OpenClaw v2026.4.29-beta.3: Model coverage expands with NVIDIA, operator field notes

The source record for v2026.4.29-beta.3 names a provider, model, voice, or context change. This is an operator's reading of that record, with the exact source facts kept next to a bounded experiment rather than a generic feature list.

Pre-release status matters here: stage the tag beside the stable package and do not infer support for every account or platform. A release tag describes the project change. Your Gateway's provider, channel, platform, and account support should be recorded separately.

A small field test

01 / Messaging and automation get active-run steering by

Source signal. Messaging and automation get active-run steering by default, visible-reply enforcement, spawned subagent routing metadata, and opt-in follow-up commitments for heartbeat-delivered reminders.

Use Messaging and automation get active-run steering by as a routing test: capture the destination before the send, interrupt the transport, and verify that retry logic did not duplicate or redirect the response.

02 / Memory grows into a people-aware wiki with provenance

Source signal. Memory grows into a people-aware wiki with provenance views, per-conversation Active Memory filters, partial recall on timeout, and bounded REM preview diagnostics.

Memory grows into a people-aware wiki with provenance changes what a run can carry forward. Use synthetic names, inspect both recalled and omitted context, and check which workspace or participant can read it.

03 / Provider/model coverage expands with NVIDIA

Source signal. Provider/model coverage expands with NVIDIA onboarding/catalogs plus faster manifest-backed model/auth paths, Bedrock Opus 4.7 thinking parity, and safer Codex/OpenAI-compatible replay and streaming behavior.

Before widening access for Provider/model coverage expands with NVIDIA, replay a narrow allow and deny case. A readable denial and a revoked test credential are part of the acceptance record.

04 / Gateway and packaged-plugin reliability focuses

Source signal. Gateway and packaged-plugin reliability focuses on slow-host startup, reusable model catalogs, event-loop readiness diagnostics, runtime-dependency repair, stale-session recovery, and version-scoped update caches.

A catalog label for Gateway and packaged-plugin reliability focuses is only the first signal. Confirm the account, endpoint, context limit, and tool policy that the running Gateway actually selected.

The details that affect a runbook

Channel fixes cluster around Slack Block Kit limits

Channel fixes cluster around Slack Block Kit limits, Telegram proxy/webhook/polling/send resilience, Discord startup/rate-limit handling, WhatsApp delivery/liveness, and Microsoft Teams/Matrix/Feishu edge cases.

The operator question around Channel fixes cluster around Slack Block Kit limits is where the reply lands after a restart. Record the conversation identifiers before testing and compare them after the channel is healthy again.

Security/tools

configured tool sections (tools.exec, tools.fs) no longer implicitly widen restrictive profiles (messaging, minimal). Users who need those tools under a restricted profile must add explicit alsoAllow entries; a startup warning identifies affected configs. Fixes #47487.

The useful evidence for Security/tools is the refusal path as well as the success path. Use synthetic credentials and record the policy that made each decision.

Agents/commitments

add opt-in inferred follow-up commitments with hidden batched extraction, per-agent/per-channel scoping, heartbeat delivery, CLI management, a simple commitments.enabled/commitments.maxPerDay config, and heartbeat-interval due-time clamping so magical check-ins do not echo immediately.

Use Agents/commitments as a routing test: capture the destination before the send, interrupt the transport, and verify that retry logic did not duplicate or redirect the response.

Operator checklist before rollout

1. Before touching persistent state, capture the version, host, provider, and workspace for Messaging and automation get active-run steering by.
2. Read the linked release record and verify a backup whenever the change can alter configuration or session data.
3. Interrupt one delivery after admission, reconnect the transport, and compare the final reply with the original conversation.
4. Force a short transport interruption and verify that retry or recovery preserves the original thread and sender.
5. Close the test with the evidence attached, and do not widen the rollout while a behavior remains unexplained.

openclaw --version
openclaw gateway status
openclaw security audit

Keep the source trail attached

Close the loop with an explicit next action: keep the candidate isolated, proceed with the staged upgrade, or stop and investigate. Do not treat a green install as proof that the runtime path is correct.

Regression cases from the source

Providers/OpenAI Codex. preserve existing wrapped Codex streams during OpenAI attribution so PI OAuth bearer injection reaches ChatGPT/Codex Responses, and strip native Codex-only unsupported payload fields without touching custom compatible endpoints. Treat the fix as a boundary condition and retain the source wording beside your local result.

Agents/tool-result guard. use the resolved runtime context token budget for non-context-engine tool-result overflow checks, so long tool-heavy sessions no longer compact early when contextTokens is larger than native contextWindow. Fixes #74917. Keep the reproduction and the post-fix result together in the acceptance record.

This page is a practical reading of the v2026.4.29-beta.3 source record; the release index and session concepts remain the references to use for migrations and support boundaries.