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 | 4,106,694,376 | I_kwDOQb6kR870xy7o | 50,950 | https://github.com/openclaw/openclaw/issues/50950 | https://api.github.com/repos/openclaw/openclaw/issues/50950 | Feature Request: Display images directly in webchat responses | ## Problem
When the assistant generates or reads an image file (PNG/JPG), the image is not displayed directly in the webchat interface. Instead, users see either:
- A broken/empty response
- Need to manually open the file path in the filesystem
This happens with:
- read tool on image files
- message tool with image a... | closed | completed | true | 1 | [] | [] | 2026-03-20T08:57:05Z | 2026-04-30T04:53:18Z | 2026-04-25T04:59:58Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | wangbiwang | 45,363,247 | MDQ6VXNlcjQ1MzYzMjQ3 | User | false |
openclaw/openclaw | 4,106,695,076 | I_kwDOQb6kR870xzGk | 50,951 | https://github.com/openclaw/openclaw/issues/50951 | https://api.github.com/repos/openclaw/openclaw/issues/50951 | Feature Request: Display images directly in webchat responses | ## Problem
When the assistant generates or reads an image file, the image is not displayed directly in webchat.
## Expected Behavior
Images should render inline in webchat.
## Proposed Solution
Add image rendering support. | closed | completed | true | 1 | [] | [] | 2026-03-20T08:57:17Z | 2026-04-30T04:53:17Z | 2026-04-25T05:00:05Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | wangbiwang | 45,363,247 | MDQ6VXNlcjQ1MzYzMjQ3 | User | false |
openclaw/openclaw | 4,106,696,629 | I_kwDOQb6kR870xze1 | 50,952 | https://github.com/openclaw/openclaw/issues/50952 | https://api.github.com/repos/openclaw/openclaw/issues/50952 | Feature Request: Display images directly in webchat responses | When the assistant generates or reads an image file, the image is not displayed directly in webchat. | closed | completed | true | 1 | [] | [] | 2026-03-20T08:57:44Z | 2026-04-30T04:53:16Z | 2026-04-25T05:00:12Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | wangbiwang | 45,363,247 | MDQ6VXNlcjQ1MzYzMjQ3 | User | false |
openclaw/openclaw | 4,106,697,266 | I_kwDOQb6kR870xzoy | 50,953 | https://github.com/openclaw/openclaw/issues/50953 | https://api.github.com/repos/openclaw/openclaw/issues/50953 | Feature Request: Display images in webchat | Images should render inline in webchat like Discord/Telegram. | closed | completed | true | 1 | [] | [] | 2026-03-20T08:57:55Z | 2026-04-30T04:53:16Z | 2026-04-25T05:00:19Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | wangbiwang | 45,363,247 | MDQ6VXNlcjQ1MzYzMjQ3 | User | false |
openclaw/openclaw | 4,106,697,607 | I_kwDOQb6kR870xzuH | 50,954 | https://github.com/openclaw/openclaw/issues/50954 | https://api.github.com/repos/openclaw/openclaw/issues/50954 | [Bug]: openclaw onboard wizard performs silent, unapproved skill installation based on fuzzy matching | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
The **openclaw onboard** command currently operates with an over-aggressive "Wizard" mode. It analyzes user workspace history to infer requirements, performs fuzzy searches on ClawHub, and installs skills without explicit user confirmation.... | closed | not_planned | true | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-03-20T08:58:01Z | 2026-04-30T04:53:15Z | 2026-04-25T05:00:26Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | ie2718 | 209,652,275 | U_kgDODH8KMw | User | false |
openclaw/openclaw | 4,106,701,686 | I_kwDOQb6kR870x0t2 | 50,958 | https://github.com/openclaw/openclaw/issues/50958 | https://api.github.com/repos/openclaw/openclaw/issues/50958 | Feature Request: Display images in webchat | Images should render inline in webchat. | closed | completed | true | 1 | [] | [] | 2026-03-20T08:59:08Z | 2026-04-30T04:53:12Z | 2026-04-25T05:00:47Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | wangbiwang | 45,363,247 | MDQ6VXNlcjQ1MzYzMjQ3 | User | false |
openclaw/openclaw | 4,106,702,735 | I_kwDOQb6kR870x0-P | 50,959 | https://github.com/openclaw/openclaw/issues/50959 | https://api.github.com/repos/openclaw/openclaw/issues/50959 | Feature Request: Display images in webchat | Test | closed | completed | true | 1 | [] | [] | 2026-03-20T08:59:25Z | 2026-04-30T04:53:12Z | 2026-04-25T05:00:54Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | wangbiwang | 45,363,247 | MDQ6VXNlcjQ1MzYzMjQ3 | User | false |
openclaw/openclaw | 4,107,392,304 | I_kwDOQb6kR8700dUw | 51,007 | https://github.com/openclaw/openclaw/issues/51007 | https://api.github.com/repos/openclaw/openclaw/issues/51007 | feat: sessions_send should include sourceSession in chat.history message metadata | ## Problem
When an agent uses `sessions_send` to send a message to another agent, the receiving agent's `chat.history` stores the message with `role: "user"` but **without any information about the source session**.
This makes it impossible for frontend clients (like AsukaDeck) to distinguish:
- Messages sent by the ... | closed | completed | true | 2 | [] | [] | 2026-03-20T11:39:39Z | 2026-04-30T04:53:08Z | 2026-04-25T05:01:30Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | komuken | 206,610,109 | U_kgDODFCevQ | User | false |
openclaw/openclaw | 4,107,394,123 | I_kwDOQb6kR8700dxL | 51,008 | https://github.com/openclaw/openclaw/issues/51008 | https://api.github.com/repos/openclaw/openclaw/issues/51008 | [Bug]: 2026.3.13 local loopback regression: operator.read failures and gateway closed (1000) on CLI commands | ### Bug type
Regression (worked before, now fails)
### Summary
After upgrading to OpenClaw 2026.3.13, the local loopback gateway behaves inconsistently.
This is broader than the already reported status/probe `missing scope: operator.read` issue.
In my case, the same gateway runtime shows:
- successful local webcha... | closed | completed | true | 5 | [
"bug",
"regression"
] | [] | 2026-03-20T11:40:02Z | 2026-04-30T04:53:07Z | 2026-04-25T05:01:37Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Taureg | 148,580,378 | U_kgDOCNsoGg | User | false |
openclaw/openclaw | 4,107,418,252 | I_kwDOQb6kR8700jqM | 51,010 | https://github.com/openclaw/openclaw/issues/51010 | https://api.github.com/repos/openclaw/openclaw/issues/51010 | WebSocket ping/pong heartbeat config to prevent browser idle disconnects | ## Problem
Control UI WebSocket connections repeatedly disconnect with code 1001 ("Going Away") originating from the client (browser), not the server. This is a known browser behavior - closing idle WebSocket connections after a certain threshold.
Flapping episodes: ~10-70 minutes of stable connection, then rapid dis... | closed | completed | true | 2 | [] | [] | 2026-03-20T11:45:27Z | 2026-04-30T04:53:06Z | 2026-04-25T05:01:44Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Nerdless-ship-it | 257,649,571 | U_kgDOD1trow | User | false |
openclaw/openclaw | 4,107,422,492 | I_kwDOQb6kR8700ksc | 51,012 | https://github.com/openclaw/openclaw/issues/51012 | https://api.github.com/repos/openclaw/openclaw/issues/51012 | WhatsApp relink succeeds, then send fails with No active listener and session drops (401) | ## Bug: WhatsApp relink succeeds, then listener is unavailable and session logs out (401)
### Environment
- OpenClaw: 2026.3.13 (61d171a)
- OS: Windows
- Date: 2026-03-20
- Channel: WhatsApp (`default`)
### Reproduction
1. `openclaw channels login --channel whatsapp --account default`
2. Scan QR in WhatsApp Linked De... | closed | completed | true | 9 | [] | [] | 2026-03-20T11:46:21Z | 2026-04-30T04:53:05Z | 2026-04-25T05:01:51Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | easyvaru-hue | 265,017,145 | U_kgDOD8vXOQ | User | false |
openclaw/openclaw | 4,107,442,717 | I_kwDOQb6kR8700pod | 51,014 | https://github.com/openclaw/openclaw/issues/51014 | https://api.github.com/repos/openclaw/openclaw/issues/51014 | Telegram main session can wedge after long-running context and stop replying | ## What happened
An OpenClaw instance running on Telegram stopped replying even though the Telegram webhook path itself remained healthy.
In this case:
- Telegram webhook stayed registered and healthy
- local webhook listener stayed up
- the gateway service stayed running
- incoming Telegram messages appeared to reac... | closed | completed | true | 3 | [] | [] | 2026-03-20T11:51:03Z | 2026-04-30T04:53:04Z | 2026-04-25T05:01:58Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | mvpbin | 25,851,317 | MDQ6VXNlcjI1ODUxMzE3 | User | false |
openclaw/openclaw | 4,107,483,640 | I_kwDOQb6kR8700zn4 | 51,018 | https://github.com/openclaw/openclaw/issues/51018 | https://api.github.com/repos/openclaw/openclaw/issues/51018 | Discord can log in behind proxy, but native command deploy / bot identity fetch fail with 'fetch failed' | ## Summary
When Discord is configured behind an HTTP proxy, OpenClaw can partially start the Discord provider and log in successfully, but some startup REST calls still fail with `fetch failed`.
In my setup, the gateway is healthy and Discord shows `logged in to discord`, but these startup steps consistently fail:
- n... | closed | completed | true | 3 | [] | [] | 2026-03-20T12:00:06Z | 2026-04-30T04:53:04Z | 2026-04-25T05:02:05Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | mrpl327 | 58,623,660 | MDQ6VXNlcjU4NjIzNjYw | User | false |
openclaw/openclaw | 4,107,492,384 | I_kwDOQb6kR87001wg | 51,022 | https://github.com/openclaw/openclaw/issues/51022 | https://api.github.com/repos/openclaw/openclaw/issues/51022 | [Bug]: Tool calls silently hang after extended session with heavy tool usage | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Description:
After a long session involving many tool calls (exec, read, write, file operations),
tool calls stop executing. The model outputs a tool call normally (confirmed in
LM Studio logs - generation completes successfully), but Ope... | closed | completed | true | 6 | [
"bug",
"bug:behavior"
] | [] | 2026-03-20T12:01:56Z | 2026-04-30T04:53:03Z | 2026-04-25T05:02:12Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Alex-un | 97,377,406 | U_kgDOBc3cfg | User | false |
openclaw/openclaw | 4,107,530,492 | I_kwDOQb6kR8700_D8 | 51,030 | https://github.com/openclaw/openclaw/issues/51030 | https://api.github.com/repos/openclaw/openclaw/issues/51030 | Sub-Agent Activity: show human-readable task labels instead of raw session IDs | ## Problem
The **Sub-Agent Activity** panel currently shows internal session keys:
```
agent:main:cron:7c98cfc0-eeb2-453e-befa-4aa5db87d51f:run:d55
a241fac6-396
```
This is database exhaust, not useful information. Users have no idea what's actually running.
