fix(ai): assistant text blocks are run together on replay by Yigtwxx · Pull Request #115743 · openclaw/openclaw

Yigtwxx · 2026-07-29T07:44:22Z

What Problem This Solves

Fixes an issue where users on any OpenAI-compatible provider would have the assistant's own prior turn replayed back to the model with sentences fused together, losing the boundary between them.

An assistant turn holding two text blocks, "Let me check the file." and "The file contains X.", was replayed as:

Let me check the file.The file contains X.

Nothing errors and nothing is logged. The corrupted text is what the model reads as its own previous turn on every subsequent request in the conversation, so the damage compounds as the transcript grows.

This affects the Chat Completions lane generally: OpenRouter, Groq, DeepSeek, Together, LM Studio, Ollama and other OpenAI-compatible endpoints.

Why This Change Was Made

Two text blocks in one assistant turn is a routine shape, not a corner case:

Streaming opens a new text block whenever the current block is not text, so a text → tool call → text turn ends with two separate text blocks.
Cross-model replay converts a thinking block into a text block, which is then adjacent to the real answer block.

convertMessages flattened them with join(""). Every neighbouring path disagrees with that choice:

the thinking blocks a few lines below in the same function join with "\n\n", and the thinking signature joins with "\n"
flattenCompletionMessagesToStringContent, the helper that performs this very operation for strict OpenAI-compatible servers, joins with "\n"
the Anthropic, Responses and Mistral lanes never flatten at all; they emit one content block per block

This change joins with "\n", matching the string-content flattener, which is the closest sibling: same input shape, same output shape, same purpose. "\n\n" was the alternative, but it is used for thinking blocks rather than for flattening text parts, so the narrower precedent seemed the safer one to follow. Happy to switch if maintainers prefer the paragraph break.

Nothing else changes: block filtering, surrogate sanitization, the thinking paths, and the content-part array shape are all untouched.

User Impact

Assistant turns that contain more than one text block are replayed with their boundaries intact, so the model sees what it actually said instead of two sentences fused mid-word. Turns with a single text block, which is the common case, serialize byte-for-byte as before.

Evidence

Real runtime proof: what an OpenAI-compatible server actually receives

Requested by review — this goes past Vitest output and captures the bytes on the wire.

A local node:http server stands in for the provider endpoint: model.baseUrl points at http://127.0.0.1:<port>/v1, the server parses the real request body, records messages[], and answers with a real SSE stream that the transport parses normally. No fetch stub, no mocked client — a real socket, the real streamOpenAICompletions path, the real serializer. It needs no provider account.

The conversation contains one assistant turn with two text blocks. Only openai-completions-messages.ts differs between the two runs.

source	`assistant.content` received by the server
current `main`	`"Let me check the file.The file contains X."`
this PR	`"Let me check the file.\nThe file contains X."`

Terminal output, unfixed source first:

--- source reverted to main ---
=== what the OpenAI-compatible server received on the wire ===
assistant content: "Let me check the file.The file contains X."
contains newline between blocks: false
fused (bug): true
RESULT: fused

and with the patch applied:

=== what the OpenAI-compatible server received on the wire ===
assistant content: "Let me check the file.\nThe file contains X."
contains newline between blocks: true
fused (bug): false
RESULT: separated

The script exits on a boolean computed from the captured payload rather than printing a fixed "PASS", so a regression cannot pass silently.

Harness (run with tsx, not committed)

const server = createServer((req, res) => {
  const chunks: Buffer[] = [];
  req.on("data", (chunk: Buffer) => chunks.push(chunk));
  req.on("end", () => {
    const payload = JSON.parse(Buffer.concat(chunks).toString("utf8"));
    for (const message of payload.messages) {
      captured.push({ content: message.content, role: message.role });
    }
    res.writeHead(200, { "content-type": "text/event-stream" });
    // ... real SSE chunks + [DONE]
  });
});
await new Promise<void>((resolve) => server.listen(0, "127.0.0.1", resolve));

const model = { /* ...,  baseUrl: `http://127.0.0.1:${port}/v1` */ };
const assistant = {
  role: "assistant",
  content: [
    { type: "text", text: "Let me check the file." },
    { type: "text", text: "The file contains X." },
  ],
  // ...
};

await streamOpenAICompletions(model, context, { apiKey: "proof-key" }).result();

const replayed = captured.find((m) => m.role === "assistant");
const pass = replayed.content.includes("file.\nThe file");
process.exit(pass ? 0 : 1);

