repo stringclasses 1
value | github_id int64 3.77B 4.38B | github_node_id stringlengths 18 24 | number int64 75 77.8k | html_url stringlengths 46 49 | api_url stringlengths 56 59 | title stringlengths 1 746 | body stringlengths 1 55.7k ⌀ | state stringclasses 2
values | state_reason stringclasses 4
values | locked bool 2
classes | comments_count int64 0 128 | labels listlengths 0 5 | assignees listlengths 0 2 | created_at stringdate 2026-01-01 16:55:34 2026-05-05 11:59:03 | updated_at stringdate 2026-03-24 00:13:29 2026-05-05 11:59:15 | closed_at stringlengths 20 20 ⌀ | author_association stringclasses 3
values | milestone_title stringclasses 0
values | snapshot_id stringclasses 76
values | extracted_at stringdate 2026-03-24 23:36:49 2026-05-05 12:00:33 | author_login stringlengths 3 38 | author_id int64 346 282M | author_node_id stringlengths 12 20 | author_type stringclasses 2
values | author_site_admin bool 2
classes |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
openclaw/openclaw | 4,044,521,762 | I_kwDOQb6kR87xEoEi | 40,832 | https://github.com/openclaw/openclaw/issues/40832 | https://api.github.com/repos/openclaw/openclaw/issues/40832 | Stable/beta bundled plugins can fail to load when published artifacts omit required package dependencies | ## Summary
There is a mismatch between the stable/beta plugin model and the runtime behavior of some bundled plugins.
In stable/beta installs, certain bundled plugins are present under `extensions/*` and can be discovered/enabled by the loader, but their declared package dependencies are not actually resolvable from ... | closed | duplicate | true | 1 | [] | [] | 2026-03-09T09:44:40Z | 2026-04-24T04:25:23Z | 2026-04-22T02:36:48Z | CONTRIBUTOR | null | 20260424T060021Z | 2026-04-24T06:00:21Z | xaeon2026 | 264,572,156 | U_kgDOD8UM_A | User | false |
openclaw/openclaw | 4,197,997,361 | I_kwDOQb6kR876OFsx | 60,020 | https://github.com/openclaw/openclaw/issues/60020 | https://api.github.com/repos/openclaw/openclaw/issues/60020 | Plugin update fails with ERR_MODULE_NOT_FOUND on first openclaw update after new release | ## Summary
Running `openclaw update` after a new release fails to update plugins on the first attempt with:
```
Failed to update dingtalk: Error [ERR_MODULE_NOT_FOUND]: Cannot find module
'/home/tom/.nvm/versions/node/v22.22.1/lib/node_modules/openclaw/dist/install.runtime-BEOb-kNW.js'
imported from '/home/tom/.nvm/v... | closed | duplicate | true | 4 | [] | [] | 2026-04-03T02:58:58Z | 2026-04-24T04:24:53Z | 2026-04-22T02:36:50Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | BrilliantWang | 120,538,413 | U_kgDOBy9FLQ | User | false |
openclaw/openclaw | 4,046,633,890 | I_kwDOQb6kR87xMrui | 41,262 | https://github.com/openclaw/openclaw/issues/41262 | https://api.github.com/repos/openclaw/openclaw/issues/41262 | Extensions dependencies not installed during update | ## Bug
### Description
After upgrading OpenClaw to 2026.3.8, the `diagnostics-otel` extension fails to load because its `node_modules` are missing:
```
[plugins] diagnostics-otel failed to load from /opt/homebrew/lib/node_modules/openclaw/extensions/diagnostics-otel/index.ts:
Error: Cannot find module '@opentelemet... | closed | duplicate | true | 3 | [] | [] | 2026-03-09T16:36:04Z | 2026-04-24T04:25:22Z | 2026-04-22T02:36:52Z | CONTRIBUTOR | null | 20260424T060021Z | 2026-04-24T06:00:21Z | reed1898 | 129,141,816 | U_kgDOB7KMOA | User | false |
openclaw/openclaw | 4,196,868,936 | I_kwDOQb6kR876JyNI | 59,901 | https://github.com/openclaw/openclaw/issues/59901 | https://api.github.com/repos/openclaw/openclaw/issues/59901 | [Bug]: 4.1: Unable to resolve bundled plugin error on image-generation-core/runtime-api.js and telegram module in cron jobs | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
### Description
After a clean install of `openclaw@2026.4.1` (via `npm uninstall -g openclaw && npm install -g openclaw@latest`), several cron jobs are failing with module resolution errors. This appears to be a bug in how b... | closed | duplicate | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-02T20:39:22Z | 2026-04-24T04:25:21Z | 2026-04-22T02:36:54Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | hansolo949 | 217,484,633 | U_kgDODPaNWQ | User | false |
openclaw/openclaw | 4,076,330,509 | I_kwDOQb6kR87y994N | 46,443 | https://github.com/openclaw/openclaw/issues/46443 | https://api.github.com/repos/openclaw/openclaw/issues/46443 | Discord (and other extensions) broken in Docker after send-deps extraction in #46301 | ## Summary
PR #46301 ("Fix configure startup stalls from outbound send-deps imports") extracted `resolveOutboundSendDep` from `src/infra/outbound/deliver.ts` into a new file `src/infra/outbound/send-deps.ts`, and updated extension imports accordingly. This breaks all affected extensions (Discord, Telegram, WhatsApp, S... | closed | duplicate | true | 4 | [] | [] | 2026-03-14T17:39:43Z | 2026-04-24T04:25:20Z | 2026-04-22T02:36:56Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | fakade2 | 46,537,030 | MDQ6VXNlcjQ2NTM3MDMw | User | false |
openclaw/openclaw | 4,087,122,486 | I_kwDOQb6kR87znIo2 | 48,875 | https://github.com/openclaw/openclaw/issues/48875 | https://api.github.com/repos/openclaw/openclaw/issues/48875 | CLI subcommands broken in Docker: ~29 missing external deps | ## Environment
| Field | Value |
|-------|-------|
| Image | `ghcr.io/openclaw/openclaw:main` |
| Version | 2026.3.14 |
| Platform | Docker (OrbStack, arm64) |
## Steps to Reproduce
1. `docker pull ghcr.io/openclaw/openclaw:main`
2. Start container (standard compose)
3. Verify gateway is healthy: `docker exec opencl... | closed | duplicate | true | 4 | [] | [] | 2026-03-17T08:58:32Z | 2026-04-24T04:25:20Z | 2026-04-22T02:36:58Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | bottlerex | 65,446,553 | MDQ6VXNlcjY1NDQ2NTUz | User | false |
openclaw/openclaw | 4,306,397,100 | I_kwDOQb6kR88AAAABAK5nrA | 69,949 | https://github.com/openclaw/openclaw/issues/69949 | https://api.github.com/repos/openclaw/openclaw/issues/69949 | [Bug]: Telegram: streaming assistant replies with MEDIA:<path> deliver as plain sendMessage instead of sendDocument | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
When the assistant emits a reply like:
```
[[reply_to_current]] segue o anexo
MEDIA:/home/node/.openclaw/media/file.pdf
```
the Telegram bot channel deliversit via `sendMessage` (plain text) instead of `sendDoc... | closed | completed | true | 1 | [
"bug"
] | [] | 2026-04-22T02:37:54Z | 2026-04-25T04:17:08Z | 2026-04-22T19:11:33Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | deon7769 | 73,475,350 | MDQ6VXNlcjczNDc1MzUw | User | false |
openclaw/openclaw | 4,306,398,667 | I_kwDOQb6kR88AAAABAK5tyw | 69,950 | https://github.com/openclaw/openclaw/issues/69950 | https://api.github.com/repos/openclaw/openclaw/issues/69950 | [Bug]: 2026.4.20版发升级锁死 | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
> openclaw@2026.4.20 openclaw D:\dalongxia\openclaw
> node scripts/run-node.mjs "doctor"
4.20卡死在这里
D:\dalongxia\openclaw>pnpm openclaw gateway start
> openclaw@2026.4.20 openclaw D:\dalongxia\openclaw
> node scripts/run-... | closed | completed | true | 1 | [
"bug"
] | [] | 2026-04-22T02:38:30Z | 2026-04-28T04:48:27Z | 2026-04-25T19:16:39Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | gg22mm | 6,648,329 | MDQ6VXNlcjY2NDgzMjk= | User | false |
openclaw/openclaw | 4,306,280,013 | I_kwDOQb6kR88AAAABAKyeTQ | 69,939 | https://github.com/openclaw/openclaw/issues/69939 | https://api.github.com/repos/openclaw/openclaw/issues/69939 | Bug: memory-core logs 'cron service unavailable' on every gateway startup (2026.4.20) | ## Bug Description
On every gateway startup (including warm restarts), the memory-core plugin logs a warning:
```
memory-core: managed dreaming cron could not be reconciled (cron service unavailable).
```
This warning fires repeatedly for ~3-5 minutes after startup, then stops. The actual cron job (`memory-core shor... | closed | completed | true | 2 | [] | [] | 2026-04-22T02:01:10Z | 2026-04-24T04:25:15Z | 2026-04-22T03:50:27Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | mlaihk | 25,972,362 | MDQ6VXNlcjI1OTcyMzYy | User | false |
openclaw/openclaw | 4,306,406,043 | I_kwDOQb6kR88AAAABAK6Kmw | 69,951 | https://github.com/openclaw/openclaw/issues/69951 | https://api.github.com/repos/openclaw/openclaw/issues/69951 | [Bug]: `/model gpt54` says switched successfully but main session persists back to MiniMax-M2.7 on 2026.4.20 | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
On OpenClaw `2026.4.20 (115f05d)` in a Feishu-backed main session, sending `/model gpt54` returns a success message (`已切换到 GPT-5.4 ✅`), but the main session does not actually persist the override. Immediately after that, `/m... | closed | completed | true | 1 | [] | [] | 2026-04-22T02:41:23Z | 2026-04-28T04:48:25Z | 2026-04-25T19:16:46Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | wolfzl | 107,912,679 | U_kgDOBm6d5w | User | false |
openclaw/openclaw | 4,306,452,937 | I_kwDOQb6kR88AAAABAK9ByQ | 69,955 | https://github.com/openclaw/openclaw/issues/69955 | https://api.github.com/repos/openclaw/openclaw/issues/69955 | [Bug]: Telegram `/think` inline-button menu ignores per-session `/model` override, always resolved against `agents.defaults.model` | ## Summary
On `openclaw@2026.4.20`, the Telegram `/think` inline-button menu is generated from **`agents.defaults.model`** instead of the **current session's model**, so the menu never reflects per-session `/model` overrides. The hardcoded argument description string (`"off, minimal, low, medium, high, xhigh"`) is als... | closed | completed | false | 1 | [] | [] | 2026-04-22T02:59:10Z | 2026-04-25T09:30:51Z | 2026-04-25T09:30:51Z | NONE | null | 20260425T120014Z | 2026-04-25T12:00:14Z | superandylin | 56,207,675 | MDQ6VXNlcjU2MjA3Njc1 | User | false |
openclaw/openclaw | 4,306,487,828 | I_kwDOQb6kR88AAAABAK_KFA | 69,959 | https://github.com/openclaw/openclaw/issues/69959 | https://api.github.com/repos/openclaw/openclaw/issues/69959 | [Bug]: non-Lark users can still hit hard CLI startup failures from bundled Feishu dependency on 2026.4.20 | ## Description
On a fresh Linux/npm global install of `openclaw@2026.4.20`, the CLI can still fail on startup with a hard module-resolution error for the bundled Feishu/Lark integration even when the operator does not use Lark/Feishu at all.