## Goal
Replace raw session IDs with human-readable task... | closed | completed | true | 4 | [] | [] | 2026-03-20T12:10:27Z | 2026-04-30T04:53:01Z | 2026-04-25T05:02:26Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | TSCarterJr | 82,057,175 | MDQ6VXNlcjgyMDU3MTc1 | User | false |
openclaw/openclaw | 4,107,531,685 | I_kwDOQb6kR8700_Wl | 51,031 | https://github.com/openclaw/openclaw/issues/51031 | https://api.github.com/repos/openclaw/openclaw/issues/51031 | [Bug]: Tool calls silently hang after compaction in extended sessions (compaction.mode=safeguard) | ## Bug Summary
After extended sessions with heavy tool usage under `compaction.mode: safeguard`, tool calls stop executing. The model generates the tool call output normally (confirmed in LM Studio logs), but OpenClaw never executes it and the session hangs indefinitely. Starting a new session immediately restores nor... | closed | completed | true | 3 | [] | [] | 2026-03-20T12:10:40Z | 2026-04-30T04:53:01Z | 2026-04-25T05:02:33Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | ai-nurmamat | 130,913,840 | U_kgDOB82WMA | User | false |
openclaw/openclaw | 4,107,619,862 | I_kwDOQb6kR8701U4W | 51,034 | https://github.com/openclaw/openclaw/issues/51034 | https://api.github.com/repos/openclaw/openclaw/issues/51034 | Fleet/dashboard: alerts not auto-resolved on recovery, causing contradictory online status + active offline alert | ## Problem
The fleet/dashboard view shows contradictory state:
- **Status:** Online (current heartbeat shows gateway is alive)
- **Alerts:** "Gateway Offline" (active alert, never cleared)
This is a zombie alert — raised correctly when the gateway went offline, but never auto-resolved when it came back up.
## Root ... | closed | not_planned | true | 2 | [] | [] | 2026-03-20T12:29:55Z | 2026-04-30T04:53:00Z | 2026-04-25T05:02:40Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | TSCarterJr | 82,057,175 | MDQ6VXNlcjgyMDU3MTc1 | User | false |
openclaw/openclaw | 4,107,652,286 | I_kwDOQb6kR8701cy- | 51,037 | https://github.com/openclaw/openclaw/issues/51037 | https://api.github.com/repos/openclaw/openclaw/issues/51037 | Telegram native slash command control is too coarse for multi-account bots | ## Summary
Telegram native slash command registration is currently too coarse-grained for multi-bot / multi-account setups.
`channels.telegram.commands.native` appears to work at the Telegram provider level, which creates an awkward trade-off:
- keep native commands enabled and expose operator/admin slash commands o... | closed | completed | true | 2 | [] | [] | 2026-03-20T12:37:36Z | 2026-04-30T04:52:58Z | 2026-04-25T05:02:54Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | dongzhenye | 5,765,843 | MDQ6VXNlcjU3NjU4NDM= | User | false |
openclaw/openclaw | 4,107,717,132 | I_kwDOQb6kR8701soM | 51,043 | https://github.com/openclaw/openclaw/issues/51043 | https://api.github.com/repos/openclaw/openclaw/issues/51043 | [Feature]: freeride switch should preview and confirm before overwriting manually configured fallback chain | ### Summary
freeride switch silently overwrites manually configured fallback chain without preview or confirmation
### Problem to solve
When running `freeride switch` to add a model to the fallback chain, freeride silently rewrites the entire fallback list — replacing the manually configured chain with an auto-gener... | closed | completed | true | 2 | [
"enhancement"
] | [] | 2026-03-20T12:51:52Z | 2026-04-30T04:52:57Z | 2026-04-25T05:03:01Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | brucecbi | 142,571,553 | U_kgDOCH94IQ | User | false |
openclaw/openclaw | 4,107,759,271 | I_kwDOQb6kR870126n | 51,050 | https://github.com/openclaw/openclaw/issues/51050 | https://api.github.com/repos/openclaw/openclaw/issues/51050 | Slack channel: support mediaLocalRoots config for outbound local file sends | ## Problem
When an agent sends a local file via Slack using the `message` tool with `filePath`, files under `workspace-*` directories are rejected by `assertLocalMediaAllowed()` with:
```
Local media path is not under an allowed directory: /home/user/.openclaw/workspace-ops/test-file.txt
```
The default allowed root... | closed | not_planned | true | 2 | [] | [] | 2026-03-20T13:01:19Z | 2026-04-30T04:52:56Z | 2026-04-25T05:03:22Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Frankla20 | 265,706,740 | U_kgDOD9Zc9A | User | false |
openclaw/openclaw | 4,107,883,369 | I_kwDOQb6kR8702VNp | 51,058 | https://github.com/openclaw/openclaw/issues/51058 | https://api.github.com/repos/openclaw/openclaw/issues/51058 | Cerebras: 400 error caused by unsupported 'store' parameter (compat.supportsStore ignored) | ## Summary
When using Cerebras provider (`cerebras/llama3.1-8b`, `cerebras/gpt-oss-120b`, etc.), every request fails with `400 status code (no body)`. The actual error from Cerebras API is:
```json
{"message":"body.store: property 'body.store' is unsupported","type":"invalid_request_error","param":"validation_error",... | closed | completed | true | 3 | [] | [] | 2026-03-20T13:25:49Z | 2026-04-30T04:52:55Z | 2026-04-25T05:03:29Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | G0ldenhawk | 208,365,016 | U_kgDODGtl2A | User | false |
openclaw/openclaw | 4,107,987,702 | I_kwDOQb6kR8702ur2 | 51,063 | https://github.com/openclaw/openclaw/issues/51063 | https://api.github.com/repos/openclaw/openclaw/issues/51063 | Agents routed through openai-codex / openai-codex-responses can reply as if tools executed without actually executing them | ### Summary
In some OpenClaw agent runs, when the selected model/provider path resolves to `openai-codex` / `openai-codex-responses`, the agent can produce execution-like status text such as:
- “Executing these exact commands now”
- “Running now”
- “I’ll return with outputs”
…but no real tool execution happens:
- no... | closed | completed | true | 3 | [] | [] | 2026-03-20T13:46:27Z | 2026-04-30T04:52:53Z | 2026-04-25T05:03:43Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | glfruit | 3,939,154 | MDQ6VXNlcjM5MzkxNTQ= | User | false |
openclaw/openclaw | 4,108,000,736 | I_kwDOQb6kR8702x3g | 51,066 | https://github.com/openclaw/openclaw/issues/51066 | https://api.github.com/repos/openclaw/openclaw/issues/51066 | [Bug]: sessions.json sessionId/sessionFile null for heartbeat-origin sessions | ## Description
Sessions initialized via `origin.provider=heartbeat` do not write `sessionId` or `sessionFile` back to `sessions.json`. This causes `sessions_history` to return empty `[]` despite real transcript content existing on disk.
## Steps to Reproduce
1. Agent session initialized by heartbeat cron (not Telegr... | closed | not_planned | true | 2 | [] | [] | 2026-03-20T13:49:04Z | 2026-04-30T04:52:52Z | 2026-04-25T05:03:56Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | kaxo-io | 206,752,748 | U_kgDODFLL7A | User | false |
openclaw/openclaw | 4,108,131,497 | I_kwDOQb6kR8703Ryp | 51,076 | https://github.com/openclaw/openclaw/issues/51076 | https://api.github.com/repos/openclaw/openclaw/issues/51076 | `openclaw agent --json` output polluted by plugin loading logs | ## Description
When running `openclaw agent --agent <id> --json`, plugin loading messages are printed to stdout, corrupting the JSON output that downstream consumers expect.
## Steps to Reproduce
1. Install a plugin that logs during load (e.g. memory-lancedb-pro)
2. Run: `openclaw agent --agent some-agent --message ... | closed | completed | true | 3 | [] | [] | 2026-03-20T14:13:52Z | 2026-04-30T04:52:52Z | 2026-04-25T05:04:03Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Peters-Pans | 69,051,746 | MDQ6VXNlcjY5MDUxNzQ2 | User | false |
openclaw/openclaw | 4,108,294,653 | I_kwDOQb6kR87035n9 | 51,093 | https://github.com/openclaw/openclaw/issues/51093 | https://api.github.com/repos/openclaw/openclaw/issues/51093 | Show & Tell: tunelo — expose localhost to internet, built for agents | Built a skill for OpenClaw that lets agents expose local services and files to the internet.
**Install:** `clawhub install tunelo`
**What the agent does:**
- User says "share these files" → `tunelo serve .` → public URL with file browser
- User says "test on my phone" → `tunelo http 3000` → public URL to dev server
-... | closed | not_planned | true | 2 | [] | [] | 2026-03-20T14:41:42Z | 2026-04-30T04:52:49Z | 2026-04-25T05:04:24Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | jiweiyuan | 23,147,684 | MDQ6VXNlcjIzMTQ3Njg0 | User | false |
openclaw/openclaw | 4,108,655,565 | I_kwDOQb6kR8705RvN | 51,116 | https://github.com/openclaw/openclaw/issues/51116 | https://api.github.com/repos/openclaw/openclaw/issues/51116 | Discord WebSocket disconnects every ~10 minutes, messages lost during reconnect window | ## Summary
Discord WebSocket connection drops approximately every 10 minutes on a stable wired 1Gbps fiber connection. Messages sent during the ~30 second reconnect window are silently lost — no retry, no delivery after reconnect.
## Environment
- OpenClaw version: 2026.3.13
- OS: macOS (Darwin 25.3.0, arm64)
- Gate... | closed | completed | true | 5 | [] | [] | 2026-03-20T15:51:42Z | 2026-04-30T04:52:48Z | 2026-04-25T05:04:38Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | papiofficial | 72,460,442 | MDQ6VXNlcjcyNDYwNDQy | User | false |
openclaw/openclaw | 4,108,676,906 | I_kwDOQb6kR8705W8q | 51,118 | https://github.com/openclaw/openclaw/issues/51118 | https://api.github.com/repos/openclaw/openclaw/issues/51118 | [Feature]: Feature Request: Add disabledTools configuration option | ### Summary
## Problem
When running agents with specific file-operation policies, structural enforcement is better than behavioral rules.