Regression test

Added packages/ai/src/openai-completions-messages.test.ts, which drives convertMessages with a two-text-block assistant turn and asserts the replayed content. It is load-bearing — against the unfixed source it reproduces the exact corruption rather than an incidental error:

FAIL packages/ai/src/openai-completions-messages.test.ts
  > convertMessages assistant text replay > keeps separate assistant text blocks apart
- Expected
+ Received
- Let me check the file.
- The file contains X.
+ Let me check the file.The file contains X.

Test Files  1 failed (1)
     Tests  1 failed (1)

The whole packages/ai suite is green, including the compatibility suite that pins Chat Completions payload shapes across provider variants:

node scripts/run-vitest.mjs run packages/ai/src/

Test Files  51 passed (51)
     Tests  906 passed (906)

oxfmt --check is clean on both touched files. The branch is rebased on current main.

AI-assisted: written and verified with an AI coding agent; the runtime proof above, the failing-test-first measurement, and the regression run were reviewed by me.

clawsweeper · 2026-07-29T13:01:17Z

Codex review: needs maintainer review before merge. Reviewed August 1, 2026, 3:57 AM ET / 07:57 UTC.

ClawSweeper review

What this changes

The PR preserves boundaries between multiple assistant text blocks during OpenAI-compatible Chat Completions transcript replay by joining them with a newline and testing the resulting request content.

Merge readiness

⚠️ Ready for maintainer review - 2 items remain

This PR is still necessary: current main concatenates separate assistant text blocks with no separator before sending an OpenAI Chat Completions replay. The branch changes that one conversion boundary to a newline, aligns it with the strict string-content compatibility helper, adds a focused regression test, and has no supported correctness or security finding.

Priority: P2
Reviewed head: 725ed96ebe5a8beda5b03b37ece3c93f9ec222a8

Review scores

Measure	Result	What it means
Overall readiness	🐚 platinum hermit (4/6)	A focused, source-proven repair with strong direct runtime evidence and no actionable correctness finding.
Proof confidence	🦞 diamond lobster (5/6)	Sufficient (terminal): The PR body provides an after-fix loopback HTTP/SSE capture of the real Chat Completions serializer receiving newline-separated assistant content, supplemented by focused regression coverage and successful supplied exact-head checks.
Patch quality	🐚 platinum hermit (4/6)	No actionable review findings were identified.

Verification

Check	Result	Evidence
Real behavior	Verified	Sufficient (terminal): The PR body provides an after-fix loopback HTTP/SSE capture of the real Chat Completions serializer receiving newline-separated assistant content, supplemented by focused regression coverage and successful supplied exact-head checks.
Evidence reviewed	7 items	Current-main defect: Current main filters assistant text content blocks and joins their sanitized text with an empty string, so two non-empty blocks replay without a boundary. Sibling serialization contract: The strict OpenAI-compatible string-content helper flattens an array of text parts with a newline, providing a direct same-protocol precedent for the proposed delimiter. Production entry point: The OpenAI Chat Completions provider calls this converter while building the streaming request parameters, making the changed value part of the provider request transcript.
Findings	None	None.
Security	None	None.

How this fits together

The AI package converts stored conversation messages into OpenAI Chat Completions request messages. That serialized transcript becomes the prior context sent to OpenAI-compatible providers, so a flattened assistant turn affects later model responses.

flowchart LR
  A[Conversation history] --> B[Assistant content blocks]
  B --> C[Chat Completions converter]
  C --> D[Serialized request transcript]
  D --> E[OpenAI-compatible provider]
  E --> F[Next model response]

Before merge

Resolve merge risk (P1) - The supplied GitHub state marks the cleanly mergeable head as behind current main. Refresh the GitHub merge result and required checks against the exact merge head if main advances again before landing.
Complete next step (P2) - No repair lane is needed because this PR already contains the narrow source fix and regression coverage; the remaining action is normal maintainer landing after confirming the current merge result.

