule
a new release about every 2 weeks
Nearly every release is documented
notes for 47 of 47 stable releases
Nothing withdrawn
no release was ever pulled
6 months old
88 releases · first in 2026
Drag across the ruler to compare, or tap a release to open it.
Release version bumped to v0.9.1 from the latest main branch.
Nothing published for this version
Your coding agent can read these notes before it upgrades. Set up the MCP server →
One column per month.
Nothing published for this version
大版本:统一任务抽象、MCP 客户端、并行多 Agent、AI 助理写权限治理,外加一整轮工作流编排与可观测打磨。
大版本:统一任务抽象、MCP 客户端、并行多 Agent、AI 助理写权限治理,外加一整轮工作流编排与可观测打磨。
serve_page(只读网页托管)始终可用。tasks / flows / mcp_servers / approvals / secrets 等表 + reports.task_id、run_id 两列,不破坏存量数据;实测 v0.8.x → v0.9.0 迁移零报错。task_id 反向归属任务、任务详情总览全部产物;「立即生成 / 新建定时」合并为单个「新建任务 ▾」,产物页只读。mcp__<server>__<tool>);认证用「命名凭证 + 请求头模板」静态注入,凭证加密存储、运行时进程内解密、从不经 API 序列化。serve_page(HTML → 可分享网页)、send_im_message(发飞书 / 钉钉)。@ 调设备 / 资源;alert_fired 触发器事件重开时也触发;get_edge_summary 省略 device_ids 即汇总全部边端。pages / workspace / tools 目录(修 serve_page / cloud_bash 权限报错);edge bundle 按实际架构重建。https://ongrid.cloud/dl/ongrid-v0.9.0-linux-amd64.tar.xz(arm64 同理)sha256
8a18b720eae0bbcff6639e1a51c132941c33b84923964c01303ec38c31552f388131fa9855193d783db143d390d71c3a5594edf355c62b927d8e54b18fb4fb4a大版本:统一任务抽象、MCP 客户端、AI 助理写权限治理,外加一整轮工作流编排与可观测打磨。
serve_page(托管只读网页)始终
可用,不受写开关限制。tasks 表 + flows/mcp_servers/approvals/secrets 等新表
reports.task_id/run_id 两列),不破坏存量数据;已实测 v0.8.x → v0.9.0 迁移零报错。task_id 反向归属任务,任务详情总览其全部产物。
「立即生成 / 新建定时任务」合并为单个「新建任务 ▾」下拉。生成动作全部归任务侧,产物页只读。mcp__<server>__<tool>)。认证采用
「命名凭证 + 请求头模板」静态注入(Bearer / API-Key 等),凭证加密存 secrets 库、运行时进程内
解密、从不经 API 序列化;trusted 标志独立于认证,只决定同步直跑 vs 走审批。serve_page(HTML → 可分享内网网页)、send_notification(发飞书 / 钉钉)。@ 调出设备 / 资源。alert_fired 触发器在事件重新打开时也触发(不止新建)。get_edge_summary 省略 device_ids 即汇总全部边端(最多 16),「巡检所有设备」
无需先查清单。Nothing published for this version
Ubuntu 22.04+, Debian 12+, RHEL/Rocky 9 are supported.
Ubuntu 22.04+, Debian 12+, RHEL/Rocky 9 are supported.
wget https://github.com/ongridio/ongrid/releases/download/v0.8.7/ongrid-v0.8.7-linux-amd64.tar.xz
tar -xf ongrid-v0.8.7-linux-amd64.tar.xz && cd ongrid-v0.8.7-linux-amd64 && sudo ./install.sh
wget https://github.com/ongridio/ongrid/releases/download/v0.8.7/ongrid-v0.8.7-linux-arm64.tar.xz
tar -xf ongrid-v0.8.7-linux-arm64.tar.xz && cd ongrid-v0.8.7-linux-arm64 && sudo ./install.sh
Full Changelog: https://github.com/ongridio/ongrid/compare/v0.8.6...v0.8.7
Ubuntu 22.04+, Debian 12+, RHEL/Rocky 9 are supported.
Ubuntu 22.04+, Debian 12+, RHEL/Rocky 9 are supported.
wget https://github.com/ongridio/ongrid/releases/download/v0.8.6/ongrid-v0.8.6-linux-amd64.tar.xz
tar -xf ongrid-v0.8.6-linux-amd64.tar.xz && cd ongrid-v0.8.6-linux-amd64 && sudo ./install.sh
wget https://github.com/ongridio/ongrid/releases/download/v0.8.6/ongrid-v0.8.6-linux-arm64.tar.xz
tar -xf ongrid-v0.8.6-linux-arm64.tar.xz && cd ongrid-v0.8.6-linux-arm64 && sudo ./install.sh
Full Changelog: https://github.com/ongridio/ongrid/compare/v0.8.5...v0.8.6
Nothing published for this version
Ubuntu 22.04+, Debian 12+, RHEL/Rocky 9 are supported.