Currently, there's no way to remove tools from the available function set. This requires relying on agent discipline rather than system-level enforcement.
## Request
Add a `disa... | closed | completed | true | 2 | [
"enhancement"
] | [] | 2026-03-20T15:56:16Z | 2026-04-30T04:52:47Z | 2026-04-25T05:04:45Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | ozzydiaz | 202,164,872 | U_kgDODAzKiA | User | false |
openclaw/openclaw | 4,108,817,983 | I_kwDOQb6kR87055Y_ | 51,132 | https://github.com/openclaw/openclaw/issues/51132 | https://api.github.com/repos/openclaw/openclaw/issues/51132 | Feishu channel: Messages not sent when Reasoning mode is enabled | ## Bug Description
When Reasoning mode is enabled, messages sent by the agent are not delivered to Feishu. The messages are lost.
## Environment
| Component | Version |
|-----------|---------|
| OpenClaw | 2026.3.13 |
| Feishu Plugin | 2026.3.13 |
| Feishu Config | streaming=true, footer.elapsed=true, footer.status=... | closed | completed | true | 2 | [] | [] | 2026-03-20T16:26:59Z | 2026-04-30T04:52:45Z | 2026-04-25T05:05:07Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | suryadi62-ai | 269,139,638 | U_kgDOEAq-tg | User | false |
openclaw/openclaw | 4,108,984,020 | I_kwDOQb6kR8706h7U | 51,149 | https://github.com/openclaw/openclaw/issues/51149 | https://api.github.com/repos/openclaw/openclaw/issues/51149 | Bug: raw internal tool/runtime errors leak into chat as user-visible replies | ## Summary
OpenClaw is sometimes surfacing raw internal tool/runtime plumbing messages directly into user chat instead of suppressing or rewriting them into safe, human-readable replies.
This is happening frequently enough in production chat usage to be a systemic UX/runtime bug, not an isolated incident.
## Examples... | closed | completed | true | 2 | [] | [] | 2026-03-20T16:59:29Z | 2026-04-30T04:52:42Z | 2026-04-25T05:05:28Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | joesinvestments | 257,641,472 | U_kgDOD1tMAA | User | false |
openclaw/openclaw | 4,109,115,116 | I_kwDOQb6kR8707B7s | 51,164 | https://github.com/openclaw/openclaw/issues/51164 | https://api.github.com/repos/openclaw/openclaw/issues/51164 | Credential leak: maherucifer/mersal skill ships with live moltbook API key in .env | ## Summary
The skills archive backup includes a live API key committed directly in a `.env` file:
**Path:** `skills/maherucifer/mersal/.env`
**Contents:** `MOLTBOOK_API_KEY=moltbook_sk_REDACTED`
This key appears to be a real, active credential for the [moltbook](https://www.moltbook.com) platform (bearer token for... | closed | not_planned | true | 6 | [] | [] | 2026-03-20T17:25:36Z | 2026-04-30T04:52:39Z | 2026-04-25T05:05:49Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Pro777 | 164,695 | MDQ6VXNlcjE2NDY5NQ== | User | false |
openclaw/openclaw | 4,109,206,562 | I_kwDOQb6kR8707YQi | 51,171 | https://github.com/openclaw/openclaw/issues/51171 | https://api.github.com/repos/openclaw/openclaw/issues/51171 | Bug: Voice/audio messages via Telegram delivered twice — causes duplicate agent replies | ## Description
When a voice message or audio clip is sent via Telegram, OpenClaw delivers it to the agent twice:
1. Once as a media attachment
2. Once as a transcript insert (after STT processing)
This causes the agent to respond twice to the same message.
## Steps to Reproduce
1. Send a voice message via Telegram ... | closed | not_planned | true | 6 | [] | [] | 2026-03-20T17:42:53Z | 2026-04-30T04:52:38Z | 2026-04-25T05:05:56Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | ramshenoy | 268,190,902 | U_kgDOD_xEtg | User | false |
openclaw/openclaw | 4,109,311,374 | I_kwDOQb6kR8707x2O | 51,177 | https://github.com/openclaw/openclaw/issues/51177 | https://api.github.com/repos/openclaw/openclaw/issues/51177 | [Feature] Add model list filtering to hide unconfigured providers | ### Summary
Control UI displays 20+ model providers in the model selector, but most users only configure 2-3 providers. This creates visual clutter and makes it harder to find configured models.
### Problem to solve
The model selector in Control UI (127.0.0.1:18789) shows all available providers from the hardcoded c... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-03-20T18:03:25Z | 2026-04-30T04:52:38Z | 2026-04-25T05:06:03Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | netmanno1-maker | 230,789,175 | U_kgDODcGQNw | User | false |
openclaw/openclaw | 4,109,592,521 | I_kwDOQb6kR87082fJ | 51,190 | https://github.com/openclaw/openclaw/issues/51190 | https://api.github.com/repos/openclaw/openclaw/issues/51190 | [Bug]: Discord channel can show running=true while connected=false, and new DMs stop reaching the agent until gateway restart | ### Bug type
Crash (process/app exits or hangs)
### Summary
On OpenClaw 2026.3.13 on macOS, the Discord channel can report running: true while the Discord account reports connected: false, and new Discord DMs do not reach the agent until the gateway is restarted.
### Steps to reproduce
1. Start OpenClaw 2026.3.13... | closed | completed | true | 2 | [
"bug",
"bug:crash"
] | [] | 2026-03-20T18:55:13Z | 2026-04-30T04:52:36Z | 2026-04-25T05:06:17Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | MLachance42 | 155,120,594 | U_kgDOCT7z0g | User | false |
openclaw/openclaw | 4,109,750,714 | I_kwDOQb6kR8709dG6 | 51,203 | https://github.com/openclaw/openclaw/issues/51203 | https://api.github.com/repos/openclaw/openclaw/issues/51203 | [Feature]: Feature Request: Immutable Approval Gate (Human-in-the-Loop execution layer) | ### Summary
Current agents lose safety instructions under context compression — as seen with Summer Yue's email deletion incident.
Theia Guard proposes an approval gate that operates outside the agent's context entirely.
Concept and architecture: https://github.com/ismailkarabulut-lang/theia-guard
### Problem to solv... | closed | completed | true | 3 | [
"enhancement"
] | [] | 2026-03-20T19:33:07Z | 2026-04-30T04:52:35Z | 2026-04-25T05:06:24Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | ismailkarabulut-lang | 235,378,335 | U_kgDODgeWnw | User | false |
openclaw/openclaw | 4,109,796,535 | I_kwDOQb6kR8709oS3 | 51,209 | https://github.com/openclaw/openclaw/issues/51209 | https://api.github.com/repos/openclaw/openclaw/issues/51209 | Fallback chain does not cascade on HTTP 401/404 provider errors | ## Problem
When the primary model provider returns HTTP errors (401 authentication\_error, 404 not\_found), OpenClaw retries the **same broken provider** repeatedly instead of cascading to the next model in the fallback chain. This effectively causes the session to hang until timeout (600s default).
## Expected Behav... | closed | completed | true | 5 | [] | [] | 2026-03-20T19:43:23Z | 2026-04-30T04:52:35Z | 2026-04-25T05:06:31Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | logos-taranov | 262,328,755 | U_kgDOD6LRsw | User | false |
openclaw/openclaw | 4,109,915,306 | I_kwDOQb6kR870-FSq | 51,255 | https://github.com/openclaw/openclaw/issues/51255 | https://api.github.com/repos/openclaw/openclaw/issues/51255 | [Feature]: onSessionEnd hook for auto-compaction and cleanup tasks | ## Feature Request
### Problem
When a session ends (via `/new`, `/compact`, disconnect, or timeout), there is currently no way for agents or hooks to perform cleanup tasks automatically. This means:
- Context compaction only happens reactively (when context is too large) or manually (`/compact`)
- Memory flush to fi... | closed | completed | true | 2 | [] | [] | 2026-03-20T20:13:28Z | 2026-04-30T04:52:33Z | 2026-04-25T05:06:45Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Maw-bot | 267,040,691 | U_kgDOD-q3sw | User | false |
openclaw/openclaw | 4,110,036,935 | I_kwDOQb6kR870-i_H | 51,266 | https://github.com/openclaw/openclaw/issues/51266 | https://api.github.com/repos/openclaw/openclaw/issues/51266 | Gateway crashes on startup with bad hook entries — graceful error handling needed | If a user adds a hook entry to `openclaw.json` that points to a non-existent path, the gateway crashes on startup with an unhandled `MODULE_NOT_FOUND` error. The error message doesn't indicate which hook entry is bad or how to fix it.
This is common with new users who clone workspace repos containing hook references a... | closed | completed | true | 2 | [] | [] | 2026-03-20T20:47:17Z | 2026-04-30T04:52:32Z | 2026-04-25T05:06:53Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | wiziswiz | 83,570,252 | MDQ6VXNlcjgzNTcwMjUy | User | false |
openclaw/openclaw | 4,110,078,965 | I_kwDOQb6kR870-tP1 | 51,273 | https://github.com/openclaw/openclaw/issues/51273 | https://api.github.com/repos/openclaw/openclaw/issues/51273 | Loopback gateway RPC fails with 'missing scope: operator.read' — shouldAttachDeviceIdentityForGatewayCall drops device identity on loopback | ## Summary
On a machine where the OpenClaw gateway is running locally and the device is paired, operator-scoped gateway RPC calls (`openclaw status`, `config.get`, `system-presence`, etc.) fail with:
```
unreachable (missing scope: operator.read)
```
This happens even when the local operator device token has the cor... | closed | completed | true | 2 | [] | [] | 2026-03-20T20:58:07Z | 2026-04-30T04:52:32Z | 2026-04-25T05:07:00Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | mband0 | 31,228,859 | MDQ6VXNlcjMxMjI4ODU5 | User | false |
openclaw/openclaw | 4,110,085,478 | I_kwDOQb6kR870-u1m | 51,274 | https://github.com/openclaw/openclaw/issues/51274 | https://api.github.com/repos/openclaw/openclaw/issues/51274 | CLI handshake timeout (3s) too short — causes silent 1000 close on loaded gateways | ## Problem
The CLI WebSocket handshake timeout is hardcoded to 3 seconds (DEFAULT_HANDSHAKE_TIMEOUT_MS = 3e3 in gateway-cli-*.js). On loaded gateways (600MB+ memory, 80+ tasks, multiple active channels), the gateway consistently fails to complete the handshake within 3s.