Agent review details

Security

None.

PR surface

Source +2, Tests +52. Total +54 across 2 files.

View PR surface stats

Area	Files	Added	Removed	Net
Source	1	3	1	+2
Tests	1	52	0	+52
Docs	0	0	0	0
Config	0	0	0	0
Generated	0	0	0	0
Other	0	0	0	0
Total	2	55	1	+54

Review metrics

None.

Merge-risk options

Maintainer options:

Decide the mitigation before merge
Land the focused newline repair and its regression test once the current GitHub merge result remains clean; retain the existing special handling for thinking content and tool calls.
Pause or close
Do not merge this PR until maintainers decide whether the risk is worth taking.

Technical review

Best possible solution:

Land the focused newline repair and its regression test once the current GitHub merge result remains clean; retain the existing special handling for thinking content and tool calls.

Do we have a high-confidence way to reproduce the issue?

Yes—current main directly joins two non-empty assistant text blocks with "" at packages/ai/src/openai-completions-messages.ts:144; the PR also supplies a loopback request-capture proof for that exact conversion path.

Is this the best way to solve the issue?

Yes—the newline is the narrowest repair because it changes only string flattening for adjacent assistant text blocks and matches packages/ai/src/transports/openai-completions-string-content.ts:21 without altering thinking, tool-call, or content-array behavior.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against 8317e23c1c2b.

Labels

Label justifications:

P2: The defect can corrupt model-visible replay text for OpenAI-compatible conversations, but the repair and blast radius are confined to one transcript conversion path.
rating: 🐚 platinum hermit: Overall readiness is 🐚 platinum hermit; proof is 🦞 diamond lobster and patch quality is 🐚 platinum hermit.
status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Sufficient (terminal): The PR body provides an after-fix loopback HTTP/SSE capture of the real Chat Completions serializer receiving newline-separated assistant content, supplemented by focused regression coverage and successful supplied exact-head checks.
proof: sufficient: Contributor real behavior proof is sufficient. The PR body provides an after-fix loopback HTTP/SSE capture of the real Chat Completions serializer receiving newline-separated assistant content, supplemented by focused regression coverage and successful supplied exact-head checks.

Evidence

What I checked:

Current-main defect: Current main filters assistant text content blocks and joins their sanitized text with an empty string, so two non-empty blocks replay without a boundary. (packages/ai/src/openai-completions-messages.ts:144, 8317e23c1c2b)
Sibling serialization contract: The strict OpenAI-compatible string-content helper flattens an array of text parts with a newline, providing a direct same-protocol precedent for the proposed delimiter. (packages/ai/src/transports/openai-completions-string-content.ts:21, 8317e23c1c2b)
Production entry point: The OpenAI Chat Completions provider calls this converter while building the streaming request parameters, making the changed value part of the provider request transcript. (packages/ai/src/providers/openai-completions.ts:728, 8317e23c1c2b)
Focused branch repair: The proposed head changes only the empty join to a newline join in the converter and adds a regression test that asserts two assistant text blocks remain separated. (packages/ai/src/openai-completions-messages.ts:144, 725ed96ebe5a)
Behavior provenance: Blame and pickaxe history tie the current empty-string join to the current converter refactor, whose latest history owner is Peter Steinberger; this is provenance for routing, not fault attribution. (packages/ai/src/openai-completions-messages.ts:144, 89394978e0a6)
Base-drift scope: The current main branch has no changes to either proposed path after the PR base, while the supplied GitHub state reports the PR cleanly mergeable but behind main. (packages/ai/src/openai-completions-messages.ts:144, 8317e23c1c2b)

Likely related people:

steipete: Blame attributes both the empty-string assistant replay line and the sibling newline flattener to the same current-main refactor commit by Peter Steinberger. (role: current converter and compatibility-helper history owner; confidence: high; commits: 89394978e0a6; files: packages/ai/src/openai-completions-messages.ts, packages/ai/src/transports/openai-completions-string-content.ts)

Rank-up moves

Optional improvements that raise the rating; they are not merge blockers.