This is distinct from the older reports about Feishu onboarding itself: the ... | closed | completed | true | 1 | [] | [] | 2026-04-22T03:11:42Z | 2026-04-25T04:17:26Z | 2026-04-22T17:49:38Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | MichealWangYZ | 49,559,224 | MDQ6VXNlcjQ5NTU5MjI0 | User | false |
openclaw/openclaw | 4,158,101,646 | I_kwDOQb6kR87315iO | 56,230 | https://github.com/openclaw/openclaw/issues/56230 | https://api.github.com/repos/openclaw/openclaw/issues/56230 | Telegram polling lacks per-token dedup guard — external scripts and hot-reload can silently create duplicate pollers | ## The Problem
There's no mechanism in `monitorTelegramProvider()` to detect or prevent duplicate polling sessions for the same bot token. If anything starts a second `getUpdates` call on the same token — whether it's a hot-reload race, an external script, or a health monitor probe — the gateway enters a permanent 409... | closed | completed | false | 2 | [] | [] | 2026-03-28T06:34:02Z | 2026-04-25T08:39:24Z | 2026-04-25T08:39:24Z | NONE | null | 20260425T090015Z | 2026-04-25T09:00:15Z | Co-Messi | 87,403,883 | MDQ6VXNlcjg3NDAzODgz | User | false |
openclaw/openclaw | 4,306,539,729 | I_kwDOQb6kR88AAAABALCU0Q | 69,966 | https://github.com/openclaw/openclaw/issues/69966 | https://api.github.com/repos/openclaw/openclaw/issues/69966 | Feature request: per-agent bootstrap profile support | ## Problem
Currently, the Profile settings in the Control UI (and the equivalent `agents.defaults` config) apply globally to ALL agents in a multi-agent setup. The profiles adjust three key parameters:
- `contextInjection`
- `bootstrapMaxChars`
- `bootstrapTotalMaxChars`
For teams with heterogeneous agents — e.g., a... | open | null | false | 1 | [] | [] | 2026-04-22T03:31:03Z | 2026-05-02T12:37:17Z | null | NONE | null | 20260502T150030Z | 2026-05-02T15:00:30Z | stayif | 7,867,389 | MDQ6VXNlcjc4NjczODk= | User | false |
openclaw/openclaw | 4,306,498,991 | I_kwDOQb6kR88AAAABAK_1rw | 69,960 | https://github.com/openclaw/openclaw/issues/69960 | https://api.github.com/repos/openclaw/openclaw/issues/69960 | [Bug]: plugins install --profile X writes to default extensions dir instead of profile's state dir | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Version: v2026.4.20, Node 22 LTS, Ubuntu 24.04
Repro: openclaw --profile ledger plugins install <path> installs to ~/.openclaw/extensions/ but the ledger profile's gateway (with OPENCLAW_STATE_DIR=~/.openclaw-ledger) scans ~... | closed | not_planned | true | 2 | [
"bug"
] | [] | 2026-04-22T03:15:39Z | 2026-04-29T04:42:19Z | 2026-04-26T07:10:31Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | FrancisLyman | 58,153,088 | MDQ6VXNlcjU4MTUzMDg4 | User | false |
openclaw/openclaw | 3,999,065,086 | I_kwDOQb6kR87uXOP- | 28,300 | https://github.com/openclaw/openclaw/issues/28300 | https://api.github.com/repos/openclaw/openclaw/issues/28300 | Theme Customization System — Preset Themes + Custom Theme Studio | ### Summary
Add a theme customization system to the Control UI: 6 curated preset themes + a Custom Theme Studio with hue-based color generation, dark/light mode, gradient accents, and per-variable color overrides.
This addresses the feedback from #28048 — instead of fixed themes, users can customize their own colors.... | open | null | false | 5 | [
"enhancement"
] | [
"BunsDev"
] | 2026-02-27T04:47:30Z | 2026-05-05T02:56:15Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | xingzihai | 87,858,694 | MDQ6VXNlcjg3ODU4Njk0 | User | false |
openclaw/openclaw | 4,306,608,499 | I_kwDOQb6kR88AAAABALGhcw | 69,968 | https://github.com/openclaw/openclaw/issues/69968 | https://api.github.com/repos/openclaw/openclaw/issues/69968 | Bug: heartbeat exec-completion prompt references phantom system messages — agent always replies "I don't see the output" | ## Summary
When an async exec command completes and triggers a heartbeat, the agent consistently replies with a variant of `"I don't see any async command output in the system messages above"`. This is a bug in `buildExecEventPrompt()` — the actual exec output is never injected into the prompt.
## Code trace
**File:... | closed | completed | true | 0 | [] | [] | 2026-04-22T03:57:14Z | 2026-04-27T04:35:00Z | 2026-04-25T03:51:31Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | pagem7632-hcm | 269,862,179 | U_kgDOEBXFIw | User | false |
openclaw/openclaw | 4,306,617,089 | I_kwDOQb6kR88AAAABALHDAQ | 69,970 | https://github.com/openclaw/openclaw/issues/69970 | https://api.github.com/repos/openclaw/openclaw/issues/69970 | [Bug]: Windows auto-update restart script hangs indefinitely on `schtasks /End`, leaves zombie cmd.exe and flashing Terminal window | ## Environment
- **OpenClaw version**: 2026.4.20 (auto-updated from prior stable)
- **OS**: Windows 11 (Build 26200+)
- **Node.js**: System-installed (`C:\Program Files\nodejs\node.exe`)
- **Gateway port**: 18789
- **Scheduled Task**: `OpenClaw Gateway` (At logon, runs `gateway.cmd`)
- **Auto-update config**: `update.... | closed | completed | true | 2 | [] | [] | 2026-04-22T04:00:34Z | 2026-04-29T04:42:46Z | 2026-04-26T06:57:11Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | yangswld | 43,212,671 | MDQ6VXNlcjQzMjEyNjcx | User | false |
openclaw/openclaw | 4,306,645,262 | I_kwDOQb6kR88AAAABALIxDg | 69,973 | https://github.com/openclaw/openclaw/issues/69973 | https://api.github.com/repos/openclaw/openclaw/issues/69973 | [Bug]: claude-cli fallback turn loses prior session context even though Claude session file still has full history | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When using `claude-cli/claude-sonnet-4-6`, a turn that falls back away from Claude can lose all prior context for the fallback model, even though the original Claude local session file still contains the fuller... | closed | completed | true | 3 | [
"bug"
] | [] | 2026-04-22T04:10:11Z | 2026-04-29T04:42:18Z | 2026-04-26T07:10:36Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | aounakram | 57,077,179 | MDQ6VXNlcjU3MDc3MTc5 | User | false |
openclaw/openclaw | 4,306,669,497 | I_kwDOQb6kR88AAAABALKPuQ | 69,977 | https://github.com/openclaw/openclaw/issues/69977 | https://api.github.com/repos/openclaw/openclaw/issues/69977 | Gateway RSS grows to 900MB+ rapidly on macOS MacBook Air after fresh restart | ## Bug Report: Gateway Memory Bloat on macOS
### Environment
- **OpenClaw version:** 2026.4.20 (upgraded from 2026.4.15)
- **macOS:** Darwin 24.6.0 (macOS 26 on MacBook Air)
- **RAM:** 8GB MacBook Air
- **Node:** v22.22.0
### Problem
The openclaw-gateway process RSS grows to 900MB+ within minutes of a fresh restart, ... | closed | not_planned | true | 1 | [] | [] | 2026-04-22T04:19:33Z | 2026-04-29T04:42:17Z | 2026-04-26T07:10:41Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | GenieWeenie | 152,430,369 | U_kgDOCRXnIQ | User | false |
openclaw/openclaw | 4,306,670,619 | I_kwDOQb6kR88AAAABALKUGw | 69,978 | https://github.com/openclaw/openclaw/issues/69978 | https://api.github.com/repos/openclaw/openclaw/issues/69978 | Feature Request: Add /watchdog command and suppress duplicate watchdog warnings per run | ## Summary
The TUI streaming watchdog serves an important purpose (detecting silently-dropped runs), but the current UX has two pain points:
1. **Duplicate warnings for the same run** — After the watchdog fires and prints the system message, if the backend continues emitting deltas for the same `runId` (which can hap... | open | null | false | 1 | [] | [] | 2026-04-22T04:19:59Z | 2026-05-02T17:55:12Z | null | CONTRIBUTOR | null | 20260502T180029Z | 2026-05-02T18:00:29Z | JIRBOY | 49,422,297 | MDQ6VXNlcjQ5NDIyMjk3 | User | false |
openclaw/openclaw | 4,306,647,659 | I_kwDOQb6kR88AAAABALI6aw | 69,974 | https://github.com/openclaw/openclaw/issues/69974 | https://api.github.com/repos/openclaw/openclaw/issues/69974 | [Bug]: tools: [] in createAgentSession disables all tools including customTools after pi 0.68.1 upgrade | # Beta blocker: openclaw — `createAgentSession({ tools: [] })` silently disables all tools in pi 0.68.1
## Bug type
Regression (worked before, now fails)
## Beta release blocker
Yes
## Summary
After upgrading `@mariozechner/pi-*` packages from `0.67.68` to `0.68.1` (commit `0f4dc42767`), all tool calls are silent... | closed | completed | true | 2 | [
"bug"
] | [] | 2026-04-22T04:11:05Z | 2026-04-25T04:16:29Z | 2026-04-22T21:44:34Z | CONTRIBUTOR | null | 20260425T060021Z | 2026-04-25T06:00:21Z | cxyhhhhh | 49,286,167 | MDQ6VXNlcjQ5Mjg2MTY3 | User | false |
openclaw/openclaw | 4,037,380,622 | I_kwDOQb6kR87wpYoO | 38,580 | https://github.com/openclaw/openclaw/issues/38580 | https://api.github.com/repos/openclaw/openclaw/issues/38580 | Feature request: Read pinned Telegram messages via bot API | ## Summary
Allow the OpenClaw assistant to fetch pinned messages from a Telegram chat on demand.