Ubuntu 22.04+, Debian 12+, RHEL/Rocky 9 are supported.
wget https://github.com/ongridio/ongrid/releases/download/v0.8.5/ongrid-v0.8.5-linux-amd64.tar.xz
tar -xf ongrid-v0.8.5-linux-amd64.tar.xz && cd ongrid-v0.8.5-linux-amd64 && sudo ./install.sh
wget https://github.com/ongridio/ongrid/releases/download/v0.8.5/ongrid-v0.8.5-linux-arm64.tar.xz
tar -xf ongrid-v0.8.5-linux-arm64.tar.xz && cd ongrid-v0.8.5-linux-arm64 && sudo ./install.sh
Full Changelog: https://github.com/ongridio/ongrid/compare/v0.8.4...v0.8.5
Nothing published for this version
Nothing published for this version
Nothing published for this version
Install (Ubuntu 22.04+/Debian 12+/RHEL/Rocky 9):
Install (Ubuntu 22.04+/Debian 12+/RHEL/Rocky 9):
AMD64:
wget https://github.com/ongridio/ongrid/releases/download/v0.8.4/ongrid-v0.8.4-linux-amd64.tar.xz
tar -xf ongrid-v0.8.4-linux-amd64.tar.xz && cd ongrid-v0.8.4-linux-amd64 && sudo ./install.sh
ARM64:
wget https://github.com/ongridio/ongrid/releases/download/v0.8.4/ongrid-v0.8.4-linux-arm64.tar.xz
tar -xf ongrid-v0.8.4-linux-arm64.tar.xz && cd ongrid-v0.8.4-linux-arm64 && sudo ./install.sh
UI polish: Health page (compact layout, indigo accents, de-emphasized OK status), Traces light-mode color fix, report schedules grid, Home hero spacin
UI polish: Health page (compact layout, indigo accents, de-emphasized OK status), Traces light-mode color fix, report schedules grid, Home hero spacing. Platform health check (Settings -> Health, admin-only). Edge deployment guide + Telegram/Slack community badges. ssh.service log-filter fix.
Install (Ubuntu 22.04+/Debian 12+/RHEL/Rocky 9): wget https://github.com/ongridio/ongrid/releases/download/v0.8.3/ongrid-v0.8.3-linux-amd64.tar.xz tar -xf ongrid-v0.8.3-linux-amd64.tar.xz && cd ongrid-v0.8.3-linux-amd64 && sudo ./install.sh
Nothing published for this version
Nothing published for this version
Builds on v0.8.1. The report list now reads calm and consistent with the other list pages:
Builds on v0.8.1. The report list now reads calm and consistent with the other list pages:
Includes all v0.8.0/v0.8.1 content (Scheduled Reports, chat tool-id root fix, report i18n).
Builds on v0.8.0 (Scheduled Reports + chat tool-id root fix + report i18n).
Builds on v0.8.0 (Scheduled Reports + chat tool-id root fix + report i18n).
All v0.8.0 fixes included (report i18n: title / narrative / markdown all follow the requester's UI locale).
定时/手动生成的运维报告:日报 / 周报 / 月报 + 自定义 cron,时区感知调度器。
compose.GetToolCallID(ctx),不再用完成顺序去猜 call_XX id;从源头消除合成 id 孤儿消息。完整变更见仓库提交历史。
Eight changes since v0.7.176. Five of them harden the chat-completion path against eino ToolsNode's parallel persistence drops + the dirty histories D
Eight changes since v0.7.176. Five of them harden the chat-completion path against eino ToolsNode's parallel persistence drops + the dirty histories DeepSeek-strict providers refuse to replay. One unifies LLM timeouts across every call site. Two clean up the curl-pipe edge uninstaller. And the roadmap doc lands open-core.
