fix(sessions): clear stale contextTokens on model switch by yuweuii · Pull Request #38044 · openclaw/openclaw

yuweuii · 2026-03-06T15:35:51Z

Summary

Describe the problem and fix in 2–5 bullets:

Problem: session-scoped contextTokens can stay pinned to an older/smaller model window after a model switch in the same session.
Why it matters: switching back to a larger-context model can still report and budget against the stale smaller window, causing premature compaction and incorrect status output.
What changed: clear cached contextTokens when a model override/default-model selection changes (or when stored runtime model identity is already stale), so status/runtime paths recompute from the newly active model.
What did NOT change (scope boundary): this does not change model selection itself, model discovery, or compaction thresholds; it only invalidates stale cached context-window state on model switch.

Change Type (select all)

Scope (select all touched areas)

Linked Issue/PR

User-visible / Behavior Changes

After switching from a smaller-context model back to a larger-context model in the same session, /status and related session views now reflect the larger window again instead of staying pinned to the old smaller value.
Existing sessions no longer need a /new just to recover the correct context window after a model switch.

Security Impact (required)

New permissions/capabilities? (No)
Secrets/tokens handling changed? (No)
New/changed network calls? (No)
Command/tool execution surface changed? (No)
Data access scope changed? (No)
If any Yes, explain risk + mitigation:

Repro + Verification

Environment

OS: macOS 26.3
Runtime/container: Node.js workspace / local source checkout
Model/provider: local provider models with different configured context windows
Integration/channel (if any): session store + status output
Relevant config (redacted): one small-window local model (4096) and one larger-window local model (65536)

Steps

Start with a session entry that has a small-model override and persisted contextTokens = 4096.
Switch that same session back to a larger model/default selection.
Inspect the resulting session state / status output.

Expected

The session should recompute contextTokens from the currently active model, so the larger window is restored.

Actual

Before this patch, the session kept the stale 4096 override even after the active model switched back to the larger one.

Evidence

Attach at least one:

Failing test/log before + passing after
Trace/log snippets
Screenshot/recording
Perf numbers (if relevant)

Minimal local repro before fix:

before {"resolved":{"provider":"sunapi386","model":"llama-3-lexi-uncensored:8b"},"context":4096,"rawContextTokens":4096}
after {"resolved":{"provider":"local","model":"llama3.1:8b"},"context":4096,"rawContextTokens":4096}

Control check for the larger model window:

Human Verification (required)

What you personally verified (not just CI), and how:

Verified scenarios:
- reproduced the bug locally with a minimal session-entry script (small-window override -> switch back to larger model -> stale contextTokens remained)
- verified that clearing stale contextTokens fixes the reproduced case
- verified status output now re-resolves to the larger window after the switch
Edge cases checked:
- switching between two explicit overrides
- switching back to the default model
- unchanged/already-aligned runtime model does not clear state unnecessarily
What you did not verify:
- full end-to-end interactive TUI reproduction on every provider family
- multi-process races involving concurrent session-store writers

Compatibility / Migration

Backward compatible? (Yes)
Config/env changes? (No)
Migration needed? (No)
If yes, exact upgrade steps:

Failure Recovery (if this breaks)

How to disable/revert this change quickly:
- revert this PR commit
Files/config to restore:
- src/sessions/model-overrides.ts
- src/sessions/model-overrides.test.ts
- src/auto-reply/status.test.ts
Known bad symptoms reviewers should watch for:
- model switch stops invalidating stale session context windows
- /status remains pinned to the previous smaller window after switching back to a larger model

Risks and Mitigations

List only real risks for this PR. Add/remove entries as needed. If none, write None.

Risk:
- Clearing contextTokens on model switch could briefly remove a cached window until the new model is re-resolved.
- Mitigation:
  - all affected read paths already have normal fallbacks (agent.contextTokens, configured model windows, discovered windows, or DEFAULT_CONTEXT_TOKENS), so the state becomes recomputed rather than empty/broken.

greptile-apps · 2026-03-06T15:40:50Z

Greptile Summary

This PR fixes a session-scoped bug where contextTokens would remain pinned to an older/smaller model's context window after a model switch within the same session, causing /status and compaction budgeting to incorrectly use the stale, smaller window even after switching back to a larger-context model.

The fix adds a focused deletion of entry.contextTokens inside applyModelOverrideToSessionEntry whenever the selection changes or the stored runtime model is misaligned, mirroring the existing runtime-field clearing logic. All read paths already have fallbacks (agent.contextTokens, configured windows, or DEFAULT_CONTEXT_TOKENS), so the brief invalidation is safe.

Key changes:

src/sessions/model-overrides.ts: New block deletes contextTokens when selectionUpdated is true or when the stored runtime model is present but stale relative to the active selection.
src/sessions/model-overrides.test.ts: Existing fixtures gain contextTokens values; a new test covers the "switch back to default model clears stale window" scenario.
src/auto-reply/status.test.ts: New integration test verifies the status output recomputes context window size from the agent after the model switch.
The condition for clearing contextTokens is intentionally slightly broader than the runtime-field condition — it fires on any selection change even when no runtime model has been recorded — which is correct and intentional per the PR description.
The "retains aligned" test would benefit from including an explicit contextTokens assertion to guard against future regressions in the "do not clear when already aligned" code path.

Confidence Score: 4/5

Safe to merge. The fix is focused and correct, with all affected read paths having fallbacks.
The implementation is sound and the deletion logic is precise. The two deletion scenarios are well-tested, and the preservation scenario is tested (though without explicit assertion of the contextTokens field itself). Minor: adding an explicit contextTokens assertion to the aligned-case test would strengthen coverage of the "do not clear when already aligned" path, but its absence does not block the PR.
No blocking issues. Consider adding explicit contextTokens assertion to the aligned-case test in src/sessions/model-overrides.test.ts before merging for better coverage of the non-deletion path.