## Use case
Pinned messages are a natural way to pass persistent context across sessions — project status, standing instructions, key links. Currently the bot only receives messages sent directly to it and cannot access pi... | open | null | false | 3 | [] | [] | 2026-03-07T04:56:53Z | 2026-05-04T10:02:49Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | therickod-promeus | 259,231,431 | U_kgDOD3OOxw | User | false |
openclaw/openclaw | 4,037,381,993 | I_kwDOQb6kR87wpY9p | 38,581 | https://github.com/openclaw/openclaw/issues/38581 | https://api.github.com/repos/openclaw/openclaw/issues/38581 | [Bug]: Gateway forwards raw provider error JSON to Discord instead of summarizing | ## Summary
When an LLM provider returns an error payload (e.g. `server_error`), the gateway currently forwards the *raw provider JSON blob* through to Discord as message content (e.g. `Codex error: {...json...}`), instead of masking/summarizing it.
This is noisy/hard-to-read for users and exposes internal/low-level fi... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-07T04:57:27Z | 2026-04-27T05:01:20Z | 2026-04-24T19:37:44Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | dennisw3255 | 255,210,905 | U_kgDODzY1mQ | User | false |
openclaw/openclaw | 4,037,385,083 | I_kwDOQb6kR87wpZt7 | 38,584 | https://github.com/openclaw/openclaw/issues/38584 | https://api.github.com/repos/openclaw/openclaw/issues/38584 | [Control UI] Streaming responses create multiple message bubbles instead of updating single bubble | ### Bug type
Regression (worked before, now fails)
### Summary
In Control UI, streaming responses with thinking/reasoning content display each chunk as a separate message bubble instead of accumulating in a single bubble. This fragments the thinking process across multiple bubbles.
### Steps to reproduce
1. Open C... | closed | not_planned | true | 2 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-07T04:58:41Z | 2026-04-27T05:01:19Z | 2026-04-24T19:37:49Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | carlos999-hqsama | 245,906,921 | U_kgDODqg96Q | User | false |
openclaw/openclaw | 4,037,394,738 | I_kwDOQb6kR87wpcEy | 38,587 | https://github.com/openclaw/openclaw/issues/38587 | https://api.github.com/repos/openclaw/openclaw/issues/38587 | [Bug]: Dashboard/Control-UI 无法打开 - 页面返回 404 Not Found | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
问题描述
OpenClaw 版本：2026.3.2 (85377a2)
操作系统：Windows 11
安装方式：pnpm 全局安装
通过飞书发送消息功能正常，但无法打开控制面板 Dashboard。
### Steps to reproduce
复现步骤
安装 OpenClaw：pnpm add -g openclaw
启动 gateway：openclaw gateway start
浏览器访问：http://127.0.0.1:18789/
页面返回：404 N... | closed | completed | true | 4 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-07T05:02:54Z | 2026-04-27T05:01:19Z | 2026-04-24T19:37:53Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | 490912111 | 171,170,599 | U_kgDOCjPbJw | User | false |
openclaw/openclaw | 4,037,405,932 | I_kwDOQb6kR87wpezs | 38,590 | https://github.com/openclaw/openclaw/issues/38590 | https://api.github.com/repos/openclaw/openclaw/issues/38590 | Feature request: macOS Keychain support for API key storage | ## Feature Request
### Summary
Add support for storing API keys (Anthropic, Google, OpenRouter, etc.) in the macOS Keychain instead of plaintext in `openclaw.json` and `auth-profiles.json`.
### Problem
Currently, all API keys and tokens are stored as plaintext in:
- `~/.openclaw/openclaw.json`
- `~/.openclaw/agents/<... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T05:08:58Z | 2026-04-28T04:43:23Z | 2026-04-26T01:45:17Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | EvilRebel | 25,450,463 | MDQ6VXNlcjI1NDUwNDYz | User | false |
openclaw/openclaw | 4,037,442,367 | I_kwDOQb6kR87wpns_ | 38,596 | https://github.com/openclaw/openclaw/issues/38596 | https://api.github.com/repos/openclaw/openclaw/issues/38596 | Discord: health monitor restart loop — post-connection zombie sessions evade circuit breakers | ## Problem
Discord connections successfully complete the handshake (HELLO → READY/RESUMED), run for a period, then become zombie/unstable. The health monitor detects this ("stuck" / "stale-socket") and restarts the provider, but the cycle repeats indefinitely — creating an effective infinite restart loop where **messa... | closed | completed | true | 4 | [] | [] | 2026-03-07T05:22:59Z | 2026-04-29T04:34:47Z | 2026-04-26T18:09:05Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | jeanmonet | 15,112,920 | MDQ6VXNlcjE1MTEyOTIw | User | false |
openclaw/openclaw | 4,037,445,222 | I_kwDOQb6kR87wpoZm | 38,597 | https://github.com/openclaw/openclaw/issues/38597 | https://api.github.com/repos/openclaw/openclaw/issues/38597 | CJK text splitting breaks mid-character at Discord 2000-char boundary | ## Problem
When a response exceeds Discord's 2000-character limit, `chunkDiscordText()` splits the message at the nearest whitespace boundary. However, CJK text (Chinese, Japanese, Korean) has no word-separating spaces, so the split falls at an arbitrary character position, often breaking mid-word or mid-sentence.
##... | closed | completed | true | 7 | [] | [] | 2026-03-07T05:24:14Z | 2026-05-02T04:28:45Z | 2026-04-29T09:08:01Z | CONTRIBUTOR | null | 20260502T060142Z | 2026-05-02T06:01:42Z | p3nchan | 5,032,148 | MDQ6VXNlcjUwMzIxNDg= | User | false |
openclaw/openclaw | 4,037,452,704 | I_kwDOQb6kR87wpqOg | 38,601 | https://github.com/openclaw/openclaw/issues/38601 | https://api.github.com/repos/openclaw/openclaw/issues/38601 | [Bug] PTY silently disabled in sandbox mode without warning when pty:true is set | ## Description
When a user sets `pty: true` on the `exec` tool but the session is running in sandbox mode, PTY is silently disabled without any warning to the user.
## Code Evidence
In `dist/pi-embedded-DgYXShcG.js` (and other builds):
```javascript
const usePty = params.pty === true && !sandbox;
```
The PTY is fo... | open | null | false | 2 | [] | [] | 2026-03-07T05:27:30Z | 2026-05-04T06:15:30Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | Jackey0903 | 97,658,517 | U_kgDOBdImlQ | User | false |
openclaw/openclaw | 4,037,456,607 | I_kwDOQb6kR87wprLf | 38,603 | https://github.com/openclaw/openclaw/issues/38603 | https://api.github.com/repos/openclaw/openclaw/issues/38603 | Telegram: sendMessageTelegram missing isSilentReply guard (causes flash of streamed text before NO_REPLY suppression) | ## Bug Report
### Summary
The `sendMessageTelegram` adapter is missing the `isSilentReplyText` guard that exists in `sendMessageSlack`. While this isn't directly responsible for the flash bug, it's a defensive-coding gap that should be addressed.
The actual flash bug occurs when the agent outputs visible narration/t... | open | null | false | 4 | [] | [] | 2026-03-07T05:29:30Z | 2026-05-04T09:03:13Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | susanadmin | 266,227,360 | U_kgDOD95OoA | User | false |
openclaw/openclaw | 4,037,462,424 | I_kwDOQb6kR87wpsmY | 38,604 | https://github.com/openclaw/openclaw/issues/38604 | https://api.github.com/repos/openclaw/openclaw/issues/38604 | [Security] Sandbox containers have no default pidsLimit - vulnerable to fork bomb attacks | ## Summary
Sandbox containers are created without a default `pidsLimit`, making them vulnerable to fork bomb attacks that can exhaust host resources.
## Evidence
In `dist/sandbox-BTk3jOUP.js`:
```javascript
return {
image: agentDocker?.image ?? globalDocker?.image ?? DEFAULT_SANDBOX_IMAGE,
// ... other defa... | open | null | false | 3 | [] | [] | 2026-03-07T05:32:04Z | 2026-05-04T09:03:15Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | Jackey0903 | 97,658,517 | U_kgDOBdImlQ | User | false |
openclaw/openclaw | 4,037,474,163 | I_kwDOQb6kR87wpvdz | 38,607 | https://github.com/openclaw/openclaw/issues/38607 | https://api.github.com/repos/openclaw/openclaw/issues/38607 | Discord: CJK text splits at wrong character boundaries when message exceeds 2000 chars | ## Problem
When a Discord message exceeds 2000 characters (Discord's hard API limit), OpenClaw's `chunkDiscordText()` function splits the message using whitespace (`/\s/`) as word boundary detection. This works for Latin text, but **CJK languages (Chinese, Japanese, Korean) have no whitespace between words**, so the s... | closed | not_planned | true | 5 | [] | [] | 2026-03-07T05:38:12Z | 2026-04-28T04:42:43Z | 2026-04-26T02:41:59Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | p3nchan | 5,032,148 | MDQ6VXNlcjUwMzIxNDg= | User | false |
openclaw/openclaw | 4,037,525,593 | I_kwDOQb6kR87wp8BZ | 38,622 | https://github.com/openclaw/openclaw/issues/38622 | https://api.github.com/repos/openclaw/openclaw/issues/38622 | Workspace file injector does not follow symlinks | ## Describe the bug
When a workspace file (e.g. `SOUL.md`) is a symlink to a shared file in another workspace directory, OpenClaw's workspace file injection shows it as `[MISSING]` rather than following the symlink and loading the content.
## To reproduce
1. Create `workspace-shared/SOUL.md` with content
2. In anoth... | open | null | false | 3 | [] | [] | 2026-03-07T05:59:38Z | 2026-05-04T06:15:31Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | shizenchan | 264,457,523 | U_kgDOD8NNMw | User | false |
openclaw/openclaw | 4,037,532,663 | I_kwDOQb6kR87wp9v3 | 38,623 | https://github.com/openclaw/openclaw/issues/38623 | https://api.github.com/repos/openclaw/openclaw/issues/38623 | Feature Request: Video Input Support for Multimodal Models (e.g., Qwen3.5-VL) | ## 🎬 Feature Request: Video Input Support
### Background
Several modern multimodal models now natively support video input, including:
- **Qwen3.5 / Qwen3.5-VL** (Alibaba): Supports video analysis up to several minutes
- **GPT-4o**: Video understanding capabilities
- **Claude 3.5 Sonnet**: Video frame analysis
How... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-07T06:02:17Z | 2026-04-28T04:43:22Z | 2026-04-26T01:45:41Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | LobayClaw | 261,543,597 | U_kgDOD5bWrQ | User | false |
openclaw/openclaw | 4,037,542,647 | I_kwDOQb6kR87wqAL3 | 38,625 | https://github.com/openclaw/openclaw/issues/38625 | https://api.github.com/repos/openclaw/openclaw/issues/38625 | feat: version field in health response + POST /admin/update endpoint | ## Context
This request comes from [jhgaylor/crab](https://github.com/jhgaylor/crab/issues/41), a fleet monitoring tool for OpenClaw deployments. Two gateway API additions would unlock version display and update orchestration from Crab without requiring SSH credentials.