AI 助查 · PromQL returned HTTP 502 immediately on every translate. queryTranslateTimeout was hard-coded at 6 s, tuned for a Haiku-class default; DeepSeek v4-flash averages 5–8 s for short JSON output and tripped the wall. Bump to 20 s; also surface plain-text response bodies into ApiError.message so the UI no longer drops the actionable detail.pkg/llm/client.go::defaultTimeout, query_translate.go::queryTranslateTimeout, alert/investigator::SummarizerTimeout, and aiops/investigator::defaultLLMTimeout. DB / tool / probe / monitor-sync timeouts deliberately untouched.b528bfb0-..., an assistant emitted 4 parallel host_bash tool_calls but only 2 of 4 role=tool rows landed in chat_messages. Replay then sent an envelope DeepSeek rejected with 400 insufficient tool messages following tool_calls. The error classifier ALSO mis-categorised this as "余额不足/限流" because of a bare insufficient substring match — operator chased provider config while the actual bug was history corruption. Two fixes: (a) replay precheck — before emitting an assistant with hydrated ToolCalls, every call_id must have a hoistable role=tool row, or the whole turn is dropped; (b) classifier tightened (insufficient_quota, insufficient balance, credit balance is too low, etc.) + a distinct case for the protocol error that tells the user truthfully: this session's history is poisoned, open a new one.PersistenceHandler now tracks the per-assistant tool batch, marks each tool_call_id seen in persistToolEnd, and at the next ChatModel.OnStart (and on chatruntime defer for browser-close cases) writes a stub role=tool row for every missing call_id. Stub body honestly says {"error":"tool response was not persisted ...","autoheal":true}. Adds ongrid_chat_tool_response_loss_total{outcome,tool_name} counter + info-level traces on every callback firing so the underlying eino bug becomes observable instead of being discoverable only via user-reported HTTP 400. Tests cover the 4-of-4 / 2-of-4 / sequential-batches / no-batch / finalize-idempotent cases.https://<server>/uninstall.sh) printed [OK] while every agent process kept running. Two issues: (1) only node_exporter + process_exporter got rm -f'd, leaving promtail + plugin work dir orphaned; (2) the if systemctl list-unit-files | grep -q ... precondition silently skipped on hosts where formatting confused the anchored grep. Fix: wholesale rm -rf of /usr/local/lib/ongrid-edge + /var/lib/ongrid-edge; stop units unconditionally + pkill -9 -f /usr/local/{bin,lib}/ongrid-edge as a belt-and-suspenders.ROADMAP.md + ROADMAP.zh-CN.md land at the open-core root with nested-bullet structure, A–L sections, status legend ◯ □ ◐ ✓. Sections A (root-cause RCA), B–G (value chain + ops), H (sandbox), I (security/compliance), J (ecosystem), K (SOP), L (periodic agent jobs).wget https://github.com/ongridio/ongrid/releases/download/v0.7.179/ongrid-v0.7.179-linux-amd64.tar.xz
tar -xJf ongrid-v0.7.179-linux-amd64.tar.xz
cd ongrid-v0.7.179-linux-amd64
bash upgrade.sh
In-place; no schema migration. Edge agents pick up #46/#47 on their next bundle-upgrade heartbeat (ADR-024).
59f7cb572bb514f3373de5b4dfb8af433610df4044247391a4f6964047e16052 ongrid-v0.7.179-linux-amd64.tar.xz
Seven fixes since v0.7.169 — five of them resolve chat-stream HTTP 400 failures we tripped in production after rolling DeepSeek v4. The remaining two
Seven fixes since v0.7.169 — five of them resolve chat-stream HTTP 400 failures we tripped in production after rolling DeepSeek v4. The remaining two clear up UI auto-scroll and an upstream node_exporter gap on modern kernels.
DeepSeek v4 enforces the OpenAI tool-call protocol strictly: every assistant message with tool_calls MUST be followed by tool responses pairing each tool_call_id. Several historical shortcuts in our persistence + replay layers produced subtly malformed history that older permissive providers tolerated and DeepSeek v4 rejected.
Content=NULL assistant rows (the natural shape for a turn that emitted only tool_calls). The following tool messages became orphans and the next request 400'd. Hydrated Message.ToolCalls via JOIN, added chat_tool_calls.llm_call_id column, factored the pairing logic into a new toolreplay package shared between the legacy agent path and the graph kernel.PersistenceHandler exposed WithMessageID / WithToolCallID seams that production code never called → chat_tool_calls.message_id was empty → NOT NULL violation → silent log warn → 0 rows written. Tool messages received synthetic name|type IDs. Handler now self-wires via lastAssistantID + a pendingLLMCalls FIFO populated from msg.ToolCalls.AgentTool) called the routing chat model with no opts and defaulted to openai. On installs configured for Zhipu / Anthropic / DeepSeek only, the specialist call hard-failed with "provider not configured". Coordinator LLM choice now propagates through context (basetool.WithLLMChoice) into SpawnRequest{Provider, Model} and the worker's chat model opts.AgentTool finishes seconds after parallel sibling tools, persistence stores rows in created_at order, which interleaves the next assistant turn between the parent and its delayed tool response. DeepSeek then 400'd with "insufficient tool messages following tool_calls message". Replay now indexes tool messages by tool_call_id and hoists each immediately after its parent assistant via emitToolByCallID.useEffect measured distance AFTER React had already appended the new message, so scrollHeight - scrollTop - clientHeight always cleared the 80px threshold and the code bailed. Replaced with a stickToBottomRef flipped by the scroller's own onScroll (120px slack at the bottom).node_exporter 1.8.2's conntrack collector reads /proc/sys/net/nf_conntrack_count, which doesn't exist on modern kernels (the file lives at /proc/sys/net/netfilter/nf_conntrack_count). The collector registered but emitted nothing, leaving the manager's "conntrack 利用率" panel empty on every edge. hostmetrics plugin now wraps the subprocess with a 15s producer writing node_nf_conntrack_entries{,_limit} to a managed textfile collector dir. #45 adds a guard so the shim stays quiet on older kernels (where the built-in collector is already emitting the same series).wget https://github.com/ongridio/ongrid/releases/download/v0.7.176/ongrid-v0.7.176-linux-amd64.tar.xz
tar -xJf ongrid-v0.7.176-linux-amd64.tar.xz
cd ongrid-v0.7.176-linux-amd64
bash upgrade.sh
In-place; no data migration. Edge agents pick up the conntrack fix automatically on their next bundle-upgrade heartbeat (ADR-024).