**Result:** Every CLI command fails with "gate... | closed | completed | true | 3 | [] | [] | 2026-03-20T20:59:39Z | 2026-04-30T04:52:31Z | 2026-04-25T05:07:07Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | cygnostik | 5,272,292 | MDQ6VXNlcjUyNzIyOTI= | User | false |
openclaw/openclaw | 4,110,098,734 | I_kwDOQb6kR870-yEu | 51,275 | https://github.com/openclaw/openclaw/issues/51275 | https://api.github.com/repos/openclaw/openclaw/issues/51275 | [Bug]: Sandbox security errors posted as channel messages — leaks internal paths and PII | ### Summary
When an agent fails to start due to a sandbox security violation (e.g., a bind mount source outside allowed roots), the full error message is posted to the configured channel as if it were the agent's reply. This leaks internal file paths, usernames, config file locations, and potentially PII to public-fac... | closed | completed | true | 3 | [] | [] | 2026-03-20T21:02:43Z | 2026-04-30T04:52:30Z | 2026-04-25T05:07:14Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | alildizzy | 264,213,712 | U_kgDOD7-U0A | User | false |
openclaw/openclaw | 4,110,449,787 | I_kwDOQb6kR871AHx7 | 51,310 | https://github.com/openclaw/openclaw/issues/51310 | https://api.github.com/repos/openclaw/openclaw/issues/51310 | docs: add dedicated secrets / exec-provider page | ## Summary
The docs currently have no dedicated page for OpenClaw's secret provider system. Coverage is scattered across config references and setup guides. As the community builds exec provider integrations (e.g. Azure Key Vault, macOS Keychain, 1Password), there is no clear place to discover, document, or list them.... | closed | completed | true | 1 | [] | [] | 2026-03-20T22:41:54Z | 2026-04-30T04:52:29Z | 2026-04-25T05:07:28Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Arry8 | 263,019,747 | U_kgDOD61c4w | User | false |
openclaw/openclaw | 4,110,617,782 | I_kwDOQb6kR871Awy2 | 51,328 | https://github.com/openclaw/openclaw/issues/51328 | https://api.github.com/repos/openclaw/openclaw/issues/51328 | [Bug]: ACP session transcript exists but session APIs may return empty history | Title: [Bug]: ACP session transcript exists but session APIs may return empty history
Bug type
Behavior bug (incorrect output/state without crash)
Summary
In observed Claude ACP one-shot runs, `sessions_history` and `sessions_list(...messageLimit=...)` returned an empty session even though the raw ACP transcript JSON... | closed | completed | true | 3 | [] | [] | 2026-03-20T23:42:56Z | 2026-04-30T04:52:28Z | 2026-04-25T05:07:35Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | kyGaribaldi | 84,607,777 | MDQ6VXNlcjg0NjA3Nzc3 | User | false |
openclaw/openclaw | 4,110,625,616 | I_kwDOQb6kR871AytQ | 51,330 | https://github.com/openclaw/openclaw/issues/51330 | https://api.github.com/repos/openclaw/openclaw/issues/51330 | Handle stale session .jsonl.lock files more gracefully | ## Summary
OpenClaw can leave stale `*.jsonl.lock` files behind in agent session stores after interrupted or failed runs, and there does not appear to be a built-in stale-lock recovery path on startup.
## Why this matters
When this happens, later local-agent runs can fail with session-store lock errors even though the... | closed | completed | true | 2 | [] | [] | 2026-03-20T23:46:05Z | 2026-04-30T04:52:27Z | 2026-04-25T05:07:42Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | emilclaw | 269,614,170 | U_kgDOEBH8Wg | User | false |
openclaw/openclaw | 4,110,703,462 | I_kwDOQb6kR871BFtm | 51,342 | https://github.com/openclaw/openclaw/issues/51342 | https://api.github.com/repos/openclaw/openclaw/issues/51342 | Feature request: Make staleEventThresholdMs configurable (WhatsApp restart loop on low-traffic accounts) | ## Summary
OpenClaw's WhatsApp channel health monitor restarts the WhatsApp Web socket **every ~35 minutes** on personal/low-traffic accounts. This creates a perpetual restart cycle that causes cron job delivery failures when jobs fire during the brief reconnection window. We observed **4,004 stale-socket restarts in ... | closed | completed | true | 2 | [] | [] | 2026-03-21T00:14:38Z | 2026-04-30T04:52:26Z | 2026-04-25T05:07:56Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | raistlin88 | 26,916,023 | MDQ6VXNlcjI2OTE2MDIz | User | false |
openclaw/openclaw | 4,110,796,915 | I_kwDOQb6kR871Bchz | 51,351 | https://github.com/openclaw/openclaw/issues/51351 | https://api.github.com/repos/openclaw/openclaw/issues/51351 | Stabilize Slack thread-root image inheritance | ## Summary
Ensure image-only/file-only Slack thread root messages reliably propagate media to later thread replies.
## Why
A key user flow is: root message contains an image/file with little or no text, then a later thread reply asks the agent to inspect it.
## Tasks
- review existing local changes in Slack media/thr... | closed | completed | true | 1 | [] | [] | 2026-03-21T00:43:43Z | 2026-04-30T04:52:25Z | 2026-04-25T05:08:03Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | UGDev-Chane | 180,237,537 | U_kgDOCr404Q | User | false |
openclaw/openclaw | 4,110,797,075 | I_kwDOQb6kR871BckT | 51,352 | https://github.com/openclaw/openclaw/issues/51352 | https://api.github.com/repos/openclaw/openclaw/issues/51352 | Validate Slack multi-image and forwarded attachment handling | ## Summary
Confirm and tighten support for multi-image Slack messages and forwarded/shared Slack attachments.
## Why
Slack image workflows are not only single-image direct uploads. We need to understand whether downstream processing uses all media entries and whether forwarded/shared attachments survive normalization.... | closed | completed | true | 1 | [] | [] | 2026-03-21T00:43:44Z | 2026-04-30T04:52:24Z | 2026-04-25T05:08:10Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | UGDev-Chane | 180,237,537 | U_kgDOCr404Q | User | false |
openclaw/openclaw | 4,110,797,108 | I_kwDOQb6kR871Bck0 | 51,353 | https://github.com/openclaw/openclaw/issues/51353 | https://api.github.com/repos/openclaw/openclaw/issues/51353 | Add Slack attachment vision regression tests | ## Summary
Add targeted regression tests for Slack image/file vision flows.
## Tasks
- direct Slack image attachment test
- image-only/file-only root thread + reply test
- multi-image Slack message test
- forwarded/shared Slack attachment test where practical
- inaccessible/invalid attachment failure-path test
- run t... | closed | completed | true | 1 | [] | [] | 2026-03-21T00:43:45Z | 2026-04-30T04:52:24Z | 2026-04-25T05:08:17Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | UGDev-Chane | 180,237,537 | U_kgDOCr404Q | User | false |
openclaw/openclaw | 4,110,946,919 | I_kwDOQb6kR871CBJn | 51,369 | https://github.com/openclaw/openclaw/issues/51369 | https://api.github.com/repos/openclaw/openclaw/issues/51369 | Feishu file upload: sanitizeFileNameForUpload causes garbled Chinese filenames | ## Bug Description
When sending files with Chinese filenames via the Feishu channel, the uploaded file displays with a garbled name (percent-encoded). For example, `测试中文.pdf` appears as `%E6%B5%8B%E8%AF%95%E4%B8%AD%E6%96%87.pdf`.
## Root Cause
In `extensions/feishu/src/media.ts`, the `sanitizeFileNameForUpload` func... | closed | completed | true | 4 | [] | [] | 2026-03-21T01:43:16Z | 2026-04-30T04:52:22Z | 2026-04-25T05:08:31Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | DavidDing16 | 20,034,738 | MDQ6VXNlcjIwMDM0NzM4 | User | false |
openclaw/openclaw | 4,110,948,587 | I_kwDOQb6kR871CBjr | 51,370 | https://github.com/openclaw/openclaw/issues/51370 | https://api.github.com/repos/openclaw/openclaw/issues/51370 | [Bug]: Discord plugin does not reconnect after network outage during initial startup | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When the Discord plugin's initial startup occurs while the network is unavailable (e.g., ISP outage), all API calls (`fetchDiscordGatewayInfo`, `deployDiscordCommands`, `fetchUser("@me")`) fail with `fetch failed`. The plugin logs "logged i... | closed | completed | true | 4 | [] | [] | 2026-03-21T01:44:23Z | 2026-04-30T04:52:21Z | 2026-04-25T05:08:38Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | bhooker-github | 52,179,622 | MDQ6VXNlcjUyMTc5NjIy | User | false |
openclaw/openclaw | 4,111,054,583 | I_kwDOQb6kR871Cbb3 | 51,382 | https://github.com/openclaw/openclaw/issues/51382 | https://api.github.com/repos/openclaw/openclaw/issues/51382 | [Bug]: Web Control UI cron job badge shows incorrect count (4 instead of 3) | ### Bug type
Crash (process/app exits or hangs)
### Summary
# Bug Report: Web Control UI Cron Job Count Mismatch
## Summary
Web Control UI displays incorrect cron job count in the top-right corner badge.
## Environment
- **OpenClaw Version:** 2026.3.13
- **OS:** Windows 11 (Build 10.0.26200)
- **Node:** v24.14.0
-... | closed | not_planned | true | 3 | [
"bug",
"bug:crash"
] | [] | 2026-03-21T02:31:07Z | 2026-04-30T04:52:20Z | 2026-04-25T05:08:45Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Chopin-1008 | 213,917,579 | U_kgDODMAfiw | User | false |
openclaw/openclaw | 4,111,297,742 | I_kwDOQb6kR871DWzO | 51,410 | https://github.com/openclaw/openclaw/issues/51410 | https://api.github.com/repos/openclaw/openclaw/issues/51410 | Duplicate tool call ID error with Kimi/moonshot model | # Bug Report: Duplicate Tool Call ID Error with Kimi/moonshot Model
## Summary
OpenClaw throws "duplicated tool call ID" error when using moonshot/kimi-k2.5 model, causing HTTP 400 errors and failed responses.