---
## Request 1 — `version` in `GET /health` ... | open | null | false | 2 | [] | [] | 2026-03-07T06:05:23Z | 2026-05-04T06:04:17Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | agent-five | 264,713,178 | U_kgDOD8cz2g | User | false |
openclaw/openclaw | 4,037,555,796 | I_kwDOQb6kR87wqDZU | 38,626 | https://github.com/openclaw/openclaw/issues/38626 | https://api.github.com/repos/openclaw/openclaw/issues/38626 | Subagent lifecycle observability + async supervision controls | ## Problem\nOperators need deterministic visibility/control for async subagent workflows (timeline, errors, artifacts, supervision actions).\n\n## Scope\n- Emit per-subagent lifecycle events: spawn, queued/running, tool start/end, warning/error, completion\n- Persist run summary with artifact pointers (logs/results/pat... | open | null | false | 2 | [] | [] | 2026-03-07T06:09:41Z | 2026-05-04T10:02:50Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | flynnbot-ryan[bot] | 262,420,930 | BOT_kgDOD6Q5wg | Bot | false |
openclaw/openclaw | 4,037,581,539 | I_kwDOQb6kR87wqJrj | 38,632 | https://github.com/openclaw/openclaw/issues/38632 | https://api.github.com/repos/openclaw/openclaw/issues/38632 | [Bug] process.chdir() race condition causes nested workspace directory creation | ## Summary
When multiple agent runs execute concurrently (e.g., heartbeat + user message), `process.chdir()` calls in `attempt.js` and `compact.js` create a race condition that corrupts the global cwd, causing workspace bootstrap to create nested `data/agents/<id>/workspace/` directories inside the existing workspace.... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-07T06:20:23Z | 2026-04-27T05:01:18Z | 2026-04-24T19:38:17Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Zenwh | 129,019,575 | U_kgDOB7Cutw | User | false |
openclaw/openclaw | 4,037,581,569 | I_kwDOQb6kR87wqJsB | 38,633 | https://github.com/openclaw/openclaw/issues/38633 | https://api.github.com/repos/openclaw/openclaw/issues/38633 | [Docs] 文档中 pidsLimit 示例值为 256，但代码无默认值，容易误导用户 | ## 问题描述
在分析沙箱安全配置时，我发现文档中的 `pidsLimit` 示例值与代码实现不一致：
### 文档示例 (docs/install/docker.md:623)
```json
pidsLimit: 256,
```
### 代码实现 (src/agents/sandbox/config.ts:108)
```typescript
pidsLimit: agentDocker?.pidsLimit ?? globalDocker?.pidsLimit, // 无默认值！
```
## 影响
1. **用户误以为有默认值**：文档示例显示 `pidsLimit: 256`，用户可能认为这是默认值
2. *... | open | null | false | 3 | [] | [] | 2026-03-07T06:20:25Z | 2026-05-04T06:04:18Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | Jackey0903 | 97,658,517 | U_kgDOBdImlQ | User | false |
openclaw/openclaw | 4,037,594,696 | I_kwDOQb6kR87wqM5I | 38,639 | https://github.com/openclaw/openclaw/issues/38639 | https://api.github.com/repos/openclaw/openclaw/issues/38639 | Telegram /models does not surface configured custom providers from models.providers | ## Summary
Custom providers added under `models.providers` can be usable at runtime, but they may not appear in Telegram `/models` provider selection.
In my case, a custom provider `packycode` with model `gpt-5.4` was configured and could be set as the default model, but Telegram `/models` continued to show only built... | closed | completed | true | 3 | [] | [] | 2026-03-07T06:27:07Z | 2026-04-29T04:38:04Z | 2026-04-26T10:02:52Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | booooodv | 35,971,518 | MDQ6VXNlcjM1OTcxNTE4 | User | false |
openclaw/openclaw | 4,037,639,608 | I_kwDOQb6kR87wqX24 | 38,651 | https://github.com/openclaw/openclaw/issues/38651 | https://api.github.com/repos/openclaw/openclaw/issues/38651 | Cron scheduler not loading jobs (nextWakeAtMs=null) | ## Bug Description
Cron scheduler shows `nextWakeAtMs: null` and doesn't execute jobs.
## Environment
- OpenClaw version: 0.1.7
- macOS
- Gateway running but cron module not initializing
## Steps to Reproduce
1. Add cron jobs via `cron add` command
2. Jobs saved to jobs.json correctly
3. Restart gateway
4. Check cron... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-07T06:47:08Z | 2026-04-27T05:01:17Z | 2026-04-24T19:38:35Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | yvhkddd | 45,888,867 | MDQ6VXNlcjQ1ODg4ODY3 | User | false |
openclaw/openclaw | 4,037,689,577 | I_kwDOQb6kR87wqkDp | 38,657 | https://github.com/openclaw/openclaw/issues/38657 | https://api.github.com/repos/openclaw/openclaw/issues/38657 | [Bug] Onboard UI fails to overwrite existing provider configurations | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When attempting to update an existing model provider configuration through the Onboard UI, the changes are not persisted to the configuration files. The UI appears to save successfully, but the underlying JSON files retain the original (pot... | open | null | false | 4 | [
"bug",
"bug:behavior"
] | [] | 2026-03-07T07:04:58Z | 2026-05-04T06:15:34Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | Fa1eon | 212,091,325 | U_kgDODKRBvQ | User | false |
openclaw/openclaw | 4,037,740,103 | I_kwDOQb6kR87wqwZH | 38,663 | https://github.com/openclaw/openclaw/issues/38663 | https://api.github.com/repos/openclaw/openclaw/issues/38663 | [Bug]: Cron isolated jobs don't execute in sandbox despite `sandbox.mode="all"` | ### Bug type
Regression (worked before, now fails)
### Summary
### Description
When a cron job is configured with `sessionTarget: "isolated"` and the agent has `sandbox.mode: "all"`, the job still executes in the Gateway process instead of the sandbox container.
### Current Behavior
- **Manual trigger (via chat)*... | closed | completed | true | 2 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-07T07:26:13Z | 2026-04-27T05:01:16Z | 2026-04-24T19:38:39Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | songchenghao | 4,641,267 | MDQ6VXNlcjQ2NDEyNjc= | User | false |
openclaw/openclaw | 4,037,780,063 | I_kwDOQb6kR87wq6Jf | 38,672 | https://github.com/openclaw/openclaw/issues/38672 | https://api.github.com/repos/openclaw/openclaw/issues/38672 | [Bug]: npm error code 3221225477 | ### Bug type
Regression (worked before, now fails)
### Summary
Installation failed with `npm install -g openclaw@latest`.
### Steps to reproduce
npm error code 3221225477
npm error path C:\Program Files\nodejs\node_global\node_modules\openclaw\node_modules\node-llama-cpp
npm error command failed
npm error comman... | closed | completed | true | 7 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-07T07:44:28Z | 2026-04-27T05:01:15Z | 2026-04-24T19:38:48Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | czx0222 | 94,733,462 | U_kgDOBaWElg | User | false |
openclaw/openclaw | 4,037,783,475 | I_kwDOQb6kR87wq6-z | 38,673 | https://github.com/openclaw/openclaw/issues/38673 | https://api.github.com/repos/openclaw/openclaw/issues/38673 | [Feature]: Per-Agent TTS Configuration | ## Problem Description
Currently, TTS (Text-to-Speech) configuration in OpenClaw is global and shared across all agents. The `messages.tts` configuration in `openclaw.json` applies to all agents and channels.
## Current Limitations
1. **Global TTS Settings**: All agents use the same TTS provider, voice, and settings... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-07T07:46:23Z | 2026-04-28T04:42:58Z | 2026-04-26T02:16:50Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | pata00 | 1,174,163 | MDQ6VXNlcjExNzQxNjM= | User | false |
openclaw/openclaw | 4,037,797,838 | I_kwDOQb6kR87wq-fO | 38,679 | https://github.com/openclaw/openclaw/issues/38679 | https://api.github.com/repos/openclaw/openclaw/issues/38679 | Webchat: use agent identity name in session dropdown instead of raw session keys | ## Current behaviour
The Control UI session dropdown displays raw session keys like `webchat:g-agent-main-main`, `webchat:g-agent-butler-main`, etc. These are opaque and hard to distinguish, especially with 10+ agents.
## Proposed behaviour
Use the agent's `identity.name` and `id` (or role) from `agents.list[]` to r... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T07:54:28Z | 2026-04-27T05:01:14Z | 2026-04-24T19:39:01Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | waynegault | 39,455,578 | MDQ6VXNlcjM5NDU1NTc4 | User | false |
openclaw/openclaw | 4,037,807,105 | I_kwDOQb6kR87wrAwB | 38,682 | https://github.com/openclaw/openclaw/issues/38682 | https://api.github.com/repos/openclaw/openclaw/issues/38682 | Gateway does not auto-recover after VPN reconnection (macOS LaunchAgent may not respawn gateway) | # Draft GitHub Issue
Title: **Gateway does not auto-recover after VPN reconnection (macOS LaunchAgent may not respawn gateway)**
## Summary
When the host VPN disconnects, the OpenClaw gateway can transition from `running` → `stopped` due to repeated outbound failures (Telegram API / WhatsApp). After VPN connectivity ... | closed | completed | true | 4 | [
"stale"
] | [] | 2026-03-07T07:59:00Z | 2026-04-27T05:01:13Z | 2026-04-24T19:39:05Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | LindalyX-Lee | 256,247,180 | U_kgDOD0YFjA | User | false |
openclaw/openclaw | 4,037,847,044 | I_kwDOQb6kR87wrKgE | 38,686 | https://github.com/openclaw/openclaw/issues/38686 | https://api.github.com/repos/openclaw/openclaw/issues/38686 | [Bug]: Feishu Channel Fails to Send Replies Despite Successful Agent Processing | ### Bug type
Regression (worked before, now fails)
### Summary
My Feishu channel in OpenClaw is unable to send any replies back to the user. The system successfully receives messages from Feishu, dispatches them to the correct agent, and the agent processes the request and generates a response (visible in the OpenCl... | closed | completed | true | 2 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-07T08:16:29Z | 2026-04-27T05:01:12Z | 2026-04-24T19:39:09Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | toretangtang | 153,609,078 | U_kgDOCSfjdg | User | false |
openclaw/openclaw | 4,037,905,705 | I_kwDOQb6kR87wrY0p | 38,697 | https://github.com/openclaw/openclaw/issues/38697 | https://api.github.com/repos/openclaw/openclaw/issues/38697 | Feature: log actual model used in cron job state (actualModel field) | ## Problem
When a cron job is configured with `model: openai-codex/gpt-5.2` but the provider fails and OpenClaw falls back to a different model, there's no way to know which model actually executed the job.