_{Last reviewed commit: 7731104}

greptile-apps · 2026-03-06T15:40:54Z

src/sessions/model-overrides.test.ts

@@ -85,4 +89,30 @@ describe("applyModelOverrideToSessionEntry", () => {
    expect(entry.model).toBe("gpt-5.2");
    expect(entry.updatedAt).toBe(before);
  });


The "retains aligned runtime model fields" test does not include a contextTokens value in the entry, so the code path where contextTokens is NOT cleared (when model and provider are already aligned) is only exercised implicitly. Adding a contextTokens field and asserting it is preserved after the call would explicitly document and guard the "no unnecessary clearing" contract.

Suggested change

it("retains aligned runtime model fields when selection and runtime already match", () => {

const before = Date.now() - 5_000;

const entry: SessionEntry = {

sessionId: "sess-3",

updatedAt: before,

modelProvider: "openai",

model: "gpt-5.2",

providerOverride: "openai",

modelOverride: "gpt-5.2",

contextTokens: 65_536,

};

const result = applyModelOverrideToSessionEntry({

entry,

selection: {

provider: "openai",

model: "gpt-5.2",

},

});

expect(result.updated).toBe(false);

expect(entry.modelProvider).toBe("openai");

expect(entry.model).toBe("gpt-5.2");

expect(entry.contextTokens).toBe(65_536);

expect(entry.updatedAt).toBe(before);

});

Prompt To Fix With AI

This is a comment left during a code review. Path: src/sessions/model-overrides.test.ts Line: 68-91 Comment: The "retains aligned runtime model fields" test does not include a `contextTokens` value in the entry, so the code path where `contextTokens` is NOT cleared (when model and provider are already aligned) is only exercised implicitly. Adding a `contextTokens` field and asserting it is preserved after the call would explicitly document and guard the "no unnecessary clearing" contract. ```suggestion it("retains aligned runtime model fields when selection and runtime already match", () => { const before = Date.now() - 5_000; const entry: SessionEntry = { sessionId: "sess-3", updatedAt: before, modelProvider: "openai", model: "gpt-5.2", providerOverride: "openai", modelOverride: "gpt-5.2", contextTokens: 65_536, }; const result = applyModelOverrideToSessionEntry({ entry, selection: { provider: "openai", model: "gpt-5.2", }, }); expect(result.updated).toBe(false); expect(entry.modelProvider).toBe("openai"); expect(entry.model).toBe("gpt-5.2"); expect(entry.contextTokens).toBe(65_536); expect(entry.updatedAt).toBe(before); }); ``` How can I resolve this? If you propose a fix, please make it concise.

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

yuweuii · 2026-03-06T15:42:36Z

Added one more local regression pass beyond the focused new tests for this patch.

Also verified these related paths after the fix:

src/gateway/session-utils.test.ts
src/agents/context.test.ts
src/auto-reply/reply/model-selection.test.ts
src/sessions/model-overrides.test.ts
src/auto-reply/status.test.ts

Local result:

pnpm exec vitest run src/sessions/model-overrides.test.ts src/auto-reply/status.test.ts src/gateway/session-utils.test.ts src/agents/context.test.ts src/auto-reply/reply/model-selection.test.ts --reporter=dot
65/65 passed
pnpm build:plugin-sdk:dts passed

That re-checks the adjacent session-model/context-window/status paths so clearing stale contextTokens on model switch does not regress related behavior.

jalehman · 2026-03-08T17:59:19Z

Merged via squash.

Prepared head SHA: bac2df4
Merge commit: 6c9b49a

Thanks @yuweuii!

@jalehman

) Merged via squash. Prepared head SHA: bac2df4 Co-authored-by: yuweuii <[email protected]> Co-authored-by: jalehman <[email protected]> Reviewed-by: @jalehman

@jalehman

) Merged via squash. Prepared head SHA: bac2df4 Co-authored-by: yuweuii <[email protected]> Co-authored-by: jalehman <[email protected]> Reviewed-by: @jalehman

yuweuii · 2026-03-09T00:34:27Z

Thanks Josh — really appreciate you helping get this over the line, especially the test cleanup.

@jalehman

) Merged via squash. Prepared head SHA: bac2df4 Co-authored-by: yuweuii <[email protected]> Co-authored-by: jalehman <[email protected]> Reviewed-by: @jalehman

@jalehman

) Merged via squash. Prepared head SHA: bac2df4 Co-authored-by: yuweuii <[email protected]> Co-authored-by: jalehman <[email protected]> Reviewed-by: @jalehman

@jalehman

) Merged via squash. Prepared head SHA: bac2df4 Co-authored-by: yuweuii <[email protected]> Co-authored-by: jalehman <[email protected]> Reviewed-by: @jalehman

@jalehman

) Merged via squash. Prepared head SHA: bac2df4 Co-authored-by: yuweuii <[email protected]> Co-authored-by: jalehman <[email protected]> Reviewed-by: @jalehman

@jalehman

) Merged via squash. Prepared head SHA: bac2df4 Co-authored-by: yuweuii <[email protected]> Co-authored-by: jalehman <[email protected]> Reviewed-by: @jalehman

@jalehman

) Merged via squash. Prepared head SHA: bac2df4 Co-authored-by: yuweuii <[email protected]> Co-authored-by: jalehman <[email protected]> Reviewed-by: @jalehman