## Environment
- **OpenClaw Version:** 2026.3.13 (61d171a)
- **Model:** moonshot/kimi-k2.5
- **OS:** Ubuntu ... | closed | completed | true | 2 | [] | [] | 2026-03-21T03:59:51Z | 2026-04-30T04:52:18Z | 2026-04-25T05:09:06Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | arvil22 | 107,420,474 | U_kgDOBmcbOg | User | false |
openclaw/openclaw | 4,111,410,263 | I_kwDOQb6kR871DyRX | 51,423 | https://github.com/openclaw/openclaw/issues/51423 | https://api.github.com/repos/openclaw/openclaw/issues/51423 | Compaction breaks tool_use/tool_result pairing — API rejects orphaned tool_result | ## Summary
Session compaction (or context pruning) can remove a `tool_use` block while leaving its corresponding `tool_result` in the message history. Anthropic's API then rejects the request with:
```
messages.80.content.1: unexpected tool_use_id found in tool_result blocks: <id>.
Each tool_result block must have a ... | closed | completed | true | 4 | [] | [] | 2026-03-21T04:48:51Z | 2026-04-30T04:52:17Z | 2026-04-25T05:09:13Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | coywolffuturist | 264,284,025 | U_kgDOD8CneQ | User | false |
openclaw/openclaw | 4,111,413,250 | I_kwDOQb6kR871DzAC | 51,424 | https://github.com/openclaw/openclaw/issues/51424 | https://api.github.com/repos/openclaw/openclaw/issues/51424 | Feature: Support passing slash commands to ACP harness sessions | ## Problem
When using ACP sessions (Claude Code, Codex, etc.) via Discord, users cannot send slash commands like `/help` or `/compact` to the harness because:
1. Discord intercepts `/` prefixed messages as bot commands
2. Quoting the command (`'/help'`) passes it literally to the harness, which rejects the malformed ... | closed | completed | true | 2 | [] | [] | 2026-03-21T04:50:32Z | 2026-04-30T04:52:16Z | 2026-04-25T05:09:20Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | rollingstone87 | 194,603,715 | U_kgDOC5lqww | User | false |
openclaw/openclaw | 4,111,440,313 | I_kwDOQb6kR871D5m5 | 51,428 | https://github.com/openclaw/openclaw/issues/51428 | https://api.github.com/repos/openclaw/openclaw/issues/51428 | [Feature]: While hovering Lobster shadow became blue but it will look more better if it will be red shadow | ### Summary
While hovering Lobster shadow became blue but it will look more better if it will be red shadow
### Problem to solve
by changing shadow color of Lobster to red #FF4D4D
### Proposed solution
by changing shadow color of Lobster to red #FF4D4D
### Alternatives considered
_No response_
### Impact
... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-03-21T05:03:55Z | 2026-04-30T04:52:15Z | 2026-04-25T05:09:27Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | aadityakumarsah | 148,959,473 | U_kgDOCODw8Q | User | false |
openclaw/openclaw | 4,111,485,243 | I_kwDOQb6kR871EEk7 | 51,430 | https://github.com/openclaw/openclaw/issues/51430 | https://api.github.com/repos/openclaw/openclaw/issues/51430 | Feature Request: Cron Job Message Context Preservation | # Feature Request: Cron Job Message Context Preservation
## Problem Statement
When a cron job sends a message to the user, the agent (AI assistant) cannot see the original message content in the session context. This creates a **context gap** when the user replies to the cron job with references like "execute this su... | closed | completed | true | 3 | [] | [] | 2026-03-21T05:27:54Z | 2026-04-30T04:52:15Z | 2026-04-25T05:09:34Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | foreverpupu | 154,308,979 | U_kgDOCTKRcw | User | false |
openclaw/openclaw | 4,111,515,162 | I_kwDOQb6kR871EL4a | 51,437 | https://github.com/openclaw/openclaw/issues/51437 | https://api.github.com/repos/openclaw/openclaw/issues/51437 | [Feature]: Hi, just wondering why there haven't been any updates this past week? | ### Summary
Hi, just wondering why there haven't been any updates this past week?
### Problem to solve
Hi, just wondering why there haven't been any updates this past week?
### Proposed solution
Hi, just wondering why there haven't been any updates this past week?
### Alternatives considered
Hi, just wondering w... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-03-21T05:40:51Z | 2026-04-30T04:52:13Z | 2026-04-25T05:09:55Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | weg123159 | 203,663,910 | U_kgDODCOqJg | User | false |
openclaw/openclaw | 4,111,522,672 | I_kwDOQb6kR871ENtw | 51,438 | https://github.com/openclaw/openclaw/issues/51438 | https://api.github.com/repos/openclaw/openclaw/issues/51438 | CLI → Gateway WebSocket connections timeout in v2026.3.12+ (Regression) | ### Bug type
Regression (worked before, now fails)
### Summary
After upgrading from v2026.3.11 to v2026.3.12 or v2026.3.13, all CLI commands that connect to the gateway (openclaw logs, openclaw status, etc.) fail with WebSocket connection timeouts in the 3–10 second range. Downgrading to v2026.3.11 resolves the issu... | closed | completed | true | 5 | [
"bug",
"regression"
] | [] | 2026-03-21T05:44:21Z | 2026-04-30T04:52:12Z | 2026-04-25T05:10:02Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | zoe-ludvigh | 266,191,022 | U_kgDOD93Arg | User | false |
openclaw/openclaw | 4,111,527,922 | I_kwDOQb6kR871EO_y | 51,439 | https://github.com/openclaw/openclaw/issues/51439 | https://api.github.com/repos/openclaw/openclaw/issues/51439 | [Bug]: UI Model selection Normalisation & Deduplication | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
HI I have only two models but the chat selection shows 4.
### Steps to reproduce
<img width="618" height="179" alt="Image" src="https://github.com/user-attachments/assets/249011ce-73d3-40f2-bd1c-9a8927454a1f" />
### Expected behavior
2... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-21T05:46:49Z | 2026-04-30T04:52:11Z | 2026-04-25T05:10:09Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | kingmaxmmxweisun | 158,125,378 | U_kgDOCWzNQg | User | false |
openclaw/openclaw | 4,111,531,105 | I_kwDOQb6kR871EPxh | 51,440 | https://github.com/openclaw/openclaw/issues/51440 | https://api.github.com/repos/openclaw/openclaw/issues/51440 | Control UI and TUI chat crash with TypeError: Cannot read properties of undefined (reading 'map') after sending a message | ## Summary
Both **Control UI chat** and **`openclaw tui`** crash with the same error after sending a message:
```text
TypeError: Cannot read properties of undefined (reading 'map')
```
This is reproducible from the chat flow itself (not just page load / initial render).
## Reproduction
### Control UI
1. Open Contr... | closed | completed | true | 4 | [] | [] | 2026-03-21T05:48:06Z | 2026-04-30T04:52:11Z | 2026-04-25T05:10:16Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | jaVVVid | 267,270,014 | U_kgDOD-43fg | User | false |
openclaw/openclaw | 4,111,547,809 | I_kwDOQb6kR871ET2h | 51,447 | https://github.com/openclaw/openclaw/issues/51447 | https://api.github.com/repos/openclaw/openclaw/issues/51447 | Discord: asVoice conflicts with required components schema — cannot send voice messages with pre-recorded audio | ## Bug
The `message` tool requires `components` for Discord sends, but Discord voice messages (`asVoice: true`) cannot have components. This makes it **impossible to send pre-recorded audio as inline playable voice messages in Discord**.
## Steps to Reproduce
1. Try sending a voice message with pre-recorded audio:
... | closed | completed | true | 3 | [] | [] | 2026-03-21T05:58:26Z | 2026-04-30T04:52:10Z | 2026-04-25T05:10:23Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | cyperx84 | 47,339,338 | MDQ6VXNlcjQ3MzM5MzM4 | User | false |
openclaw/openclaw | 4,111,564,671 | I_kwDOQb6kR871EX9_ | 51,452 | https://github.com/openclaw/openclaw/issues/51452 | https://api.github.com/repos/openclaw/openclaw/issues/51452 | [Bug] Persistent HTTP 401 with Valid Aliyun ModelStudio Key (Direct API Calls Work, CLI Fails) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
## 🐛 Describe the Bug
I am encountering a persistent `HTTP 401: Incorrect API key provided` error when running OpenClaw with Aliyun ModelStudio (DashScope), despite the API key being fully valid and functional in direct API calls.
The iss... | closed | completed | true | 5 | [
"bug",
"bug:behavior"
] | [] | 2026-03-21T06:05:45Z | 2026-04-30T04:52:09Z | 2026-04-25T05:10:30Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Arsnature2026 | 263,112,573 | U_kgDOD67HfQ | User | false |
openclaw/openclaw | 4,111,565,984 | I_kwDOQb6kR871EYSg | 51,453 | https://github.com/openclaw/openclaw/issues/51453 | https://api.github.com/repos/openclaw/openclaw/issues/51453 | [Bug]: config_guard falsely flags models.providers.zai.models token refresh as breaking change every hour | ### Bug type
Regression (worked before, now fails)
### Summary
config_guard detects hourly JWT token refresh in models.providers.zai.models as a breaking config change (isBreaking: true), triggering a warning dialog and preventing Gateway from starting. The only difference between .bak and .known-good files is the J... | closed | not_planned | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-21T06:06:24Z | 2026-04-30T04:52:08Z | 2026-04-25T05:10:37Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | HenryFredrich | 60,771,929 | MDQ6VXNlcjYwNzcxOTI5 | User | false |
openclaw/openclaw | 4,111,567,657 | I_kwDOQb6kR871EYsp | 51,454 | https://github.com/openclaw/openclaw/issues/51454 | https://api.github.com/repos/openclaw/openclaw/issues/51454 | [Resource] Awesome OpenClaw Tutorials & Tools Collection - Share Your Favorites! | ## 📚 OpenClaw Learning Resources
Hey OpenClaw community! Let's build a curated list of tutorials, tools, and resources for the ecosystem.