Currently `cron.state` includes `lastStatus`, `lastDurationMs`, `lastRunStatus` etc., but not the model that wa... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-07T08:39:47Z | 2026-04-27T05:01:11Z | 2026-04-24T19:39:25Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | logos-taranov | 262,328,755 | U_kgDOD6LRsw | User | false |
openclaw/openclaw | 4,038,008,670 | I_kwDOQb6kR87wrx9e | 38,714 | https://github.com/openclaw/openclaw/issues/38714 | https://api.github.com/repos/openclaw/openclaw/issues/38714 | Request: Add Discord reaction event support to Hooks system | ## Summary
Currently, OpenClaw can detect Discord reactions (via the config option), but these events are not exposed through the Hooks system. This makes it impossible to create automation based on user reactions (e.g., react with ✅ to save a message to memory).
## Current Behavior
- Discord reactions are detected... | open | null | false | 2 | [] | [] | 2026-03-07T09:19:37Z | 2026-05-04T06:15:41Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | windhood-jza | 80,495,136 | MDQ6VXNlcjgwNDk1MTM2 | User | false |
openclaw/openclaw | 4,038,034,877 | I_kwDOQb6kR87wr4W9 | 38,716 | https://github.com/openclaw/openclaw/issues/38716 | https://api.github.com/repos/openclaw/openclaw/issues/38716 | feat(slack): Support Slack RTS (Real-Time Search) API via assistant action tokens | ## Summary
OpenClaw's Slack integration currently doesn't extract or use the `action_token` from `assistant_thread_started` events. This token is required to call Slack's `assistant.search.context` (Real-Time Search) API, which enables workspace-wide search across channels, messages, and files.
## Context
- Slack rece... | open | reopened | false | 3 | [] | [] | 2026-03-07T09:28:57Z | 2026-05-04T06:04:35Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | wkdd7 | 114,310,251 | U_kgDOBtA8aw | User | false |
openclaw/openclaw | 4,038,049,434 | I_kwDOQb6kR87wr76a | 38,719 | https://github.com/openclaw/openclaw/issues/38719 | https://api.github.com/repos/openclaw/openclaw/issues/38719 | Bug: Message ordering conflict when switching models mid-session | ## Description
When switching models using runtime model override (e.g., from /reasoning or /model command), OpenClaw throws "Message ordering conflict - please try again" error.
## Steps to Reproduce
1. Start a session with default model (e.g., llama/Qwen3.5-35B)
2. Send a message to switch to a different model (e.g.... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T09:35:17Z | 2026-04-27T05:01:09Z | 2026-04-24T19:39:39Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | richardng1505 | 118,802,129 | U_kgDOBxTG0Q | User | false |
openclaw/openclaw | 4,038,054,919 | I_kwDOQb6kR87wr9QH | 38,721 | https://github.com/openclaw/openclaw/issues/38721 | https://api.github.com/repos/openclaw/openclaw/issues/38721 | [Bug] gateway shutdown timed out: node report shows active child process handle; leaves dirty shutdown + lane wait | ## Summary
On OpenClaw **2026.3.2** (Ubuntu 24.04, Node 22.22.0), gateway shutdown frequently logs:
- `[gateway] shutdown timed out; exiting without full cleanup`
and is often followed by `lane wait exceeded`.
## Key evidence (Node diagnostic reports)
We enabled Node `--report-on-signal` and captured reports during ... | open | null | false | 3 | [] | [] | 2026-03-07T09:37:34Z | 2026-05-04T06:15:43Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | carlosrenatoy | 196,261,309 | U_kgDOC7K1vQ | User | false |
openclaw/openclaw | 4,038,060,204 | I_kwDOQb6kR87wr-is | 38,723 | https://github.com/openclaw/openclaw/issues/38723 | https://api.github.com/repos/openclaw/openclaw/issues/38723 | WhatsApp: add message delete support | ## Feature Request
**Current behavior:**
The WhatsApp channel plugin currently supports `poll`, `react`, and `send` actions only. There is no way to delete a sent message.
**Desired behavior:**
Add a `delete` action to the WhatsApp message tool, allowing the agent to delete its own previously sent messages (using the... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-07T09:39:36Z | 2026-04-28T04:42:42Z | 2026-04-26T02:42:04Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | nicomiachon | 48,127,448 | MDQ6VXNlcjQ4MTI3NDQ4 | User | false |
openclaw/openclaw | 4,038,070,256 | I_kwDOQb6kR87wsA_w | 38,726 | https://github.com/openclaw/openclaw/issues/38726 | https://api.github.com/repos/openclaw/openclaw/issues/38726 | Feature: Show model usage (incl. local models) in OpenClaw dashboard | # Feature: Show model usage (incl. local models) in OpenClaw dashboard
## Problem
Users currently rely on provider dashboards (OpenAI/OpenRouter/Claude) or external tools (CodexBar) to see usage. Local model runs (e.g., Ollama) have no visibility. There is no unified, first‑party “usage (not cost)” view in the OpenCla... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-07T09:43:59Z | 2026-04-27T05:01:09Z | 2026-04-24T19:39:50Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | TylonHH | 18,073,299 | MDQ6VXNlcjE4MDczMjk5 | User | false |
openclaw/openclaw | 4,038,071,626 | I_kwDOQb6kR87wsBVK | 38,728 | https://github.com/openclaw/openclaw/issues/38728 | https://api.github.com/repos/openclaw/openclaw/issues/38728 | Feishu TTS audio message fails - needs file_type=opus instead of mp3 | ## Bug Description
When sending TTS audio messages to Feishu, the OpenClaw Feishu plugin uploads audio with file_type=mp3, but Feishu API requires file_type=opus for audio messages.
## Steps to Reproduce
1. Configure ElevenLabs TTS in OpenClaw
2. Send a message that triggers TTS response
3. Error 400: "Invalid request... | closed | completed | true | 1 | [
"stale"
] | [] | 2026-03-07T09:44:38Z | 2026-04-26T04:30:22Z | 2026-04-23T13:58:19Z | NONE | null | 20260426T060034Z | 2026-04-26T06:00:34Z | liuy-byte | 46,175,527 | MDQ6VXNlcjQ2MTc1NTI3 | User | false |
openclaw/openclaw | 4,038,072,213 | I_kwDOQb6kR87wsBeV | 38,730 | https://github.com/openclaw/openclaw/issues/38730 | https://api.github.com/repos/openclaw/openclaw/issues/38730 | Cron session accumulation can silently exhaust agent context window — no guardrails or warnings | # Draft GitHub Issue
Title: **Cron session accumulation can silently exhaust agent context window — no guardrails or warnings**
## Problem
High-frequency OpenClaw cron jobs (e.g. `every 15m` / `every 30m`) continuously create isolated run sessions and grow the session store without visible guardrails. In a real incid... | open | null | false | 2 | [] | [] | 2026-03-07T09:44:59Z | 2026-05-04T06:15:44Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | LindalyX-Lee | 256,247,180 | U_kgDOD0YFjA | User | false |
openclaw/openclaw | 4,038,084,981 | I_kwDOQb6kR87wsEl1 | 38,737 | https://github.com/openclaw/openclaw/issues/38737 | https://api.github.com/repos/openclaw/openclaw/issues/38737 | Feature: gateway-level auto-escalation from specialist agents to coordinator with synchronous handoff | ## Summary
In a multi-agent setup with specialist channels (for example: stock/dev/ops/content) plus a coordinator agent (`main`), prompt-only routing is not reliable enough.
If a specialist session receives a task that is:
- clearly **outside its domain**, or
- **composite** across multiple domains,
the gateway sho... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-07T09:49:30Z | 2026-04-28T04:42:41Z | 2026-04-26T02:42:08Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | skernelx | 250,769,487 | U_kgDODvJwTw | User | false |
openclaw/openclaw | 4,038,124,214 | I_kwDOQb6kR87wsOK2 | 38,745 | https://github.com/openclaw/openclaw/issues/38745 | https://api.github.com/repos/openclaw/openclaw/issues/38745 | Telegram: model list buttons broken for providers with dots in name (e.g. hf.co) | ## Bug
When using `/models` in a Telegram chat, providers with dots in their ID (e.g. `hf.co`) display models as **plain text** instead of inline keyboard buttons. Providers without dots (e.g. `openai-codex`) display correctly as buttons.
## Root cause
In `src/telegram/model-buttons.ts`, `parseModelCallbackData()` u... | open | null | false | 2 | [] | [] | 2026-03-07T10:03:01Z | 2026-05-04T06:04:43Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | justkekalot | 200,369,881 | U_kgDOC_Fm2Q | User | false |
openclaw/openclaw | 4,038,148,083 | I_kwDOQb6kR87wsT_z | 38,755 | https://github.com/openclaw/openclaw/issues/38755 | https://api.github.com/repos/openclaw/openclaw/issues/38755 | cron.run false failure: tool timeout at 60s while gateway later returns success | ## Summary
`cron.run` can report a hard failure (`gateway timeout after 60000ms`) even when the run actually completes successfully shortly after.
This creates false incident noise (`⚠️ Cron failed`) in production operations.
## Environment
- OpenClaw 2026.3.2
- Windows 11
- Local gateway: `ws://127.0.0.1:18789`
## ... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T10:14:56Z | 2026-04-27T05:01:08Z | 2026-04-24T19:39:54Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | amibars | 149,697,253 | U_kgDOCOwy5Q | User | false |
openclaw/openclaw | 4,306,726,610 | I_kwDOQb6kR88AAAABALNu0g | 69,980 | https://github.com/openclaw/openclaw/issues/69980 | https://api.github.com/repos/openclaw/openclaw/issues/69980 | [Bug] --thinking flag doesn't enable thinking for MiniMax via Anthropic API | # Issue: `--thinking` flag doesn't enable thinking for MiniMax models via Anthropic API
## Problem Description
The `--thinking <level>` flag in cron jobs sets the `thinking` field in the payload, but the HTTP API layer checks the `reasoning` field instead. This causes thinking to never be enabled for MiniMax models (... | closed | not_planned | true | 1 | [] | [] | 2026-04-22T04:39:47Z | 2026-04-28T04:48:24Z | 2026-04-25T19:17:00Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | saulyuhaozhang-web | 271,183,012 | U_kgDOECnspA | User | false |
openclaw/openclaw | 4,038,157,172 | I_kwDOQb6kR87wsWN0 | 38,759 | https://github.com/openclaw/openclaw/issues/38759 | https://api.github.com/repos/openclaw/openclaw/issues/38759 | Add support for GPT-5.4 (released 2026-03-05) | ## Description
OpenAI released GPT-5.4 API on March 5, 2026 (2 days ago), but OpenClaw currently only supports up to GPT-5.3-codex.