) Cherry-picked from openclaw/openclaw@6c9b49a10. Discarded gutted files: CHANGELOG.md, src/process/supervisor/adapters/child.test.ts, src/sessions/model-overrides.ts, src/sessions/model-overrides.test.ts. Co-authored-by: jalehman <[email protected]>

) (#1283) Cherry-picked from openclaw/openclaw@6c9b49a10. Discarded gutted files: CHANGELOG.md, src/process/supervisor/adapters/child.test.ts, src/sessions/model-overrides.ts, src/sessions/model-overrides.test.ts. Co-authored-by: yuweuii <[email protected]> Co-authored-by: jalehman <[email protected]>

…14 rebase The daily sync was failing for 22 days because GITHUB_TOKEN cannot push commits containing workflow file changes. Switch to FORK_PAT (fine-grained PAT with workflow scope) for the checkout/push steps. Also updates FORK-PATCHES.md to reflect: - Base bumped from v2026.2.21 to v2026.3.14 - Patch openclaw#1 (contextTokens) dropped (absorbed upstream as openclaw#38044) - pi-ai dependency patch version mismatch noted (0.54.0 vs 0.58.0) - Patch numbering updated (8 patches -> 7)

@jalehman

) Merged via squash. Prepared head SHA: bac2df4 Co-authored-by: yuweuii <[email protected]> Co-authored-by: jalehman <[email protected]> Reviewed-by: @jalehman

@jalehman

) Merged via squash. Prepared head SHA: bac2df4 Co-authored-by: yuweuii <[email protected]> Co-authored-by: jalehman <[email protected]> Reviewed-by: @jalehman