### 🔥 Must-Read Tutorials
- [Official Docs](https://docs.openclaw.ai)
- [OpenClaw GitHub](https://github.com/openclaw/openclaw)
### 🛠️ Essential Tools
- **OpenClaw Skills** - E... | closed | not_planned | true | 1 | [] | [] | 2026-03-21T06:07:10Z | 2026-04-30T04:52:08Z | 2026-04-25T05:10:44Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | jingchang0623-crypto | 237,525,016 | U_kgDODihYGA | User | false |
openclaw/openclaw | 4,111,586,008 | I_kwDOQb6kR871EdLY | 51,456 | https://github.com/openclaw/openclaw/issues/51456 | https://api.github.com/repos/openclaw/openclaw/issues/51456 | Feature Request: Cancel/Undo window for incoming messages | ## Feature Request: Cancel/Undo Window for Incoming Messages
**Submitted by:** @jardysuntan
### Problem
When a message is accidentally sent to the bot (typo, wrong channel, premature send), there is no way to cancel it before the agent starts processing. This is especially costly with expensive models like Opus where... | closed | not_planned | true | 2 | [] | [] | 2026-03-21T06:14:39Z | 2026-04-30T04:52:07Z | 2026-04-25T05:10:51Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | jardysuntan | 262,049,598 | U_kgDOD56PPg | User | false |
openclaw/openclaw | 4,111,592,937 | I_kwDOQb6kR871Ee3p | 51,459 | https://github.com/openclaw/openclaw/issues/51459 | https://api.github.com/repos/openclaw/openclaw/issues/51459 | Feature: Auto-LCM recall enforcement after context compaction | # Feature Request: Auto-LCM Recall Enforcement After Compaction
## Problem
Agents frequently fail to use LCM tools (`lcm_grep`, `lcm_expand_query`) after context compaction, even when explicit rules exist in AGENTS.md. This leads to:
- Incorrect responses based on incomplete summaries
- Repeated mistakes that were d... | closed | not_planned | true | 2 | [] | [] | 2026-03-21T06:18:27Z | 2026-04-30T04:52:06Z | 2026-04-25T05:10:58Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | jonoboyer | 9,364,810 | MDQ6VXNlcjkzNjQ4MTA= | User | false |
openclaw/openclaw | 4,111,605,702 | I_kwDOQb6kR871Eh_G | 51,461 | https://github.com/openclaw/openclaw/issues/51461 | https://api.github.com/repos/openclaw/openclaw/issues/51461 | Feishu DM stop/停止 does not preempt active run; abort appears delayed until dispatch completes | ## Summary
In Feishu direct-message chats, sending `stop`, `/stop`, or `停止` during an active run does not interrupt promptly. In practice, the stop message appears to be processed only after the current dispatch finishes.
This does not match the documented out-of-band abort behavior.
## Docs / expected behavior
The d... | closed | completed | true | 5 | [] | [] | 2026-03-21T06:24:33Z | 2026-04-30T04:52:05Z | 2026-04-25T05:11:05Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | chengnanlxg | 127,652,346 | U_kgDOB5vR-g | User | false |
openclaw/openclaw | 4,111,607,516 | I_kwDOQb6kR871Eibc | 51,463 | https://github.com/openclaw/openclaw/issues/51463 | https://api.github.com/repos/openclaw/openclaw/issues/51463 | Gateway self-SIGTERMs after ~45s: server.mjs device approval timeout kills healthy gateway | ## Bug
On container restart, `server.mjs` spawns the gateway process but kills it with SIGTERM after ~45 seconds if "device approval" does not complete in time. This causes a crash loop when combined with a watchdog auto-restart.
## Environment
- OpenClaw 2026.3.13 (61d171a)
- Docker (Hostinger KVM VPS)
- WhatsApp + ... | closed | not_planned | true | 4 | [] | [] | 2026-03-21T06:25:43Z | 2026-04-30T04:52:05Z | 2026-04-25T05:11:12Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | mids-neo | 117,532,760 | U_kgDOBwFoWA | User | false |
openclaw/openclaw | 4,111,615,286 | I_kwDOQb6kR871EkU2 | 51,466 | https://github.com/openclaw/openclaw/issues/51466 | https://api.github.com/repos/openclaw/openclaw/issues/51466 | [Bug] Telegram /tts audio reaches TTS path but Edge provider times out; no voice note delivered on v2026.3.13 | ## Summary
On OpenClaw v2026.3.13, Telegram TTS does not successfully return a voice note in my environment.
The key point is:
- TTS is not completely unsupported on Telegram
- /tts audio clearly enters the TTS path
- but the free Edge/Microsoft provider times out, so no audio is produced and no Telegram voice note i... | closed | completed | true | 2 | [] | [] | 2026-03-21T06:30:14Z | 2026-04-30T04:52:04Z | 2026-04-25T05:11:19Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | spadesv | 30,253,580 | MDQ6VXNlcjMwMjUzNTgw | User | false |
openclaw/openclaw | 4,111,629,697 | I_kwDOQb6kR871En2B | 51,467 | https://github.com/openclaw/openclaw/issues/51467 | https://api.github.com/repos/openclaw/openclaw/issues/51467 | Feature Request: Multi-Account Feishu Routing in Bindings | **Is your feature request related to a problem? Please describe.**
Currently, OpenClaw supports multiple Feishu accounts (bots) via the `channels.feishu.accounts` configuration. However, the `bindings` routing system cannot distinguish which account (bot) received the message. This makes it impossible to route messages... | closed | completed | true | 2 | [] | [] | 2026-03-21T06:37:04Z | 2026-04-30T04:52:03Z | 2026-04-25T05:11:26Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | DoneHome | 10,095,576 | MDQ6VXNlcjEwMDk1NTc2 | User | false |
openclaw/openclaw | 4,326,889,709 | I_kwDOQb6kR88AAAABAecY7Q | 71,380 | https://github.com/openclaw/openclaw/issues/71380 | https://api.github.com/repos/openclaw/openclaw/issues/71380 | [Bug]: Parallel tool results can be replaced by synthetic transcript-repair errors, stalling Discord turn | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
During a Discord-hosted agent turn on OpenClaw `2026.4.23`, a `multi_tool_use.parallel` validation step emitted synthetic transcript-repair tool results instead of the actual child tool results:
```text
[openc... | closed | completed | true | 1 | [] | [] | 2026-04-25T03:23:59Z | 2026-04-30T04:52:01Z | 2026-04-25T05:12:22Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | sherly-techtree | 258,044,850 | U_kgDOD2Fzsg | User | false |
openclaw/openclaw | 4,180,899,540 | I_kwDOQb6kR875M3bU | 58,485 | https://github.com/openclaw/openclaw/issues/58485 | https://api.github.com/repos/openclaw/openclaw/issues/58485 | [Bug]: Isolated cron sessions ignore `payload.model` override — always use agent default model | **OpenClaw version:** 2026.3.28
**OS:** macOS 26.3.1 (Apple Silicon)
**Description:**
When a cron job specifies `payload.model` (e.g., `anthropic/claude-haiku-4-5-20251001`), isolated cron sessions ignore it and use the agent's default model instead.
**Steps to reproduce:**
1. Create a cron job with `sessionTarget: "... | closed | completed | true | 0 | [] | [] | 2026-03-31T17:11:43Z | 2026-04-30T04:51:56Z | 2026-04-25T05:15:22Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | liu51115 | 31,350,941 | MDQ6VXNlcjMxMzUwOTQx | User | false |
openclaw/openclaw | 4,177,692,324 | I_kwDOQb6kR875Aoak | 58,285 | https://github.com/openclaw/openclaw/issues/58285 | https://api.github.com/repos/openclaw/openclaw/issues/58285 | Cron isolated jobs ignore payload.model override — always use agent primary model | ## Bug description
Isolated cron jobs (`sessionTarget: "isolated"`, `payload.kind: "agentTurn"`) ignore the `payload.model` field and always execute with the agent's primary model.
## Expected behavior
Per the [cron-jobs docs](https://docs.openclaw.ai/automation/cron-jobs#model-and-thinking-overrides):
> Resolution... | closed | completed | true | 0 | [] | [] | 2026-03-31T09:47:02Z | 2026-04-30T04:51:57Z | 2026-04-25T05:15:22Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | dotwobot-dev | 264,529,073 | U_kgDOD8RksQ | User | false |
openclaw/openclaw | 4,174,463,060 | I_kwDOQb6kR8740UBU | 57,968 | https://github.com/openclaw/openclaw/issues/57968 | https://api.github.com/repos/openclaw/openclaw/issues/57968 | Cron payload.model override ignored — agent primary model wins via 'live session model switch' | ## Bug
Cron jobs with `payload.model` set to a different model than the agent's primary model always fall back to the agent's primary model. The docs state that job payload override has highest priority, but the code disagrees.
## Environment
- OpenClaw 2026.3.28 (f9b1079)
- Agent primary model: `anthropic/claude-op... | closed | completed | true | 0 | [] | [] | 2026-03-30T22:30:59Z | 2026-04-30T04:51:57Z | 2026-04-25T05:15:22Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | pbanuru | 55,062,649 | MDQ6VXNlcjU1MDYyNjQ5 | User | false |
openclaw/openclaw | 4,211,708,386 | I_kwDOQb6kR877CZHi | 61,879 | https://github.com/openclaw/openclaw/issues/61879 | https://api.github.com/repos/openclaw/openclaw/issues/61879 | Isolated cron agentTurn ignores model override and still executes with agent default model | ## Summary
When creating an isolated cron job with an explicit model override like `--model "ollama/gemma3:12b"`, the cron job metadata correctly shows the overridden model, but the actual run still executes with the agent’s default model (in this case `gpt-5.4` / `openai-codex`).
This makes it impossible to reliably ... | closed | completed | true | 2 | [] | [] | 2026-04-06T12:46:53Z | 2026-04-30T04:51:54Z | 2026-04-25T05:15:23Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | gjamsue | 41,173,877 | MDQ6VXNlcjQxMTczODc3 | User | false |
openclaw/openclaw | 4,182,583,600 | I_kwDOQb6kR875TSkw | 58,575 | https://github.com/openclaw/openclaw/issues/58575 | https://api.github.com/repos/openclaw/openclaw/issues/58575 | [Bug]: --model flag on isolated cron jobs does not override session model | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
`openclaw cron add/edit --model anthropic/claude-haiku-4-5 --session isolated` sets the model correctly in `payload.model`, but the isolated session always initializes with the default model (claude-sonnet-4-6) instead.
###... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-31T21:49:22Z | 2026-04-30T04:51:56Z | 2026-04-25T05:15:23Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | litefut | 62,533,755 | MDQ6VXNlcjYyNTMzNzU1 | User | false |
openclaw/openclaw | 4,181,639,958 | I_kwDOQb6kR875PsMW | 58,533 | https://github.com/openclaw/openclaw/issues/58533 | https://api.github.com/repos/openclaw/openclaw/issues/58533 | Cron sessions inherit persisted model override instead of honoring payload.model, causing thundering herd under overload | ## Summary
Cron jobs that specify `model: openai/gpt-5.4-nano` in their payload are not using that model when they execute in isolated sessions. Instead, `resolvePersistedLiveSelection()` forces the session back to `anthropic/claude-sonnet-4-6`, overriding the cron's declared model intent. Under Anthropic overload con... | closed | completed | true | 1 | [] | [] | 2026-03-31T19:05:46Z | 2026-04-30T04:51:53Z | 2026-04-25T05:15:24Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Eugene9D | 167,811,285 | U_kgDOCgCY1Q | User | false |
openclaw/openclaw | 4,174,161,404 | I_kwDOQb6kR874zKX8 | 57,947 | https://github.com/openclaw/openclaw/issues/57947 | https://api.github.com/repos/openclaw/openclaw/issues/57947 | [Bug]: Cron payload model override ignored — always resolves to agent default | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Cron jobs with payload.model set to a specific model (e.g., openai-codex/gpt-5.3-codex) always resolve to the agents.defaults.model.primary model instead. The payload model override is completely ignored.