## Current Status
```bash
$ openclaw models list --provider openai-codex
openai-codex/gpt-5.3-codex text+image 266k no yes default
```
## Expected Behavior
Op... | closed | completed | true | 4 | [
"stale"
] | [] | 2026-03-07T10:21:05Z | 2026-04-26T04:30:00Z | 2026-04-23T19:36:44Z | NONE | null | 20260426T060034Z | 2026-04-26T06:00:34Z | Allenyang007 | 169,037,251 | U_kgDOChNNww | User | false |
openclaw/openclaw | 4,038,159,056 | I_kwDOQb6kR87wsWrQ | 38,762 | https://github.com/openclaw/openclaw/issues/38762 | https://api.github.com/repos/openclaw/openclaw/issues/38762 | browser act 的 ref 参数位置处理不一致，错误信息误导 | ### 问题
当调用 `browser action=act` 时，`ref` 参数的位置处理不一致：
| 调用方式 | 结果 |
|---------|------|
| `request={"kind": "click", "ref": "e146"}` (ref 在 request 内) | ✅ 成功 |
| `ref="e146", request={"kind": "click"}` (ref 在顶层) | ❌ 错误 |
### 错误信息（误导性）
```
Can't reach the OpenClaw browser control service. Restart the OpenClaw gateway (... | open | null | false | 2 | [] | [] | 2026-03-07T10:22:28Z | 2026-05-04T06:15:45Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | Lumos-789 | 57,990,306 | MDQ6VXNlcjU3OTkwMzA2 | User | false |
openclaw/openclaw | 4,038,178,795 | I_kwDOQb6kR87wsbfr | 38,771 | https://github.com/openclaw/openclaw/issues/38771 | https://api.github.com/repos/openclaw/openclaw/issues/38771 | [Bug]: Docker build fails due to invalid node:22-bookworm digest | ### Bug type
Regression (worked before, now fails)
### Summary
Running the provided docker-setup.sh script fails during the Docker image build step because the Dockerfile references a pinned digest for node:22-bookworm that cannot be resolved by Docker.
The build fails during metadata resolution for the base image.... | closed | completed | true | 2 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-07T10:31:07Z | 2026-04-27T05:01:07Z | 2026-04-24T19:40:06Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Issacwww | 15,892,569 | MDQ6VXNlcjE1ODkyNTY5 | User | false |
openclaw/openclaw | 4,038,194,883 | I_kwDOQb6kR87wsfbD | 38,775 | https://github.com/openclaw/openclaw/issues/38775 | https://api.github.com/repos/openclaw/openclaw/issues/38775 | [Feature Request] Usage-aware load balancing for multi-account auth profiles | ## Problem
The current multi-account auth rotation is **reactive** — it only switches accounts *after* hitting a rate limit or billing error, then applies a fixed cooldown timer. This leads to:
1. **Wasted capacity**: Account A hits its limit → cooldown → all traffic goes to Account B → Account B hits its limit → bot... | closed | not_planned | false | 2 | [] | [] | 2026-03-07T10:38:14Z | 2026-05-03T22:32:13Z | 2026-05-03T22:32:13Z | NONE | null | 20260504T000036Z | 2026-05-04T00:00:36Z | EdwardOptimization | 248,197,022 | U_kgDODssvng | User | false |
openclaw/openclaw | 4,038,217,541 | I_kwDOQb6kR87wsk9F | 38,778 | https://github.com/openclaw/openclaw/issues/38778 | https://api.github.com/repos/openclaw/openclaw/issues/38778 | [Bug]: 飞书机器人@飞书机器人 场景下，被@的机器人不会回应 | ### Bug type
Regression (worked before, now fails)
### Summary
[Bug]: 飞书机器人@飞书机器人 场景下，被@的机器人不会回应
### Steps to reproduce
机器人a通过 <at user_id="xxx">机器人b </at> @机器人b时，机器人b 不回应
### Expected behavior
希望机器B 回应机器人A
### Actual behavior
希望机器B不回应机器人A
### OpenClaw version
2026.3.3
### Operating system
ubuntu
### ... | closed | not_planned | true | 5 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-07T10:51:47Z | 2026-04-27T05:01:06Z | 2026-04-24T19:40:17Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | GuanZhuiGuo | 39,402,145 | MDQ6VXNlcjM5NDAyMTQ1 | User | false |
openclaw/openclaw | 4,038,222,794 | I_kwDOQb6kR87wsmPK | 38,782 | https://github.com/openclaw/openclaw/issues/38782 | https://api.github.com/repos/openclaw/openclaw/issues/38782 | Bug: /new and /reset startup greeting uses wrong time-of-day | ## Bug: `/new` / `/reset` startup greeting can use wrong time-of-day (morning/evening)
### Summary
When a new session is started via `/new` or `/reset`, the assistant often greets with the wrong time-of-day (for example "Good evening" in the morning).
This is caused by missing explicit current-time context in the res... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-07T10:55:23Z | 2026-04-27T05:01:05Z | 2026-04-24T19:40:21Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | matthiasroder | 9,212,056 | MDQ6VXNlcjkyMTIwNTY= | User | false |
openclaw/openclaw | 4,038,235,246 | I_kwDOQb6kR87wspRu | 38,784 | https://github.com/openclaw/openclaw/issues/38784 | https://api.github.com/repos/openclaw/openclaw/issues/38784 | [Bug]: Azure OpenAI models report 0 context tokens — supportsUsageInStreaming forcefully disabled | ### Bug type
Regression (worked before, now fails)
### Summary
`normalizeModelCompat()` forces `supportsUsageInStreaming: false` for all non-`api.openai.com` endpoints, including Azure OpenAI (`*.openai.azure.com`). This prevents pi-ai from sending `stream_options: { include_usage: true }`, so Azure never returns us... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T11:03:52Z | 2026-04-28T04:43:20Z | 2026-04-26T01:47:11Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Hua688 | 5,764,877 | MDQ6VXNlcjU3NjQ4Nzc= | User | false |
openclaw/openclaw | 4,038,246,757 | I_kwDOQb6kR87wssFl | 38,791 | https://github.com/openclaw/openclaw/issues/38791 | https://api.github.com/repos/openclaw/openclaw/issues/38791 | Cron: isolated job delivery defaults to 'none' instead of 'announce' | ## Bug Description
When creating an isolated cron job, the `delivery.mode` defaults to `"none"` instead of `"announce"` as documented.
## Expected Behavior
From the [cron-jobs docs](https://docs.openclaw.ai/automation/cron-jobs):
> Default behavior: if `delivery` is omitted, isolated jobs announce a summary (`deliv... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T11:12:02Z | 2026-04-27T05:01:05Z | 2026-04-24T19:40:25Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | imwyvern | 1,765,672 | MDQ6VXNlcjE3NjU2NzI= | User | false |
openclaw/openclaw | 4,038,249,490 | I_kwDOQb6kR87wsswS | 38,792 | https://github.com/openclaw/openclaw/issues/38792 | https://api.github.com/repos/openclaw/openclaw/issues/38792 | Agent loop silently stalls: no user response after API error + long tool-call chains | ## Bug Description
When the agent enters a long chain of tool-only turns (no text response to the user) and hits an API error mid-chain, OpenClaw silently retries and continues the tool-call loop. The user receives no visible response and no error notification.
## Timeline (from session log)
1. **02:04** — User send... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T11:13:24Z | 2026-04-28T04:43:20Z | 2026-04-26T01:47:21Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | imwyvern | 1,765,672 | MDQ6VXNlcjE3NjU2NzI= | User | false |
openclaw/openclaw | 4,038,254,674 | I_kwDOQb6kR87wsuBS | 38,795 | https://github.com/openclaw/openclaw/issues/38795 | https://api.github.com/repos/openclaw/openclaw/issues/38795 | Bug Report: imageModel 自动切换未生效 | # Bug Report: imageModel 自动切换未生效
## 问题描述
配置了 `agents.defaults.imageModel.primary` 和 `agents.defaults.imageModel.fallbacks`，但当 WebChat 渠道发送图片时，模型没有自动切换到 imageModel。
## 环境信息
- **OpenClaw 版本**: 2026.3.2
- **Node 版本**: 24.13.0
- **OS**: Windows 10.0.19045 (x64)
- **渠道**: WebChat
## 配置
```json5
{
"agents": {
"de... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-07T11:15:57Z | 2026-04-27T05:01:04Z | 2026-04-24T19:40:30Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | joel-duqiong | 13,653,416 | MDQ6VXNlcjEzNjUzNDE2 | User | false |
openclaw/openclaw | 4,038,258,381 | I_kwDOQb6kR87wsu7N | 38,797 | https://github.com/openclaw/openclaw/issues/38797 | https://api.github.com/repos/openclaw/openclaw/issues/38797 | [Feature]: Per-agent memory isolation for memory-lancedb plugin | ## Summary
The `memory-lancedb` plugin must support per-agent memory isolation. Currently, all agents on an OpenClaw instance share a single LanceDB database with no agent-level scoping — auto-capture and auto-recall hooks leak memories across agent boundaries.
---
## Problem to solve
In multi-agent deployments whe... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-07T11:17:35Z | 2026-04-28T04:43:19Z | 2026-04-26T01:47:28Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | mikezupper | 70,441,620 | MDQ6VXNlcjcwNDQxNjIw | User | false |
openclaw/openclaw | 4,038,259,047 | I_kwDOQb6kR87wsvFn | 38,799 | https://github.com/openclaw/openclaw/issues/38799 | https://api.github.com/repos/openclaw/openclaw/issues/38799 | [Feature]: Windows automation skills - bridging the platform gap | # [Feature]: Windows automation skills - bridging the platform gap
## Problem
OpenClaw ships 10+ macOS-specific skills (apple-notes, apple-reminders, things-mac, imsg, peekaboo) but has **zero Windows-equivalent skills**. Given that the recommended Windows setup is WSL2, there's a natural bridge available: WSL2 can c... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-07T11:17:54Z | 2026-04-27T05:01:03Z | 2026-04-24T19:40:34Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | aloth | 12,732,979 | MDQ6VXNlcjEyNzMyOTc5 | User | false |
openclaw/openclaw | 4,038,259,470 | I_kwDOQb6kR87wsvMO | 38,800 | https://github.com/openclaw/openclaw/issues/38800 | https://api.github.com/repos/openclaw/openclaw/issues/38800 | [Bug]: Google Chat integration crashes on startup (TypeError in ESM loader due to @nolyfill/domexception override) | ### Bug type
Crash (process/app exits or hangs)
### Summary
### Describe the bug
When configuring the Google Chat provider, the gateway crashes internally when attempting to get an access token via `google-auth-library`. This is caused by a chain reaction starting from the `node-domexception` override in `package.js... | closed | completed | true | 2 | [
"bug",
"stale",
"bug:crash"
] | [] | 2026-03-07T11:18:07Z | 2026-04-27T05:01:02Z | 2026-04-24T19:40:38Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | cflev | 17,141,683 | MDQ6VXNlcjE3MTQxNjgz | User | false |
openclaw/openclaw | 4,038,259,810 | I_kwDOQb6kR87wsvRi | 38,801 | https://github.com/openclaw/openclaw/issues/38801 | https://api.github.com/repos/openclaw/openclaw/issues/38801 | Telegram polling hangs after receiving large file (~6MB PDF) | ## Bug Description
Sending a large file (~6MB PDF) via Telegram DM to the bot causes the Telegram polling to hang completely. After the file is sent, **no further messages are received** until the gateway is restarted.