@altaywtf

* refactor: share Apple talk config parsing * refactor: add canonical talk config payload * refactor: centralize talk silence timeout defaults * test: cover invalid talk config inputs * test: decouple ios talk parsing coverage * fix: resolve live config paths in status and gateway metadata (openclaw#39952) * fix: resolve live config paths in status and gateway metadata * fix: resolve remaining runtime config path references * test: cover gateway config.set config path response * fix(web-search): restore OpenRouter compatibility for Perplexity (openclaw#39937) (openclaw#39937) * Zalo: fix provider lifecycle restarts (openclaw#39892) * Zalo: fix provider lifecycle restarts * Zalo: add typing indicators, smart webhook cleanup, and API type fixes * fix review * add allow list test secrect * Zalo: bound webhook cleanup during shutdown * Zalo: bound typing chat action timeout * Zalo: use plugin-safe abort helper import * fix(plugins): ship Feishu bundled runtime dependency (openclaw#39990) * fix: ship feishu bundled runtime dependency * test: align feishu bundled dependency specs * fix(hooks): use resolveAgentIdFromSessionKey in runBeforeReset (openclaw#39875) Merged via squash. Prepared head SHA: 00a2b24 Co-authored-by: rbutera <[email protected]> Co-authored-by: altaywtf <[email protected]> Reviewed-by: @altaywtf * CLI: include commit hash in --version output (openclaw#39712) * CLI: include commit hash in --version output * fix(version): harden commit SHA resolution and keep output consistent * CLI: keep install checks compatible with commit-tagged version output * fix(cli): include commit hash in root version fast path * test(cli): allow null commit-hash mocks * Installer: share version parser across install scripts * Installer: avoid sourcing helpers from stdin cwd * CLI: note commit-tagged version output * CLI: anchor commit hash resolution to module root * CLI: harden commit hash resolution * CLI: fix commit hash lookup edge cases * CLI: prefer live git metadata in dev builds * CLI: keep git lookup inside package root * Infra: tolerate invalid moduleUrl hints * CLI: cache baked commit metadata fallbacks * CLI: align changelog attribution with prep gate * CLI: restore changelog contributor credit --------- Co-authored-by: echoVic <[email protected]> Co-authored-by: echoVic <[email protected]> * fix: fail closed talk provider selection * fix: align talk config secret schemas * refactor: require canonical talk resolved payload * test: add talk config contract fixtures * refactor: split talk gateway config loaders * refactor: avoid checkout during prep head verification * refactor: dedupe prep branch push flow * fix: treat model api drift as baseUrl refresh * refactor: extract pure models config merge helpers * refactor: expand provider capability registry * refactor: reuse one models.json read per write * fix: publish models.json atomically * refactor: scope prep push results to env artifacts * refactor: fold implicit provider injection into resolver * fix(telegram): use message previews in DMs * CI: scope CodeQL JavaScript analysis * feat: allow compaction model override via config (openclaw#38753) Merged via squash. Prepared head SHA: a3d6d6c Co-authored-by: starbuck100 <[email protected]> Co-authored-by: jalehman <[email protected]> Reviewed-by: @jalehman * fix(sessions): clear stale contextTokens on model switch (openclaw#38044) Merged via squash. Prepared head SHA: bac2df4 Co-authored-by: yuweuii <[email protected]> Co-authored-by: jalehman <[email protected]> Reviewed-by: @jalehman * fix: prefer bundled channel plugins over npm duplicates (openclaw#40094) * fix: prefer bundled channel plugins over npm duplicates * fix: tighten bundled plugin review follow-ups * fix: address check gate follow-ups * docs: add changelog for bundled plugin install fix * fix: align lifecycle test formatting with CI oxfmt * docs: update Brave Search API docs for Feb 2026 plan restructuring (openclaw#40111) Merged via squash. Prepared head SHA: c651f07 Co-authored-by: remusao <[email protected]> Co-authored-by: gumadeiras <[email protected]> Reviewed-by: @gumadeiras * Add too-many-prs override label handling * CI: satisfy provider merge fixture typing * Tests: reduce web search secret-scan noise * Web search: rename Perplexity auth source helper * Docs: use placeholder OpenRouter key in Perplexity guide * Docs: use placeholder OpenRouter key in web tool docs * Fixtures: normalize talk config API key placeholder * Tests: lower entropy git commit fixtures * Chore: widen xxxxx detect-secrets allowlist * Chore: refresh detect-secrets baseline * Web search: allowlist Perplexity auth source type name * Chore: refresh detect-secrets baseline after docs line changes * Chore: refresh detect-secrets baseline after final scan * Chore: refresh detect-secrets baseline for Feishu docs * CLI: set local gateway mode in setup * Tests: format daemon lifecycle CLI coverage * refactor: use model compat for anthropic tool payload normalization * refactor: move bundled extension gap allowlists into manifests * refactor: thread config runtime env through models config * refactor: split models registry loading from persistence * refactor: centralize transcript provider quirks * refactor: decompose implicit provider resolution * refactor: extract openai stream wrappers * refactor: validate bundled extension release metadata * refactor: extract static provider builders * refactor: extract provider stream wrappers * refactor: extract bundled extension manifest parser * test: standardize hermetic provider env snapshots * test: add implicit provider matrix coverage * fix(acp): persist spawned child session history (openclaw#40137) Merged via squash. Prepared head SHA: 62de5d5 Co-authored-by: mbelinky <[email protected]> Co-authored-by: mbelinky <[email protected]> Reviewed-by: @mbelinky * fix: require talk resolved payload * refactor: dedupe android talk config parsing * test: expand talk config contract fixtures * refactor: split android talk voice resolution * test: isolate plugin loader from mocked module cache * test: isolate legacy plugin-sdk root import check * test: isolate git commit resolution fallbacks * refactor: simplify plugin sdk compatibility aliases * refactor: centralize acp session resolution guards * refactor: neutralize context engine runtime bridge * refactor: split doctor config analysis helpers * refactor: extract ios watch reply coordinator * fix: restore acp session meta narrowing * refactor: extract qmd process runner * fix: restore gate after rebase * docs: add Browserbase as hosted remote CDP option Add Browserbase documentation section alongside the existing Browserless section in the browser docs. Includes signup instructions, CDP connection configuration, and environment variable setup for both English and Chinese (zh-CN) translations. Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]> * Revert "docs: add Browserbase as hosted remote CDP option" This reverts commit c469657. * docs: add Browserbase as hosted remote CDP option Add Browserbase documentation section alongside the existing Browserless section in the browser docs. Includes signup instructions, CDP connection configuration, and environment variable setup. Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]> * docs: fix duplicate heading lint error Rename "Configuration" sub-heading to "Profile setup" to avoid MD024/no-duplicate-heading conflict with the existing top-level "Configuration" heading. Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]> * docs: fix Browserbase section to match official docs Browserbase requires creating a session via their API to get a CDP connect URL, unlike Browserless which uses a static endpoint. Updated to show the correct curl-based session creation flow, removed unverified static WebSocket URL, and added the 5-minute connect timeout note from official docs. Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]> * docs: restore direct wss://connect.browserbase.com URL Browserbase exposes a direct WebSocket connect endpoint that auto-creates a session, similar to how Browserless works. Simplified the section to use this static URL pattern instead of requiring manual session creation via the API. Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]> * docs: fact-check Browserbase section against official docs - Fix CAPTCHA/stealth/proxy claims: these are Developer plan+ only, not available on free tier - Fix free tier limits: 1 browser hour, 15-min session duration (not "60 minutes of monthly usage") - Add link to pricing page for paid plan details - Simplify structure to match Browserless section format - Remove sub-headings to match Browserless section style Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]> * docs: simplify Browserbase section, drop pricing details Restore platform-level feature description (CAPTCHA solving, stealth mode, proxies) without plan-specific pricing gating. Keep free tier note brief. Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]> * feat(browser): support direct WebSocket CDP URLs for Browserbase Browserbase uses direct WebSocket connections (wss://) rather than the standard HTTP-based /json/version CDP discovery flow used by Browserless. This change teaches the browser tool to accept ws:// and wss:// URLs as cdpUrl values: when a WebSocket URL is detected, OpenClaw connects directly instead of attempting HTTP discovery. Changes: - config.ts: accept ws:// and wss:// in cdpUrl validation - cdp.helpers.ts: add isWebSocketUrl() helper - cdp.ts: skip /json/version when cdpUrl is already a WebSocket URL - chrome.ts: probe WSS endpoints via WebSocket handshake instead of HTTP - cdp.test.ts: add test for direct WebSocket target creation - docs/tools/browser.md: update Browserbase section with correct URL format and notes Co-Authored-By: Claude Opus 4.6 <[email protected]> * test+docs: comprehensive coverage and generic framing - Add 12 new tests covering: isWebSocketUrl detection, parseHttpUrl WSS acceptance/rejection, direct WS target creation with query params, SSRF enforcement on WS URLs, WS reachability probing bypasses HTTP - Reframe docs section as generic "Direct WebSocket CDP providers" with Browserbase as one example — any WSS-based provider works - Update security tips to mention WSS alongside HTTPS Co-Authored-By: Claude Opus 4.6 <[email protected]> * fix(browser): update existing tests for ws/wss protocol support Two pre-existing tests still expected ws:// URLs to be rejected by parseHttpUrl, which now accepts them. Switch the invalid-protocol fixture to ftp:// and tighten the assertion to match the full "must be http(s) or ws(s)" error message. Co-Authored-By: Claude Opus 4.6 <[email protected]> * fix(browser): preserve wss:// cdpUrl in legacy default profile resolution * chore: remove vendor-specific references from code comments * style(browser): fix oxfmt formatting in config.ts Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]> * fix: preserve loopback ws cdp tab ops (openclaw#31085) (thanks @shrey150) * fix: share context engine registry across bundled chunks (openclaw#40115) Merged via squash. Prepared head SHA: 6af4820 Co-authored-by: jalehman <[email protected]> Co-authored-by: jalehman <[email protected]> Reviewed-by: @jalehman * fix(browser): rewrite 0.0.0.0 and [::] wildcard addresses in CDP WebSocket URLs Containerized browsers (e.g. browserless in Docker) report `ws://0.0.0.0:<internal-port>` in their `/json/version` response. `normalizeCdpWsUrl` rewrites loopback WS hosts to the external CDP host:port, but `0.0.0.0` and `[::]` were not treated as addresses needing rewriting, causing OpenClaw to try connecting to `ws://0.0.0.0:3000` literally — which always fails. Fixes openclaw#17752 Co-Authored-By: Claude Opus 4.6 <[email protected]> * fix: normalize wildcard remote CDP websocket URLs (openclaw#17760) (thanks @joeharouni) * fix(browser): wait for extension tabs after relay drop (openclaw#32331) * fix: wait for extension relay tab reconnects (openclaw#32461) (thanks @AaronWander) * fix(infra): make browser relay bind address configurable Add browser.relayBindHost config option so the Chrome extension relay server can bind to a non-loopback address (e.g. 0.0.0.0 for WSL2). Defaults to 127.0.0.1 when unset, preserving current behavior. Closes openclaw#39214 Co-Authored-By: Claude Opus 4.6 <[email protected]> * fix(browser): add IP validation, fix upgrade handler for non-loopback bind - Zod schema: validate relayBindHost with ipv4/ipv6 instead of bare string - Upgrade handler: allow non-loopback connections when bindHost is explicitly non-loopback (e.g. 0.0.0.0 for WSL2), keeping loopback-only default - Test: verify actual bind address via relay.bindHost instead of just checking reachability on 127.0.0.1 which passes regardless - Expose bindHost on ChromeExtensionRelayServer type for inspection Co-Authored-By: Claude Opus 4.6 <[email protected]> * fix: make browser relay bind address configurable (openclaw#39364) (thanks @mvanhorn) * docs: add WSL2 + Windows remote Chrome CDP troubleshooting (openclaw#39407) (thanks @Owlock) * macos: add remote gateway token field for remote mode * macos: clarify remote token placeholder text * macos: add mode-toggle remote token sync coverage * tests: document remote token persistence across mode toggle * fix(macos): preserve unsupported remote gateway tokens * docs(changelog): credit macos remote token author * fix(macos): improve tailscale gateway discovery (openclaw#40167) Sanitized test tailnet hostnames and re-ran the targeted macOS gateway