687f50938e8960640455416940c169545c9a820be167cd2bd3a302fc6400125b ongrid-v0.7.176-linux-amd64.tar.xz
Single fix: IM bridge (Lark / Slack / Telegram) was passing empty agent.RunOptions to the chat runtime, bypassing the operator-set default_provider /
Single fix: IM bridge (Lark / Slack / Telegram) was passing empty agent.RunOptions to the chat runtime, bypassing the operator-set default_provider / <provider>_default_model. The agent layer then fell through to its hard-coded gpt-5.4 fallback, which (on installs without an OpenAI key) gets routed by the LLM router to whichever relay shares the same key — and the relay rejected gpt-5.4 with "this model has beta-limitations…". End users saw "助手执行失败" in chat apps.
After this fix, the IM path resolves its provider+model from the same LLMSettingsResolver the web SPA picker writes to. HTTP chat path is unchanged.
Home model dropdown now auto-flips up/down based on viewport space
Five fixes bundled:
Nothing published for this version
Hardening release. First-time install on a clean CN box now works end-to-end.
Hardening release. First-time install on a clean CN box now works end-to-end.
/etc/docker/daemon.json before pulling imagesONGRID_PUBLIC_URL auto-suggested from cloud metadata (Tencent / Aliyun / AWS / GCP / Azure IMDS) instead of internal NIC IPLIBARCHIVE.xattr.com.apple.provenance noise on Linux extractwget https://github.com/ongridio/ongrid/releases/download/v0.7.167/ongrid-v0.7.167-linux-amd64.tar.xz
tar -xf ongrid-v0.7.167-linux-amd64.tar.xz && cd ongrid-v0.7.167-linux-amd64
sudo ./install.sh
Linux amd64 only (single distro target this release).
Nothing published for this version
发布包 703 MB → 381 MB(−46%):打包改用 xz -9e;edge 升级 bundle 不再双打包进主包,改为装机/升级时在主机侧从松散二进制重建(ADR-024 一键升级链路不变)。
发布包 703 MB → 381 MB(−46%):打包改用 xz -9e;edge 升级 bundle 不再双打包进主包,改为装机/升级时在主机侧从松散二进制重建(ADR-024 一键升级链路不变)。
gh release download v0.7.159 --repo ongridio/ongrid -p 'ongrid-v0.7.159-linux-amd64.tar.xz*'
tar xf ongrid-v0.7.159-linux-amd64.tar.xz && cd ongrid-v0.7.159-linux-amd64
sudo ./install.sh
解包命令不变(tar xf 自动识别 xz)。详见 ADR-032。
First tagged release of Ongrid — the AI agent for Operations.
First tagged release of Ongrid — the AI agent for Operations.
ongrid-v0.7.158-linux-amd64.tar.gz — full self-hostable stack (manager + frontier + web images, install.sh / upgrade.sh / uninstall.sh, nginx / Prometheus / Grafana / Loki / Tempo / qdrant configs). Linux amd64.edge-bundle-linux-amd64-v0.7.158.tar.gz — edge-agent one-button upgrade bundle (ADR-024).*.sha256 — per-asset checksums.gh, not a bare wget)gh release download v0.7.158 --repo ongridio/ongrid -p 'ongrid-v0.7.158-linux-amd64.tar.gz*'
sha256sum -c ongrid-v0.7.158-linux-amd64.tar.gz.sha256
tar xzf ongrid-v0.7.158-linux-amd64.tar.gz && cd ongrid-v0.7.158-linux-amd64
sudo ./install.sh # compose mode; see deploy/install/README.md for systemd
Built from the v0.7.158 cut and regression-verified on the test server: server/edge management, observability (Prometheus / Grafana / Loki / Tempo), sessions, agents & skills, notifications & channels.