Before landing, refresh the GitHub merge result if current main advances so required checks remain tied to the exact merge head.

Rating scale

Score	Internal tier	Crab rank	Meaning
6/6	S	🦀 challenger crab	Exceptional readiness
5/6	A	🦞 diamond lobster	Very strong readiness
4/6	B	🐚 platinum hermit	Good normal PR; ordinary maintainer review
3/6	C	🦐 gold shrimp	Useful, but confidence is limited
2/6	D	🦪 silver shellfish	Proof or implementation needs work
1/6	F	🧂 unranked krab	Not merge-ready
N/A	NA	🌊 off-meta tidepool	Rating does not apply

Overall follows the weaker of proof and patch quality.
Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

Workflow

ClawSweeper keeps one durable marker-backed review comment per issue or PR.
Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
Maintainers can also comment @clawsweeper review to request a fresh review only.
Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.

History

Review history (7 earlier review cycles)

reviewed 2026-07-29T12:59:43.188Z sha f1bff00 :: needs real behavior proof before merge. :: none
reviewed 2026-07-29T13:47:08.575Z sha f1bff00 :: needs maintainer review before merge. :: none
reviewed 2026-07-29T15:49:04.756Z sha b97f49b :: needs maintainer review before merge. :: none
reviewed 2026-07-29T16:51:53.357Z sha 81bc4c5 :: needs maintainer review before merge. :: none
reviewed 2026-07-31T23:48:30.915Z sha 81bc4c5 :: needs maintainer review before merge. :: none
reviewed 2026-08-01T00:08:19.566Z sha 81bc4c5 :: needs maintainer review before merge. :: none
reviewed 2026-08-01T07:38:40.472Z sha 725ed96 :: needs maintainer review before merge. :: none

Chat Completions replay flattened every assistant text block with an empty separator, so two distinct blocks came back as one word-joined sentence. The same message shape survives distinctly on the Anthropic, Responses and Mistral lanes, and the string-content flattener for strict OpenAI-compatible servers already joins with a newline. Two blocks arise routinely: streaming opens a new text block after a tool call, and cross-model replay converts a thinking block into a text block.

…block-separator

Yigtwxx · 2026-08-01T07:48:16Z

Refreshed onto current main (7aacd1314d7), which clears the P1 "branch is behind" item from the last review. Merged rather than rebased, per this repo's convention: no conflicts, and neither file this PR touches has been modified on main since the branch's base, so the diff itself is unchanged.

Exact-head CI on 725ed96ebe5: 82 passed / 0 failed / 0 pending (25 skipped). The two cancelled entries are superseded runs rather than failures — both were re-dispatched on the same head seconds later and completed green:

check	started 07:39:17	started 07:39:26–33
`Real behavior proof`	cancelled	success
`auto-response`	cancelled	success

Locally on the merged tree, the changed module's full test surface passes: openai-completions-messages.test.ts plus openai-completions-structured-content-parity.test.ts, the only other suite in the repo that consumes it — 55 passed (55).

Yigtwxx · 2026-08-01T07:53:47Z

The review above was taken while the exact-head checks were still running — they have since completed green (82 passed / 0 failed / 0 pending, detail in the comment above), so the "checks still in progress" half of the remaining item is resolved on this head.

@clawsweeper re-review

clawsweeper · 2026-08-01T07:53:50Z

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: when the review finishes, ClawSweeper will create the durable review comment if needed or update the existing comment in place.