discovery test suite before merge. * refactor(browser): scope CDP sessions and harden stale target recovery * feat: add local backup CLI (openclaw#40163) Merged via squash. Prepared head SHA: ed46625 Co-authored-by: shichangs <[email protected]> Co-authored-by: gumadeiras <[email protected]> Reviewed-by: @gumadeiras * fix(ci): scope secrets scan to branch changes * fix(ci): refresh detect-secrets baseline * fix: harden backup verify path validation * fix(plugin-sdk): lazily load legacy root alias * fix(setup-podman): cd to TMPDIR before podman load to avoid cwd permission error (openclaw#39435) * fix(setup-podman): cd to TMPDIR before podman load to avoid inherited cwd permission error * fix(podman): safe cwd in run_as_user to prevent chdir errors Co-Authored-By: Claude Opus 4.6 <[email protected]> Signed-off-by: sallyom <[email protected]> --------- Signed-off-by: sallyom <[email protected]> Co-authored-by: sallyom <[email protected]> Co-authored-by: Claude Opus 4.6 (1M context) <[email protected]> * fix(cron): consolidate announce delivery, fire-and-forget trigger, and minimal prompt mode (openclaw#40204) * fix(cron): consolidate announce delivery and detach manual runs * fix: queue detached cron runs (openclaw#40204) * Gateway/iOS: replay queued foreground actions safely after resume (openclaw#40281) Merged via squash. - Local validation: `pnpm exec vitest run --config vitest.gateway.config.ts src/gateway/server-methods/nodes.invoke-wake.test.ts` - Local validation: `pnpm build` - mb-server validation: `pnpm exec vitest run --config vitest.gateway.config.ts src/gateway/server-methods/nodes.invoke-wake.test.ts` - mb-server validation: `pnpm build` - mb-server validation: `pnpm protocol:check` * iOS: auto-load the scoped gateway canvas with safe fallback (openclaw#40282) Merged via squash. - mb-server validation: `swift test --package-path apps/shared/OpenClawKit --filter GatewayNodeSessionTests` - mb-server validation: `pnpm build` - Scope note: top-level `RootTabs` shell change was intentionally removed from this PR before merge * Update CHANGELOG.md * Update CHANGELOG.md * fix(run-openclaw-podman): add SELinux :Z mount option on enforcing/permissive hosts (openclaw#39449) * fix(run-openclaw-podman): add SELinux :Z mount option on Linux with enforcing/permissive SELinux * fix(quadlet): add SELinux :Z label to openclaw.container.in volume mount * fix(podman): add SELinux :Z mount option for Fedora/RHEL hosts Co-Authored-By: Claude Opus 4.6 <[email protected]> Signed-off-by: sallyom <[email protected]> --------- Signed-off-by: sallyom <[email protected]> Co-authored-by: sallyom <[email protected]> Co-authored-by: Claude Opus 4.6 <[email protected]> * Docker: trim runtime image payload (openclaw#40307) * Docker: shrink runtime image payload * Docker: add runtime pnpm opt-in * Docker: collapse helper entrypoint chmod layers * Docker: restore bundled pnpm runtime * Update CHANGELOG.md * docs(changelog): move post-2026.3.8 entries to unreleased (openclaw#40342) * docs(changelog): move post-2026.3.8 entries to unreleased * Update CHANGELOG.md * fix(tui): improve color contrast for light-background terminals (openclaw#40345) * fix(tui): improve colour contrast for light-background terminals (openclaw#38636) Detect light terminal backgrounds via COLORFGBG and apply a WCAG AA-compliant light palette. Adds OPENCLAW_THEME=light|dark env var override for terminals without auto-detection. Uses proper sRGB linearisation and WCAG 2.1 contrast ratios to pick whichever text palette (dark or light) has higher contrast against the detected background colour. Co-authored-by: ademczuk <[email protected]> * Update CHANGELOG.md --------- Co-authored-by: ademczuk <[email protected]> Co-authored-by: ademczuk <[email protected]> * fix(models): keep --all aligned with synthetic catalog rows * test(models): refresh list assertions after main sync * fix: normalize openai-codex gpt-5.4 transport overrides * docs: add refactor cluster backlog * refactor: dedupe plugin runtime stores * refactor: share gateway argv parsing * refactor: extract gateway port diagnostics helper * refactor: reuse broadcast route key construction * refactor: share multi-account config schema fragments * test: dedupe brave llm-context rejection cases * refactor: share channel config adapter base * fix(agents): bootstrap runtime plugins before context-engine resolution * docs(changelog): remove rebase marker * refactor: harden browser relay CDP flows * fix(config): refresh runtime snapshot from disk after write. Fixes openclaw#37175 (openclaw#37313) Merged via squash. Prepared head SHA: 69e1861 Co-authored-by: bbblending <[email protected]> Co-authored-by: gumadeiras <[email protected]> Reviewed-by: @gumadeiras * refactor: harden browser runtime profile handling * refactor(models): extract list row builders * refactor(agents): extract provider model normalization * refactor(models): split models.json planning from writes * refactor(models): split provider discovery helpers * chore(docs): drop refactor cleanup tracker * gateway: fix global Control UI 404s for symlinked wrappers and bundled package roots (openclaw#40385) Merged via squash. Prepared head SHA: 567b3ed Co-authored-by: velvet-shark <[email protected]> Co-authored-by: velvet-shark <[email protected]> Reviewed-by: @velvet-shark * Docker: improve build cache reuse (openclaw#40351) * Docker: improve build cache reuse * Tests: cover Docker build cache layout * Docker: fix sandbox cache mount continuations * Docker: document qr-import manifest scope * Docker: narrow e2e install inputs * CI: cache Docker builds in workflows * CI: route sandbox smoke through setup script * CI: keep sandbox smoke on script path * fix(tests): correct security check failure * docs(changelog): correct Control UI contributor credit (openclaw#40420) Merged via squash. Prepared head SHA: e4295fe Co-authored-by: velvet-shark <[email protected]> Co-authored-by: velvet-shark <[email protected]> Reviewed-by: @velvet-shark * fix(models): use 1M context for openai-codex gpt-5.4 (openclaw#37876) Merged via squash. Prepared head SHA: c410207 Co-authored-by: yuweuii <[email protected]> Co-authored-by: jalehman <[email protected]> Reviewed-by: @jalehman * fix(telegram): add download timeout to prevent polling loop hang (openclaw#40098) Merged via squash. Prepared head SHA: abdfa1a Co-authored-by: tysoncung <[email protected]> Co-authored-by: obviyus <[email protected]> Reviewed-by: @obviyus * ACP: add optional ingress provenance receipts (openclaw#40473) Merged via squash. Prepared head SHA: b63e46d Co-authored-by: mbelinky <[email protected]> Co-authored-by: mbelinky <[email protected]> Reviewed-by: @mbelinky * alphabetize web search providers (openclaw#40259) Merged via squash. Prepared head SHA: be6350e