## Steps to Reproduce
1. Send a ~6MB PDF document to the bot via Telegram DM
2. Send a follow-up t... | closed | completed | true | 4 | [
"stale"
] | [] | 2026-03-07T11:18:15Z | 2026-04-27T05:01:02Z | 2026-04-24T19:40:42Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | lopponep | 151,858,740 | U_kgDOCQ0uNA | User | false |
openclaw/openclaw | 4,038,265,591 | I_kwDOQb6kR87wswr3 | 38,806 | https://github.com/openclaw/openclaw/issues/38806 | https://api.github.com/repos/openclaw/openclaw/issues/38806 | [Bug]: bot replies outside thread in P2P direct message thread | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When a user sends a message inside a Feishu thread in a P2P (direct message) chat, the bot's reply is sent outside the thread as a standalone message, instead of replying within the thread.
This regression was introduced by #33789, whi... | open | null | false | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-07T11:22:47Z | 2026-05-04T06:15:48Z | null | CONTRIBUTOR | null | 20260504T090105Z | 2026-05-04T09:01:05Z | LiaoyuanNing | 259,494,737 | U_kgDOD3eTUQ | User | false |
openclaw/openclaw | 4,038,269,441 | I_kwDOQb6kR87wsxoB | 38,809 | https://github.com/openclaw/openclaw/issues/38809 | https://api.github.com/repos/openclaw/openclaw/issues/38809 | [Windows] Image payload missing for google-generative-ai (SafeOpenError regression in v2026.3.2) | ## Description
In the latest stable version (**v2026.3.2**, build `85377a2`) running on **Windows 11** (10.0.26200, x64), images received from Telegram are silently discarded when routed to a model configured with the native `google-generative-ai` API adapter.
The same images work flawlessly when using the `openai-co... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-07T11:25:33Z | 2026-04-27T05:01:01Z | 2026-04-24T19:40:53Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Skrblik | 139,452,656 | U_kgDOCE_g8A | User | false |
openclaw/openclaw | 4,038,269,507 | I_kwDOQb6kR87wsxpD | 38,810 | https://github.com/openclaw/openclaw/issues/38810 | https://api.github.com/repos/openclaw/openclaw/issues/38810 | bug(podman): OPENCLAW_GATEWAY_BIND in .env ignored — gateway always starts with --bind loopback | ## Bug Description
When using the Podman launcher (`scripts/run-openclaw-podman.sh`), setting `OPENCLAW_GATEWAY_BIND` in the `.env` file has no effect. The gateway always starts with `--bind loopback` regardless of the value in `.env`.
## Root Cause
In `scripts/run-openclaw-podman.sh`, `GATEWAY_BIND` is evaluated **... | closed | completed | true | 3 | [] | [] | 2026-03-07T11:25:36Z | 2026-04-30T04:49:41Z | 2026-04-25T05:49:31Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | majinyu666 | 31,942,171 | MDQ6VXNlcjMxOTQyMTcx | User | false |
openclaw/openclaw | 4,038,277,975 | I_kwDOQb6kR87wsztX | 38,816 | https://github.com/openclaw/openclaw/issues/38816 | https://api.github.com/repos/openclaw/openclaw/issues/38816 | [Bug]: Image tool incorrectly adds anthropic provider prefix to all model IDs / 图像工具自动为所有模型ID错误添加anthropic provider前缀 | ### 中文版本
#### 环境信息
- OpenClaw版本：2026.3.2
- 操作系统：Darwin 25.3.0 (arm64)
- 部署方式：本地部署
#### 复现步骤
1. 在`openclaw.json`中配置`agents.defaults.imageModel`列表，填写正确的无前缀模型ID：
```json
"imageModel": {
"primary": "moondream",
"fallbacks": [
"qwen2.5vl:7b",
"G-2.5-f"
]
}
```
2. 重启网关后调用`image`工具执行图像识别任务
3. 工具返回错误：`Unknown mo... | closed | completed | true | 3 | [] | [] | 2026-03-07T11:31:38Z | 2026-05-01T04:44:07Z | 2026-04-28T09:20:49Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | alainasclaw | 259,996,494 | U_kgDOD387Tg | User | false |
openclaw/openclaw | 4,038,279,290 | I_kwDOQb6kR87ws0B6 | 38,817 | https://github.com/openclaw/openclaw/issues/38817 | https://api.github.com/repos/openclaw/openclaw/issues/38817 | [Bug]: Volcengine/Kimi models fail parameter validation due to unsupported minLength keyword / 火山引擎/Kimi系列模型因不支持minLength关键字导致参数校验失败 | ### 中文版本
#### 环境信息
- OpenClaw版本：2026.3.2
- 操作系统：Darwin 25.3.0 (arm64)
- 部署方式：本地部署
- 涉及模型：火山引擎托管的Kimi 2.5/2.0全系列模型
#### 复现步骤
1. 配置使用火山引擎托管的Kimi系列模型
2. 调用包含`minLength`参数校验的工具（如message工具的modal字段配置）
3. 接口返回错误：`Keyword 'minLength' is not supported for type 'string' at path '$.properties.label'`
#### 根因分析
代码中已有`stripXaiUns... | open | null | false | 3 | [] | [] | 2026-03-07T11:32:13Z | 2026-05-04T10:02:52Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | alainasclaw | 259,996,494 | U_kgDOD387Tg | User | false |
openclaw/openclaw | 4,038,293,182 | I_kwDOQb6kR87ws3a- | 38,820 | https://github.com/openclaw/openclaw/issues/38820 | https://api.github.com/repos/openclaw/openclaw/issues/38820 | github-copilot provider: misleading 401 when models.json has wrong baseUrl | ## Summary
When `models.json` has an incorrect `baseUrl` for the `github-copilot` provider (e.g., `https://models.inference.ai.azure.com` — the GitHub Models endpoint), agent runs fail with `401 Bad credentials`. The error message gives no indication that the wrong endpoint is being used.
## Steps to Reproduce
1. Ru... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T11:39:55Z | 2026-04-27T05:01:00Z | 2026-04-24T19:40:58Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Hyungwon | 804,440 | MDQ6VXNlcjgwNDQ0MA== | User | false |
openclaw/openclaw | 4,038,294,979 | I_kwDOQb6kR87ws33D | 38,821 | https://github.com/openclaw/openclaw/issues/38821 | https://api.github.com/repos/openclaw/openclaw/issues/38821 | [Bug]: P0 Bug Report: Catastrophic and Persistent Vision System Failure in a Specific Agent Instance | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Summary of Issue
The AI agent agentb, running on the google/gemini-2.5-pro model, is experiencing a total and persistent failure of its vision system. When presented with an image via Discord attachment, the agent does not perceive the act... | closed | completed | true | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-03-07T11:40:56Z | 2026-04-29T04:38:33Z | 2026-04-26T09:46:47Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | melaniesdfm | 8,388,982 | MDQ6VXNlcjgzODg5ODI= | User | false |
openclaw/openclaw | 4,038,309,503 | I_kwDOQb6kR87ws7Z_ | 38,823 | https://github.com/openclaw/openclaw/issues/38823 | https://api.github.com/repos/openclaw/openclaw/issues/38823 | [Bug]: Cron announce delivery times out — gateway cannot self-connect for delivery (v2026.3.2) | ## Environment
- **OpenClaw version:** 2026.3.2 (85377a2)
- **OS:** Linux 6.8.0-101-generic (x64)
- **Node:** v22.22.0
- **Channel:** Telegram
## Bug Description
Cron jobs with `delivery.mode: "announce"` fail to deliver results to Telegram. The crons themselves run successfully (verified via Mission Control and logs... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T11:51:02Z | 2026-04-27T05:00:59Z | 2026-04-24T19:41:02Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | philmossman | 12,981,632 | MDQ6VXNlcjEyOTgxNjMy | User | false |
openclaw/openclaw | 4,038,312,661 | I_kwDOQb6kR87ws8LV | 38,824 | https://github.com/openclaw/openclaw/issues/38824 | https://api.github.com/repos/openclaw/openclaw/issues/38824 | [Bug] Feishu: blockStreamingDefault causes replies silently dropped + card streaming duplicates content | ## Summary
Two related Feishu plugin bugs in OpenClaw 2026.3.2:
1. **`blockStreamingDefault: "on"` causes replies to be silently dropped** — Gateway logs show `dispatch complete (queuedFinal=false, replies=0)` even though the agent generated a response.
2. **Card streaming duplicates content** — When `streaming: true... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T11:53:05Z | 2026-04-27T05:00:58Z | 2026-04-24T19:41:07Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | visense | 3,490,833 | MDQ6VXNlcjM0OTA4MzM= | User | false |
openclaw/openclaw | 4,038,323,204 | I_kwDOQb6kR87ws-wE | 38,828 | https://github.com/openclaw/openclaw/issues/38828 | https://api.github.com/repos/openclaw/openclaw/issues/38828 | [Bug]: OpenClaw browser trace start --browser-profile options are misaligned. | ### Bug type
Crash (process/app exits or hangs)
### Summary
OpenClaw browser trace start --browser-profile options are misaligned.
### Steps to reproduce
ws@DESKTOP-OKFKQOP:~/.openclaw$ openclaw browser trace start --browser-profile openclaw
🦞 OpenClaw 2026.3.2 (85377a2) — One CLI to rule them all, and one more ... | closed | completed | true | 4 | [
"bug",
"stale",
"bug:crash"
] | [] | 2026-03-07T12:00:26Z | 2026-04-27T05:00:57Z | 2026-04-24T19:41:12Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | jmjl211 | 127,288,325 | U_kgDOB5ZEBQ | User | false |
openclaw/openclaw | 4,038,325,301 | I_kwDOQb6kR87ws_Q1 | 38,829 | https://github.com/openclaw/openclaw/issues/38829 | https://api.github.com/repos/openclaw/openclaw/issues/38829 | openclaw tui terminal cannot receive real-time messages | **Environment:** - OS: macOS (Darwin 25.3.0) - OpenClaw: 2026.3.2 (85377a2) - Terminal: openclaw-tui **Bug Description:** When using `openclaw tui` for terminal chat, historical messages display correctly, but real-time new messages are not received. Messages only appear after exiting and re-entering the terminal. **St... | open | null | false | 2 | [] | [] | 2026-03-07T12:01:42Z | 2026-05-04T06:15:49Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | galen2017 | 30,015,650 | MDQ6VXNlcjMwMDE1NjUw | User | false |
openclaw/openclaw | 4,038,334,591 | I_kwDOQb6kR87wtBh_ | 38,836 | https://github.com/openclaw/openclaw/issues/38836 | https://api.github.com/repos/openclaw/openclaw/issues/38836 | Gateway restart silently drops in-flight Feishu sessions — no user notification | ## Summary
When the gateway is restarted (via `systemctl restart` or `openclaw gateway restart`), any **in-flight LLM requests** are silently dropped. The user who sent a message via Feishu receives **no reply and no notification** that a restart occurred. From the user's perspective, the bot simply stops responding.