Per the docs (auto... | closed | completed | true | 0 | [
"bug",
"regression"
] | [] | 2026-03-30T21:22:40Z | 2026-04-30T04:51:53Z | 2026-04-25T05:15:24Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | jkfaris94 | 155,022,265 | U_kgDOCT1zuQ | User | false |
openclaw/openclaw | 4,207,215,123 | I_kwDOQb6kR876xQIT | 61,294 | https://github.com/openclaw/openclaw/issues/61294 | https://api.github.com/repos/openclaw/openclaw/issues/61294 | [Bug] Cron isolated session model parameter ignored, falls back to default agent model | # [Bug] Cron isolated session `model` parameter ignored, falls back to default agent model
## Bug Type
Regression / Configuration Issue
## Summary
When using `sessionTarget: "isolated"` with `payload.kind: "agentTurn"` in cron jobs, the `payload.model` parameter is ignored at runtime. The job always uses `agents.defa... | closed | completed | true | 1 | [] | [] | 2026-04-05T10:40:50Z | 2026-04-30T04:51:50Z | 2026-04-25T05:15:25Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | MajorCommotion | 273,761,552 | U_kgDOEFFFEA | User | false |
openclaw/openclaw | 4,181,227,060 | I_kwDOQb6kR875OHY0 | 58,511 | https://github.com/openclaw/openclaw/issues/58511 | https://api.github.com/repos/openclaw/openclaw/issues/58511 | Cron sessions inherit persisted model override instead of honoring payload.model, causing thundering herd under overload | ## Summary
Cron jobs that specify `model: openai/gpt-5.4-nano` in their payload are not using that model when they execute in isolated sessions. Instead, `resolvePersistedLiveSelection()` forces the session back to `anthropic/claude-sonnet-4-6`, overriding the cron's declared model intent. Under Anthropic overload con... | closed | completed | true | 0 | [] | [] | 2026-03-31T17:57:10Z | 2026-04-30T04:51:51Z | 2026-04-25T05:15:25Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | jarvis-playrockets | 261,354,837 | U_kgDOD5P1VQ | User | false |
openclaw/openclaw | 4,164,668,218 | I_kwDOQb6kR874O8s6 | 57,112 | https://github.com/openclaw/openclaw/issues/57112 | https://api.github.com/repos/openclaw/openclaw/issues/57112 | [Bug]: Cron payload model override ignored - LiveSessionModelSwitchError on 2026.3.28 | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Cron jobs with payload.model set to a non-default model (e.g. Sonnet) are rejected with LiveSessionModelSwitchError before the LLM is called. The new per-agent model resolution logic in 2026.3.28 compares the cron's payload ... | closed | completed | true | 5 | [
"bug",
"regression"
] | [] | 2026-03-29T15:03:06Z | 2026-04-30T04:51:52Z | 2026-04-25T05:15:25Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | stemkat100 | 264,271,658 | U_kgDOD8B3Kg | User | false |
openclaw/openclaw | 4,181,263,119 | I_kwDOQb6kR875OQMP | 58,517 | https://github.com/openclaw/openclaw/issues/58517 | https://api.github.com/repos/openclaw/openclaw/issues/58517 | Heartbeat model override ignored — always falls back to default model (LiveSessionModelSwitch) | **Version:** OpenClaw 2026.3.28 (f9b1079)
**OS:** macOS 15.6.1 (arm64), Node 25.8.1
## Problem
`agents.defaults.heartbeat.model` is ignored. Heartbeat always runs on the global default model regardless of the heartbeat-specific model setting.
## Config
```json
{
"agents": {
"defaults": {
"model": { "pri... | closed | completed | true | 4 | [] | [] | 2026-03-31T18:02:41Z | 2026-04-30T04:51:48Z | 2026-04-25T05:15:26Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | davidbennett1979 | 82,350,039 | MDQ6VXNlcjgyMzUwMDM5 | User | false |
openclaw/openclaw | 4,180,561,970 | I_kwDOQb6kR875LlAy | 58,459 | https://github.com/openclaw/openclaw/issues/58459 | https://api.github.com/repos/openclaw/openclaw/issues/58459 | [Bug]: LiveSessionModelSwitchError in cron jobs with payload.model override on 2026.3.28 | ## Bug Summary
Cron jobs with `payload.model` set to a non-default model are rejected with `LiveSessionModelSwitchError` before the LLM is called. This is a regression in 2026.3.28.
## Bug Type
Regression (worked before, now fails)
## Summary
Cron jobs with `payload.model` set to a non-default model (e.g. minimax/... | closed | completed | true | 0 | [] | [] | 2026-03-31T16:21:45Z | 2026-04-30T04:51:49Z | 2026-04-25T05:15:26Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | ms7-git | 264,172,505 | U_kgDOD77z2Q | User | false |
openclaw/openclaw | 4,167,577,934 | I_kwDOQb6kR874aDFO | 57,501 | https://github.com/openclaw/openclaw/issues/57501 | https://api.github.com/repos/openclaw/openclaw/issues/57501 | isolated session 模型切换死循环 | # OpenClaw Issue: isolated session 模型切换死循环
## 问题描述
创建 isolated cron 任务时，无论使用什么模型都会触发 `LiveSessionModelSwitchError`：
- **百炼模型**（qwen3.5-plus/kimi-k2.5）看到 Runtime 中的 `model=anthropic/claude-opus-4-6` → 尝试切换到 Opus → 被拦截 → fallback → 也失败
- **Opus 模型**（claude-opus-4-6）看到 Runtime 中的 `model=anthropic/claude-opus-4-6` → 尝试切... | closed | completed | true | 1 | [] | [] | 2026-03-30T05:38:21Z | 2026-04-30T04:51:50Z | 2026-04-25T05:15:26Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | kabudaren | 161,951,903 | U_kgDOCacwnw | User | false |
openclaw/openclaw | 4,230,451,004 | I_kwDOQb6kR878J488 | 63,617 | https://github.com/openclaw/openclaw/issues/63617 | https://api.github.com/repos/openclaw/openclaw/issues/63617 | [Bug]:Kimi cron contamination caused malformed tool calls and unintended browser opens | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
Yes
### Summary
A recurring isolated cron run on Kimi received the correct prompt, then diverged into unrelated Chinese task context and emitted malformed tool calls. OpenClaw subsequently attempted unintended browser actions... | closed | completed | true | 0 | [
"bug",
"bug:behavior"
] | [] | 2026-04-09T07:46:33Z | 2026-04-30T04:51:46Z | 2026-04-25T05:15:27Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Anmeowrchy | 201,754,439 | U_kgDODAaHRw | User | false |
openclaw/openclaw | 4,218,912,493 | I_kwDOQb6kR877d37t | 62,580 | https://github.com/openclaw/openclaw/issues/62580 | https://api.github.com/repos/openclaw/openclaw/issues/62580 | [Bug]: **Bug Report: Response Contamination / Cross-Session Leakage** | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Asked Openclaw about gogcli installation. Response started in English, then switched mid-message to coherent Chinese text with
<img width="725" height="538" alt="Image" src="https://github.com/user-attachment... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-04-07T15:51:44Z | 2026-04-30T04:51:47Z | 2026-04-25T05:15:27Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | The-Lost-Remote | 266,413,231 | U_kgDOD-Ekrw | User | false |
openclaw/openclaw | 4,041,510,018 | I_kwDOQb6kR87w5IyC | 39,998 | https://github.com/openclaw/openclaw/issues/39998 | https://api.github.com/repos/openclaw/openclaw/issues/39998 | Bug: 上下文超限导致对话彻底死亡，且 token 统计与实际不一致 | ## Bug 描述
在 TUI 模式下，当上下文接近模型限制时，对话会彻底死亡，无法恢复。同时，状态栏显示的 token 使用比例与实际限制不一致。
## 现象
- 状态栏显示：`tokens 137k/205k (67%)`
- 但模型返回 `model_context_window_exceeded` 错误
- 之后任何消息都无法回复，对话彻底死亡
- 只能手动移走会话文件才能恢复正常对话
## 会话分析
- 对话文件大小：783KB
- 304 条消息，131 次工具调用，132 个 thinking 块
- 估算 tokens：97k+（不含 thinking），加 thinking 可能 120k-150k
- ... | closed | completed | true | 5 | [
"stale"
] | [] | 2026-03-08T15:27:37Z | 2026-04-30T04:51:46Z | 2026-04-25T05:15:50Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | chinabaijunjie | 33,617,869 | MDQ6VXNlcjMzNjE3ODY5 | User | false |
openclaw/openclaw | 4,326,837,043 | I_kwDOQb6kR88AAAABAeZLMw | 71,371 | https://github.com/openclaw/openclaw/issues/71371 | https://api.github.com/repos/openclaw/openclaw/issues/71371 | Control UI: messages flash-disappear on turn completion (race in session.message handler) | ## Summary
After an assistant turn completes, messages visually disappear from the Control UI chat view. Browser refresh brings them back immediately. The bug is a race condition in the `session.message` WebSocket event handler.
## Environment
- OpenClaw v2026.4.23, gateway mode: local
- WSL2 (Linux 6.6.87.2-microso... | closed | completed | true | 2 | [] | [] | 2026-04-25T03:02:55Z | 2026-04-28T04:45:04Z | 2026-04-25T05:16:13Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | WolvenRA | 35,547,932 | MDQ6VXNlcjM1NTQ3OTMy | User | false |
openclaw/openclaw | 4,216,010,251 | I_kwDOQb6kR877SzYL | 62,344 | https://github.com/openclaw/openclaw/issues/62344 | https://api.github.com/repos/openclaw/openclaw/issues/62344 | Model state bleeds from cron sessions to parent session | ## Bug: Model Override in Cron Jobs Corrupts Parent Session State
### Summary
When a cron job is configured with a model override (e.g., `model: "ollama/kimi-k2.5:cloud"`), the model state bleeds into the parent session's persistent state in `sessions.json`, causing the main session to unexpectedly switch models.