Co-authored-by: kesku <[email protected]> Co-authored-by: obviyus <[email protected]> Reviewed-by: @obviyus * fix(plugin-sdk): remove remaining bundled plugin src imports (openclaw#39638) Verified: - pnpm build - pnpm check - pnpm test:macmini Co-authored-by: Kyle <[email protected]> Co-authored-by: Tak Hoffman <[email protected]> * test: fix android talk config contract fixture * chore(acpx): move runtime test fixtures to test-utils (openclaw#40548) Verified: - pnpm install --frozen-lockfile - pnpm build - pnpm check - pnpm test:macmini * fix(agents): re-expose configured tools under restrictive profiles * fix(media): accept reader read result type * build(protocol): sync generated swift models * fix: dedupe inbound Telegram DM replies per agent (openclaw#40519) Merged via squash. Prepared head SHA: 6e235e7 Co-authored-by: obviyus <[email protected]> Co-authored-by: obviyus <[email protected]> Reviewed-by: @obviyus * fix(matrix): restore robust DM routing without the memberCount heuristic (openclaw#19736) * fix(matrix): remove memberCount heuristic from DM detection The memberCount === 2 check in isDirectMessage() misclassifies 2-person group rooms (admin channels, monitoring rooms) as DMs, routing them to the main session instead of their room-specific session. Matrix already distinguishes DMs from groups at the protocol level via m.direct account data and is_direct member state flags. Both are already checked by client.dms.isDm() and hasDirectFlag(). The memberCount heuristic only adds false positives for 2-person groups. Move resolveMemberCount() below the protocol-level checks so it is only reached for rooms not matched by m.direct or is_direct. This narrows its role to diagnostic logging for confirmed group rooms. Refs: openclaw#19739 * fix(matrix): add conservative fallback for broken DM flags Some homeservers (notably Continuwuity) have broken m.direct account data or never set is_direct on invite events. With the memberCount heuristic removed, these DMs are no longer detected. Add a conservative fallback that requires two signals before classifying as DM: memberCount === 2 AND no explicit m.room.name. Group rooms almost always have explicit names; DMs almost never do. Error handling distinguishes M_NOT_FOUND (missing state event, expected for unnamed rooms) from network/auth errors. Non-404 errors fall through to group classification rather than guessing. This is independently revertable — removing this commit restores pure protocol-based detection without any heuristic fallback. * fix(matrix): add parentPeer for DM room binding support Add parentPeer to DM routes so conversations are bindable by room ID while preserving DM trust semantics (secure 1:1, no group restrictions). Suggested by @KirillShchetinin. * fix(matrix): override DM detection for explicitly configured rooms Builds on @robertcorreiro's config-driven approach from openclaw#9106. Move resolveMatrixRoomConfig() before the DM check. If a room matches a non-wildcard config entry (matchSource === "direct") and was classified as DM, override the classification to group. This gives users a deterministic escape hatch for misclassified rooms. Wildcards are excluded from the override to avoid breaking DM routing when a "*" catch-all exists. roomConfig is gated behind isRoom so DMs never inherit group settings (skills, systemPrompt, autoReply). This commit is independently droppable if the scope is too broad. * test(matrix): add DM detection and config override tests - 15 unit tests for direct.ts: all detection paths, priority order, M_NOT_FOUND vs network error handling, edge cases (whitespace names, API failures) - 8 unit tests for rooms.ts: matchSource classification, wildcard safety for DM override, direct match priority over wildcard * Changelog: note matrix DM routing follow-up * fix(matrix): preserve DM fallback and room bindings --------- Co-authored-by: Tak Hoffman <[email protected]> * Fix cron text announce delivery for Telegram targets (openclaw#40575) Merged via squash. Prepared head SHA: 54b1513 Co-authored-by: obviyus <[email protected]> Co-authored-by: obviyus <[email protected]> Reviewed-by: @obviyus * fix: clear plugin discovery cache after plugin installation (openclaw#39752) Verified: - pnpm build - pnpm check - pnpm test:macmini Co-authored-by: GazeKingNuWu <[email protected]> Co-authored-by: Tak Hoffman <[email protected]> * test: fix windows secrets runtime ci * fix(daemon): enable LaunchAgent before bootstrap on restart restartLaunchAgent was missing the launchctl enable call that installLaunchAgent already performs. launchd can persist a "disabled" state after bootout, causing bootstrap to silently fail and leaving the gateway unloaded until a manual reinstall. Fixes openclaw#39211 Co-Authored-By: Claude Opus 4.6 <[email protected]> * fix(daemon): also enable LaunchAgent in repairLaunchAgentBootstrap The repair/recovery path had the same missing `enable` guard as `restartLaunchAgent`. If launchd persists a "disabled" state after a previous `bootout`, the `bootstrap` call in `repairLaunchAgentBootstrap` fails silently, leaving the gateway unloaded in the recovery flow. Add the same `enable` guard before `bootstrap` that was already applied to `installLaunchAgent` and (in this PR) `restartLaunchAgent`. Co-Authored-By: Claude Opus 4.6 <[email protected]> * fix(gateway): exit non-zero on restart shutdown timeout When a config-change restart hits the force-exit timeout, exit with code 1 instead of 0 so launchd/systemd treats it as a failure and triggers a clean process restart. Stop-timeout stays at exit(0) since graceful stops should not cause supervisor recovery. Closes openclaw#36822 * test(secrets): skip ACL-dependent runtime snapshot tests on windows * fix: add changelog for restart timeout recovery (openclaw#40380) (thanks @dsantoreis) * fix(browser): enforce redirect-hop SSRF checks * fix(cron): restore owner-only tools for isolated runs * test(cron): cover owner-only tool availability * fix(msteams): enforce sender allowlists with route allowlists * fix(gateway): validate config before restart to prevent crash + macOS permission loss (openclaw#35862) When 'openclaw gateway restart' is run with an invalid config, the new process crashes on startup due to config validation failure. On macOS, this causes Full Disk Access (TCC) permissions to be lost because the respawned process has a different PID. Add getConfigValidationError() helper and pre-flight config validation in both runServiceRestart() and runServiceStart(). If config is invalid, abort with a clear error message instead of crashing. The config watcher's hot-reload path already had this guard (handleInvalidSnapshot), but the CLI restart/start commands did not. AI-assisted (OpenClaw agent, fully tested) * fix(gateway): catch