... | closed | not_planned | true | 2 | [] | [] | 2026-03-07T12:07:45Z | 2026-05-03T04:35:08Z | 2026-04-30T20:15:46Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | vulturekun | 39,609,631 | MDQ6VXNlcjM5NjA5NjMx | User | false |
openclaw/openclaw | 4,038,335,311 | I_kwDOQb6kR87wtBtP | 38,837 | https://github.com/openclaw/openclaw/issues/38837 | https://api.github.com/repos/openclaw/openclaw/issues/38837 | Feishu agent: no visible feedback between dispatch and LLM response | ## Summary
When a Feishu user sends a message to a bot-bound agent, the gateway logs show `dispatching to agent` immediately, but there is **no visible feedback** to the user until the LLM response arrives (typically 10-35+ seconds later). During this time, the user has no way to know if:
- The message was received
-... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T12:08:15Z | 2026-04-27T05:00:57Z | 2026-04-24T19:41:16Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | vulturekun | 39,609,631 | MDQ6VXNlcjM5NjA5NjMx | User | false |
openclaw/openclaw | 4,038,346,890 | I_kwDOQb6kR87wtEiK | 38,839 | https://github.com/openclaw/openclaw/issues/38839 | https://api.github.com/repos/openclaw/openclaw/issues/38839 | Feature Request: Cron Job Reschedule & Retry Mechanism | I've created a feature request for the cron job reschedule mechanism. See the details in the attached file. | closed | completed | true | 4 | [
"stale"
] | [] | 2026-03-07T12:15:05Z | 2026-04-27T05:00:56Z | 2026-04-24T19:41:27Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | 31878278 | 6,007,867 | MDQ6VXNlcjYwMDc4Njc= | User | false |
openclaw/openclaw | 4,038,368,888 | I_kwDOQb6kR87wtJ54 | 38,844 | https://github.com/openclaw/openclaw/issues/38844 | https://api.github.com/repos/openclaw/openclaw/issues/38844 | [Bug]: browser upload/file chooser flow is flaky and can misreport stale-click failures as browser control timeouts | ## Summary
In the managed `openclaw` browser profile, browser automation around file chooser flows can fail intermittently in a misleading way.
There appear to be two related problems:
1. The effective automation flow can split file upload into two separate requests:
- `browser upload(paths)`
- then a separate... | open | null | false | 4 | [] | [] | 2026-03-07T12:29:24Z | 2026-05-04T06:04:52Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | tigicion | 48,045,299 | MDQ6VXNlcjQ4MDQ1Mjk5 | User | false |
openclaw/openclaw | 4,038,374,629 | I_kwDOQb6kR87wtLTl | 38,847 | https://github.com/openclaw/openclaw/issues/38847 | https://api.github.com/repos/openclaw/openclaw/issues/38847 | Feature Request: memorySearch.autoInject — automatic per-message memory retrieval | ## Problem
Currently, `memory_search` is a pull-based tool — the agent must proactively decide to call it. This creates a systematic blind spot: if the agent doesn't *know* to look for something, it never finds it.
**Concrete failure mode:** An agent gets asked about a past project. It answers confidently from its fi... | closed | completed | true | 4 | [
"stale"
] | [] | 2026-03-07T12:32:21Z | 2026-04-27T05:00:55Z | 2026-04-24T19:41:38Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | sumpter32 | 13,649,640 | MDQ6VXNlcjEzNjQ5NjQw | User | false |
openclaw/openclaw | 4,038,404,026 | I_kwDOQb6kR87wtSe6 | 38,853 | https://github.com/openclaw/openclaw/issues/38853 | https://api.github.com/repos/openclaw/openclaw/issues/38853 | [Bug]: Discord channel reconnect loop triggers Cloudflare rate-limit ban (Error 1015) | ## Description
When the Discord channel fails to connect (e.g., transient API error or brief outage), the gateway enters a tight retry loop calling `/users/@me/guilds` to resolve guild info. Each failed attempt triggers another retry within ~1-2 minutes, quickly exceeding Cloudflare's rate limit threshold. Cloudflare ... | closed | completed | true | 2 | [] | [] | 2026-03-07T12:46:41Z | 2026-05-02T04:26:49Z | 2026-04-30T00:47:04Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | djgeorg3 | 31,196,850 | MDQ6VXNlcjMxMTk2ODUw | User | false |
openclaw/openclaw | 4,038,416,472 | I_kwDOQb6kR87wtVhY | 38,858 | https://github.com/openclaw/openclaw/issues/38858 | https://api.github.com/repos/openclaw/openclaw/issues/38858 | [Feature Request] Add slash commands autocomplete to WebChat/Control UI | ## Problem
Currently, the WebChat and Control UI do not support slash commands autocomplete. When users type `/` in the chat input, there is no dropdown or popup showing available commands.
## Current Behavior
- **Discord/Telegram**: Native commands with autocomplete work well
- **WebChat/Control UI**: Users must ty... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-07T12:54:18Z | 2026-04-27T05:00:54Z | 2026-04-24T19:41:57Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | tociyou1995 | 150,312,743 | U_kgDOCPWXJw | User | false |
openclaw/openclaw | 4,038,442,468 | I_kwDOQb6kR87wtb3k | 38,867 | https://github.com/openclaw/openclaw/issues/38867 | https://api.github.com/repos/openclaw/openclaw/issues/38867 | [Bug]: Brave Search API erro422， x-subscription-token header | ### Bug type
Regression (worked before, now fails)
### Summary
web_search 工具调用 Brave Search API 时返回 422 错误
### Steps to reproduce
1. 配置 Brave API Key
2. 调用 web_search
3. 报错 "fetch failed"
### Expected behavior
## 期望行为
同时传递两个 header
### Actual behavior
当前工具只传了第一个，导致验证失败
### OpenClaw version
2026.2.17
##... | closed | not_planned | true | 2 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-07T13:05:54Z | 2026-04-27T05:00:53Z | 2026-04-24T19:42:01Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | JING22336546 | 266,294,974 | U_kgDOD99Wvg | User | false |
openclaw/openclaw | 4,038,474,714 | I_kwDOQb6kR87wtjva | 38,878 | https://github.com/openclaw/openclaw/issues/38878 | https://api.github.com/repos/openclaw/openclaw/issues/38878 | Feature Request: Support multi-device shared memory | Currently when using multiple devices to run OpenClaw, the memory is stored locally and cannot be synced across devices.
Use case:
- Main computer (for work)
- Gaming PC (for gaming)
- Both running OpenClaw
Requirement: Support multi-device shared memory
Possible solutions:
1. MCP memory service (like clou... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T13:20:52Z | 2026-04-27T05:00:51Z | 2026-04-24T19:42:14Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | carlos999-hqsama | 245,906,921 | U_kgDODqg96Q | User | false |
openclaw/openclaw | 4,038,489,884 | I_kwDOQb6kR87wtncc | 38,884 | https://github.com/openclaw/openclaw/issues/38884 | https://api.github.com/repos/openclaw/openclaw/issues/38884 | [Bug]: Control UI WebChat: queued user messages not displayed in conversation thread | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
## Description
When a user sends a message via the Control UI webchat while the agent is
mid-response (i.e., the message gets queued), the message is processed
correctly by the queue system but never appears in the webchat conversation
... | closed | completed | true | 5 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-07T13:27:34Z | 2026-04-28T04:43:18Z | 2026-04-26T01:48:00Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | darrenmklein | 84,782,409 | MDQ6VXNlcjg0NzgyNDA5 | User | false |
openclaw/openclaw | 4,038,497,369 | I_kwDOQb6kR87wtpRZ | 38,887 | https://github.com/openclaw/openclaw/issues/38887 | https://api.github.com/repos/openclaw/openclaw/issues/38887 | Telegram multi-account: stale-socket restarts kill idle bot connections | ## Bug
**Version:** 2026.3.2 (latest npm)
**Problem:** When running multiple Telegram bot accounts (multi-account setup), the health monitor triggers `stale-socket` restarts every ~35 minutes for low-traffic bot accounts. The bot starts, connects successfully, but since it receives few/no messages (e.g., a mention-on... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-07T13:31:54Z | 2026-04-27T05:00:50Z | 2026-04-24T19:42:18Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | iJaack | 6,631,681 | MDQ6VXNlcjY2MzE2ODE= | User | false |
openclaw/openclaw | 4,038,527,135 | I_kwDOQb6kR87wtwif | 38,899 | https://github.com/openclaw/openclaw/issues/38899 | https://api.github.com/repos/openclaw/openclaw/issues/38899 | [Bug]: Failed to parse input at pos 161 tool call | ### Bug type
Regression (worked before, now fails)
### Summary
Getting this parse input error with toolcall frequently with tool call. I'm running Qwen3.5 35B Q5 model via llama.cpp
### Steps to reproduce
Add these to HEARTBEAT.md
# Refresh the search index
- Run `qmd update --reindex`
# Update embeddings
- Run `q... | closed | not_planned | true | 3 | [
"bug",
"regression"
] | [] | 2026-03-07T13:49:27Z | 2026-04-28T04:43:17Z | 2026-04-26T01:48:10Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | kendy92 | 12,098,762 | MDQ6VXNlcjEyMDk4NzYy | User | false |
openclaw/openclaw | 4,038,533,144 | I_kwDOQb6kR87wtyAY | 38,902 | https://github.com/openclaw/openclaw/issues/38902 | https://api.github.com/repos/openclaw/openclaw/issues/38902 | [Bug]: run error: HTTP 422: "check open ai req parameter error" | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
centos7部署Openclaw成功且配置本地模型QWEN成功后，使用报错：run error: HTTP 422: "check open ai req parameter error"
### Steps to reproduce
1.centos7 install OpenClaw 2026.3.1
2.config local AI: DeepSeek-R1-Distill-Qwen-32B，
附：Run command "curl -i -X POST 'h... | closed | completed | true | 23 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-07T13:53:07Z | 2026-04-27T05:00:49Z | 2026-04-24T19:42:29Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | li198655 | 265,199,773 | U_kgDOD86gnQ | User | false |
openclaw/openclaw | 4,038,534,908 | I_kwDOQb6kR87wtyb8 | 38,903 | https://github.com/openclaw/openclaw/issues/38903 | https://api.github.com/repos/openclaw/openclaw/issues/38903 | [Bug]: Nova/Bedrock 500 ValidationException when switching models mid-session — extraneous 'type' key in content blocks | ## Summary
Switching to a Nova/Bedrock model mid-session causes a `ValidationException` 500 error. Fresh Nova sessions always work. The error only occurs after switching away from another model (e.g. Anthropic/Claude) and then back to Nova.
## Version
`2026.3.2`
## Steps to Reproduce
1. Start a session with any An... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-07T13:54:22Z | 2026-04-27T05:00:48Z | 2026-04-24T19:42:33Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | alexcam1901 | 43,379,154 | MDQ6VXNlcjQzMzc5MTU0 | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.