Re-review progress:

State: Complete
Detail: The durable review result and its route handoff completed.
Run: https://github.com/openclaw/clawsweeper/actions/runs/30690816510
Updated: 2026-08-01T07:59:08.429Z

steipete · 2026-08-01T09:21:28Z

Land-ready maintainer verification for the unchanged original contributor head 725ed96ebe5a8beda5b03b37ece3c93f9ec222a8:

Independently reproduced current main with the real installed OpenAI SDK v6.49.0, an actual localhost HTTP server, and real streamed SSE responses on both production entry points: streamOpenAICompletions and createOpenAICompletionsTransportStreamFn. Both currently send fused assistant transcript text: "Let me check the file.The file contains X." and "Before tool.After tool.".
Repeated the identical owner-path proof on this exact contributor head: 6/6 actual HTTP requests passed across the two transports. Adjacent text becomes "Let me check the file.\nThe file contains X."; text surrounding a real serialized tool call becomes "Before tool.\nAfter tool." while tool-call ID call_owner_42 remains intact; a single assistant block remains byte-identical on both paths. Both SDK clients consumed real SSE completion streams successfully.
Focused regression and sibling-path proof: node scripts/run-vitest.mjs packages/ai/src/openai-completions-messages.test.ts packages/ai/src/providers/openai-completions-structured-content-parity.test.ts — 55/55 tests passing.
Personally inspected the actual installed dependency contract: openai/resources/chat/completions/completions.d.ts accepts assistant string content and tool calls; completions.js posts that exact request body to /chat/completions.
Fresh structured review: .agents/skills/autoreview/scripts/autoreview --mode branch --base origin/main --engine codex --thinking xhigh — clean, no accepted/actionable findings.
ClawSweeper's optional rank-up move was applied: refreshed GitHub merge computation now reports MERGEABLE; an independent git merge-tree --write-tree origin/main 725ed96ebe5a8beda5b03b37ece3c93f9ec222a8 is clean, and neither touched file changed on main after the PR base.
Exact-head full hosted CI is green and fresh: https://github.com/openclaw/openclaw/actions/runs/30689974207

Best fix is the existing shared conversion owner: its newline matches flattenCompletionMessagesToStringContent, repairs both production transports at once, and adds no configuration, dependency, API, or provider-specific workaround. Original contributor commits/head are preserved; no rebase or fixup needed.

steipete · 2026-08-01T09:23:13Z

Merged via squash.

Prepared head SHA: 725ed96ebe5a8beda5b03b37ece3c93f9ec222a8
Landed commit: afe024e8c7990e9ced295f544c3cc5502c7136d4

* origin/main: fix(test/ui): prevent shared history pollution (openclaw#117348) fix(openrouter): apply image request transport policy (openclaw#117336) perf(gateway): skip clean transcript workers (openclaw#117342) test(testing): assert relay smoke readiness fix(testing): activate relay live smoke transport feat(channels): record account lifecycle facts and retire inferred health vocabularies (openclaw#117300) fix(canvas): remove unsupported snapshot delay hint (openclaw#117326) refactor(agents): fold cache TTL pruning into prompt projection (openclaw#117313) fix(canvas): honor paired-node invocation deadlines (openclaw#117316) fix(system-agent): make exit-and-run-a-terminal guidance surface-correct (openclaw#114633) fix(ai): assistant text blocks are run together on replay (openclaw#115743) refactor(agents): canonicalize subagent execution state (openclaw#117267)

…15743) Chat Completions replay flattened every assistant text block with an empty separator, so two distinct blocks came back as one word-joined sentence. The same message shape survives distinctly on the Anthropic, Responses and Mistral lanes, and the string-content flattener for strict OpenAI-compatible servers already joins with a newline. Two blocks arise routinely: streaming opens a new text block after a tool call, and cross-model replay converts a thinking block into a text block.