###... | closed | completed | true | 2 | [] | [] | 2026-04-07T06:56:30Z | 2026-04-30T04:51:45Z | 2026-04-25T05:17:08Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | mikronn2 | 105,823,039 | U_kgDOBk67Pw | User | false |
openclaw/openclaw | 4,182,785,725 | I_kwDOQb6kR875UD69 | 58,592 | https://github.com/openclaw/openclaw/issues/58592 | https://api.github.com/repos/openclaw/openclaw/issues/58592 | Cron isolated sessions fail with LiveSessionModelSwitchError when agent primary model differs from payload.model | ## Bug Report
### Summary
Cron jobs with `sessionTarget: "isolated"` and `payload.kind: "agentTurn"` fail immediately with `LiveSessionModelSwitchError` when the `payload.model` field specifies a different model than the agent's configured `model.primary`. This makes the `model` field in cron payloads effectively non-... | closed | completed | true | 1 | [] | [] | 2026-03-31T22:34:21Z | 2026-04-30T04:51:43Z | 2026-04-25T05:17:16Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | OneStepAt4time | 106,186,915 | U_kgDOBlRIow | User | false |
openclaw/openclaw | 4,181,310,794 | I_kwDOQb6kR875Ob1K | 58,518 | https://github.com/openclaw/openclaw/issues/58518 | https://api.github.com/repos/openclaw/openclaw/issues/58518 | Bug: LiveSessionModelSwitchError triggers for isolated sessions on heartbeat/cron when defaults differ | ## Description
When creating an isolated session via embedded workers (like cron or heartbeat), the session initially inherits the default primary model from `agents.defaults.model.primary`. However, when the heartbeat or cron payload explicitly requests a different model (e.g., `ollama/qwen3:8b`), the system triggers ... | closed | completed | true | 1 | [] | [] | 2026-03-31T18:11:11Z | 2026-04-30T04:51:42Z | 2026-04-25T05:17:19Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | banna-commits | 263,209,674 | U_kgDOD7BCyg | User | false |
openclaw/openclaw | 4,178,011,808 | I_kwDOQb6kR875B2ag | 58,307 | https://github.com/openclaw/openclaw/issues/58307 | https://api.github.com/repos/openclaw/openclaw/issues/58307 | Cron payload model field ignored — LiveSessionModelSwitchError overrides job-level model with agent default | ## Bug Report
### Summary
When a cron job specifies a `model` in its payload (e.g., `moonshot/kimi-k2.5`), the model is used to initialize the isolated session but is **not persisted** as a session store override (`providerOverride`/`modelOverride`). Mid-run, `resolveLiveSessionModelSelection()` re-resolves the model ... | closed | completed | true | 3 | [] | [] | 2026-03-31T10:34:02Z | 2026-04-30T04:51:41Z | 2026-04-25T05:17:21Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | clawdinghenry-blip | 259,933,712 | U_kgDOD35GEA | User | false |
openclaw/openclaw | 4,327,126,833 | I_kwDOQb6kR88AAAABAeq3MQ | 71,406 | https://github.com/openclaw/openclaw/issues/71406 | https://api.github.com/repos/openclaw/openclaw/issues/71406 | Discord plugin posts each agent message twice in v2026.4.22 (regression from v2026.4.12) | ## Summary
After updating from `2026.4.12` to `2026.4.22`, every message the main agent posts to a Discord thread appears **twice** in the channel, with timestamps 4–20 seconds apart. The gateway log shows only one send event, but Discord's API returns two separate messages with near-identical content from the same bo... | closed | completed | true | 1 | [] | [] | 2026-04-25T05:01:01Z | 2026-04-30T04:51:40Z | 2026-04-25T05:22:53Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | alexgross21 | 265,779,045 | U_kgDOD9d3ZQ | User | false |
openclaw/openclaw | 4,326,892,345 | I_kwDOQb6kR88AAAABAecjOQ | 71,381 | https://github.com/openclaw/openclaw/issues/71381 | https://api.github.com/repos/openclaw/openclaw/issues/71381 | Bug: Heartbeat system events render as user messages in webchat; HEARTBEAT_OK not silently suppressed | ## Description
In the webchat / Control UI, heartbeat system events are rendering as messages in the **user's message bubble** instead of being processed silently by the agent. Additionally, the agent's `HEARTBEAT_OK` response (which per docs should be silently dropped) is also rendering visibly in the chat thread.
#... | closed | completed | true | 4 | [] | [] | 2026-04-25T03:25:05Z | 2026-04-30T04:51:40Z | 2026-04-25T05:16:42Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | gerald1950ggg-ai | 271,614,232 | U_kgDOEDCBGA | User | false |
openclaw/openclaw | 4,111,685,908 | I_kwDOQb6kR871E1kU | 51,473 | https://github.com/openclaw/openclaw/issues/51473 | https://api.github.com/repos/openclaw/openclaw/issues/51473 | [Bug]: Compatibility issues with vLLM | ### Bug type
Regression (worked before, now fails)
### Summary
When using OpenClaw, the context utilization rate remains at 0 and fails to perform proper compaction when the backend is connected to vLLM.
### Steps to reproduce
当openclaw 配置使用 vllm 大模型的会出现上下文使用率一直为0 的情况，并且无法做到自动压缩
### Expected behavior
master 分支... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-21T07:04:56Z | 2026-04-30T04:51:39Z | 2026-04-25T05:27:11Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | lskdegithub | 71,971,422 | MDQ6VXNlcjcxOTcxNDIy | User | false |
openclaw/openclaw | 4,111,691,091 | I_kwDOQb6kR871E21T | 51,474 | https://github.com/openclaw/openclaw/issues/51474 | https://api.github.com/repos/openclaw/openclaw/issues/51474 | Control UI on local loopback (127.0.0.1) still requires pairing in 2026.3.13, despite docs saying local connections auto-approve | ### Bug type
Regression (worked before, now fails)
### Summary
... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-21T07:07:33Z | 2026-04-30T04:51:38Z | 2026-04-25T05:27:18Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | tothdbexec | 269,885,192 | U_kgDOEBYfCA | User | false |
openclaw/openclaw | 4,111,706,959 | I_kwDOQb6kR871E6tP | 51,478 | https://github.com/openclaw/openclaw/issues/51478 | https://api.github.com/repos/openclaw/openclaw/issues/51478 | [Bug] Sandbox mode: DingTalk sendMedia fails for workspace files | # Bug: Sandbox mode - DingTalk sendMedia fails for workspace files
## Problem
When using Docker sandbox with `workspaceAccess: "rw"`, DingTalk's `sendMedia` fails to send images/files from the sandbox workspace path (`/workspace`).
The issue is that `loadWebMedia` runs in the Gateway process on the host, receives th... | closed | completed | true | 2 | [] | [] | 2026-03-21T07:18:24Z | 2026-04-30T04:51:37Z | 2026-04-25T05:27:25Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | JasonRD | 16,793,420 | MDQ6VXNlcjE2NzkzNDIw | User | false |
openclaw/openclaw | 4,111,754,582 | I_kwDOQb6kR871FGVW | 51,494 | https://github.com/openclaw/openclaw/issues/51494 | https://api.github.com/repos/openclaw/openclaw/issues/51494 | [Bug]: api.runtime.version returns 'unknown' for plugins in npm-installed OpenClaw | ### Bug type
Behavior bug
### Summary
`api.runtime.version` (exposed to plugins via `createPluginRuntime()`) always returns `"unknown"` when OpenClaw is installed via npm. The `resolveVersion()` function in `src/plugins/runtime/index.ts` uses a hardcoded relative path `"../../../package.json"` that only works in the... | closed | completed | true | 4 | [] | [] | 2026-03-21T07:51:22Z | 2026-04-30T04:51:34Z | 2026-04-25T05:27:54Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | xiaohuaxi | 45,277,353 | MDQ6VXNlcjQ1Mjc3MzUz | User | false |
openclaw/openclaw | 4,111,766,997 | I_kwDOQb6kR871FJXV | 51,498 | https://github.com/openclaw/openclaw/issues/51498 | https://api.github.com/repos/openclaw/openclaw/issues/51498 | [Bug]: openclaw cron list/status and openclaw health --json timeout against local gateway while scheduler still appears to run jobs | ### Bug type
Regression (worked before, now fails)
### Summary
openclaw cron list/status and openclaw health --json timeout against local gateway while scheduler still appears to run jobs
### Steps to reproduce
1. Run `openclaw gateway status` on a Linux host with a user-level systemd gateway.
2. Confirm the gatew... | closed | completed | true | 3 | [
"bug",
"regression"
] | [] | 2026-03-21T07:59:40Z | 2026-04-30T04:51:33Z | 2026-04-25T05:28:01Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Parametric89 | 140,854,934 | U_kgDOCGVGlg | User | false |
openclaw/openclaw | 4,049,596,680 | I_kwDOQb6kR87xX_EI | 41,815 | https://github.com/openclaw/openclaw/issues/41815 | https://api.github.com/repos/openclaw/openclaw/issues/41815 | macOS LaunchAgent restart should use kickstart first and detach when invoked from the managed gateway process tree | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
On macOS, `openclaw gateway restart` still uses a synchronous `launchctl` restart path that does not survive when the command is invoked from inside the running gateway's own LaunchAgent process tree.
This shows up as two symptoms that hav... | closed | not_planned | true | 8 | [
"close:duplicate",
"dedupe:child"
] | [] | 2026-03-10T05:56:14Z | 2026-04-30T04:51:31Z | 2026-04-25T05:28:12Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | constansino | 65,108,260 | MDQ6VXNlcjY1MTA4MjYw | User | false |
openclaw/openclaw | 4,111,802,765 | I_kwDOQb6kR871FSGN | 51,506 | https://github.com/openclaw/openclaw/issues/51506 | https://api.github.com/repos/openclaw/openclaw/issues/51506 | [Bug] Webchat model selector prefixes all models with "anthropic/" regardless of actual provider | ## Description
The webchat control UI model selector incorrectly prefixes all model IDs with `anthropic/`, ignoring the actual provider name. This causes "model not allowed" errors when switching to non-Anthropic models.
## Steps to Reproduce
1. Configure multiple providers in `openclaw.json` (e.g., `anthropic`, `zhip... | closed | completed | true | 2 | [] | [] | 2026-03-21T08:23:57Z | 2026-04-30T04:51:30Z | 2026-04-25T05:28:15Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | 765007025 | 35,289,203 | MDQ6VXNlcjM1Mjg5MjAz | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.