* origin/main: fix(test/ui): prevent shared history pollution (openclaw#117348) fix(openrouter): apply image request transport policy (openclaw#117336) perf(gateway): skip clean transcript workers (openclaw#117342) test(testing): assert relay smoke readiness fix(testing): activate relay live smoke transport feat(channels): record account lifecycle facts and retire inferred health vocabularies (openclaw#117300) fix(canvas): remove unsupported snapshot delay hint (openclaw#117326) refactor(agents): fold cache TTL pruning into prompt projection (openclaw#117313) fix(canvas): honor paired-node invocation deadlines (openclaw#117316) fix(system-agent): make exit-and-run-a-terminal guidance surface-correct (openclaw#114633) fix(ai): assistant text blocks are run together on replay (openclaw#115743) refactor(agents): canonicalize subagent execution state (openclaw#117267)

openclaw-barnacle Bot added the size: S label Jul 29, 2026

Yigtwxx force-pushed the fix/completions-text-block-separator branch 3 times, most recently from 53dc9a7 to f1bff00 Compare July 29, 2026 12:43

Yigtwxx force-pushed the fix/completions-text-block-separator branch from f1bff00 to b97f49b Compare July 29, 2026 15:44

Yigtwxx force-pushed the fix/completions-text-block-separator branch from b97f49b to 81bc4c5 Compare July 29, 2026 16:44

Yigtwxx mentioned this pull request Jul 29, 2026

fix(web): provider caches split one URL across case-only variants #112390

Closed

clawsweeper Bot added rating: 🦞 diamond lobster Very strong PR readiness with only minor maintainer review expected. and removed rating: 🐚 platinum hermit Good normal PR readiness with ordinary maintainer review expected. labels Aug 1, 2026

Merge remote-tracking branch 'origin/main' into fix/completions-text-…

725ed96

…block-separator

clawsweeper Bot added rating: 🐚 platinum hermit Good normal PR readiness with ordinary maintainer review expected. and removed rating: 🦞 diamond lobster Very strong PR readiness with only minor maintainer review expected. labels Aug 1, 2026

steipete merged commit afe024e into openclaw:main Aug 1, 2026
108 of 110 checks passed

github-actions Bot mentioned this pull request Aug 1, 2026

📡 Upstream Digest — 2026-08-01 09:51 UTC curtismercier/openclaw-mods#1463

Open

clawsweeper Bot mentioned this pull request Aug 1, 2026

[Bug]: Signal context assembly can omit adjacent turns and inject older retrieved context #117510

Closed

Yigtwxx deleted the fix/completions-text-block-separator branch August 2, 2026 07:22

Navigation Menu

Uh oh!

Uh oh!

fix(ai): assistant text blocks are run together on replay - #115743

fix(ai): assistant text blocks are run together on replay#115743
steipete merged 2 commits into
openclaw:mainfrom
Yigtwxx:fix/completions-text-block-separator

Yigtwxx commented Jul 29, 2026 •

edited

Loading

Uh oh!

clawsweeper Bot commented Jul 29, 2026 •

edited

Loading

Security

PR surface

Review metrics

Merge-risk options

Technical review

Labels

Evidence

Rank-up moves

Rating scale

Workflow

History

Uh oh!

Yigtwxx commented Aug 1, 2026

Uh oh!

Yigtwxx commented Aug 1, 2026

Uh oh!

clawsweeper Bot commented Aug 1, 2026 •

edited

Loading

Uh oh!

steipete commented Aug 1, 2026

Uh oh!

Uh oh!

steipete commented Aug 1, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Uh oh!

Conversation

Yigtwxx commented Jul 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What Problem This Solves

Why This Change Was Made

User Impact

Evidence

Real runtime proof: what an OpenAI-compatible server actually receives

Regression test

Uh oh!

clawsweeper Bot commented Jul 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

ClawSweeper review

What this changes

Merge readiness

Review scores

Verification

How this fits together

Before merge

Security

PR surface

Review metrics

Merge-risk options

Technical review

Labels

Evidence

Rank-up moves

Rating scale

Workflow

History

Uh oh!

Yigtwxx commented Aug 1, 2026

Uh oh!

Yigtwxx commented Aug 1, 2026

Uh oh!

clawsweeper Bot commented Aug 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

steipete commented Aug 1, 2026

Uh oh!

Uh oh!

steipete commented Aug 1, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Yigtwxx commented Jul 29, 2026 •

edited

Loading

clawsweeper Bot commented Jul 29, 2026 •

edited

Loading

clawsweeper Bot commented Aug 1, 2026 •

edited

Loading