startup failure in run loop to prevent process exit (openclaw#35862) When an in-process restart (SIGUSR1) triggers a config-triggered restart and the new config is invalid, params.start() throws and the while loop exits, killing the process. On macOS this loses TCC permissions. Wrap params.start() in try/catch: on failure, set server=null, log the error, and wait for the next SIGUSR1 instead of crashing. * test: add runServiceStart config pre-flight tests (openclaw#35862) Address Greptile review: add test coverage for runServiceStart path. The error message copy-paste issue was already fixed in the DRY refactor (uses params.serviceNoun instead of hardcoded 'restart'). * fix: address bot review feedback on openclaw#35862 - Remove dead 'return false' in runServiceStart (Greptile) - Include stack trace in run-loop crash guard error log (Greptile) - Only catch startup errors on subsequent restarts, not initial start (Codex P1) - Add JSDoc note about env var false positive edge case (Codex P1) * fix: move config pre-flight before onNotLoaded in runServiceRestart (Codex P2) The config check was positioned after onNotLoaded, which could send SIGUSR1 to an unmanaged process before config was validated. * fix: release gateway lock on restart failure + reply to Codex reviews - Release gateway lock when in-process restart fails, so daemon restart/stop can still manage the process (Codex P2) - P1 (env mismatch) already addressed: best-effort by design, documented in JSDoc * fix(gateway): detect launchd supervision via XPC_SERVICE_NAME On macOS, launchd sets XPC_SERVICE_NAME on managed processes but does not set LAUNCH_JOB_LABEL or LAUNCH_JOB_NAME. Without checking XPC_SERVICE_NAME, isLikelySupervisedProcess() returns false for launchd-managed gateways, causing restartGatewayProcessWithFreshPid() to fork a detached child instead of returning "supervised". The detached child holds the gateway lock while launchd simultaneously respawns the original process (KeepAlive=true), leading to an infinite lock-timeout / restart loop. Co-Authored-By: Claude Opus 4.6 <[email protected]> * fix: detect launchd supervision via xpc service name (openclaw#20555) (thanks @dimat) * fix(node-host): bind bun and deno approval scripts * fix(skills): pin validated download roots * fix(telegram): abort in-flight getUpdates fetch on shutdown When the gateway receives SIGTERM, runner.stop() stops the grammY polling loop but does not abort the in-flight getUpdates HTTP request. That request hangs for up to 30 seconds (the Telegram API timeout). If a new gateway instance starts polling during that window, Telegram returns a 409 Conflict error, causing message loss and requiring exponential backoff recovery. This is especially problematic with service managers (launchd, systemd) that restart the process immediately after SIGTERM. Wire an AbortController into the fetch layer so every Telegram API request (especially the long-polling getUpdates) aborts immediately on shutdown: - bot.ts: Accept optional fetchAbortSignal in TelegramBotOptions; wrap the grammY fetch with AbortSignal.any() to merge the shutdown signal. - monitor.ts: Create a per-iteration AbortController, pass its signal to createTelegramBot, and abort it from the SIGTERM handler, force-restart path, and finally block. Co-Authored-By: Claude Opus 4.6 <[email protected]> * fix(telegram): use manual signal forwarding to avoid cross-realm AbortSignal AbortSignal.any() fails in Node.js when signals come from different module contexts (grammY's internal signal vs local AbortController), producing: "The signals[0] argument must be an instance of AbortSignal. Received an instance of AbortSignal". Replace with manual event forwarding that works across all realms. Co-Authored-By: Claude Opus 4.6 <[email protected]> * fix: abort telegram getupdates on shutdown (openclaw#23950) (thanks @Gkinthecodeland) * fix(cron): stagger missed jobs on restart to prevent gateway overload When the gateway restarts with many overdue cron jobs, they are now executed with staggered delays to prevent overwhelming the gateway. - Add missedJobStaggerMs config (default 5s between jobs) - Add maxMissedJobsPerRestart limit (default 5 jobs immediately) - Prioritize most overdue jobs by sorting by nextRunAtMs - Reschedule deferred jobs to fire gradually via normal timer Fixes openclaw#18892 * fix: stagger missed cron jobs on restart (openclaw#18925) (thanks @rexlunae) * build: update app deps except carbon * refactor: extract telegram polling session * refactor: split cron startup catch-up flow * refactor: flatten supervisor marker hints * docs: reorder 2026.3.8 changelog by impact * build: sync pnpm lockfile * chore: refresh secrets baseline * docs: move 2026.3.8 entries back to unreleased * test: fix Node 24+ test runner and subagent registry mocks * test: fix Windows fake runtime bin fixtures * fix: normalize windows runtime shim executables * chore: prepare 2026.3.8-beta.1 release * chore: update appcast for 2026.3.8-beta.1 * test: fix windows runtime and restart loop harnesses * fix(update): re-enable launchd service before updater bootstrap * chore: prepare 2026.3.8 npm release * test: narrow gateway loop signal harness * fix(launchd): harden macOS launchagent install permissions * fix(onboard): avoid persisting talk fallback on fresh setup * build: bump unreleased version to 2026.3.9 * fix: stabilize launchd paths and appcast secret scan * build: sync plugin versions for 2026.3.9 * fix(ui): preserve control-ui auth across refresh (openclaw#40892) Merged via squash. Prepared head SHA: f9b2375 Co-authored-by: velvet-shark <[email protected]> Co-authored-by: velvet-shark <[email protected]> Reviewed-by: @velvet-shark * fix(kimi-coding): fix kimi tool format: use native Anthropic tool schema instead of OpenAI … (openclaw#40008) Verified: - pnpm install --frozen-lockfile - pnpm build - pnpm check - pnpm test:macmini Co-authored-by: opriz <[email protected]> Co-authored-by: Tak Hoffman <[email protected]> * fix(swiftformat): exclude HostEnvSecurityPolicy.generated.swift from formatters (openclaw#39969) * test(context-engine): add bundle chunk isolation tests for registry (openclaw#40460) Merged via squash. Prepared head SHA: 44622ab Co-authored-by: dsantoreis <[email protected]> Co-authored-by: jalehman <[email protected]> Reviewed-by: @jalehman * fix(agents): bound compaction retry wait and drain embedded runs on restart (openclaw#40324) Merged via squash. Prepared head SHA: cfd9956 Co-authored-by: cgdusek <[email protected]> Co-authored-by: jalehman <[email protected]> Reviewed-by: @jalehman * Allow ACP sessions.patch lineage fields on ACP session keys (openclaw#40995) Merged via squash. Prepared head SHA: c1191ed Co-authored-by: xaeon2026 <[email protected]> Co-authored-by: mbelinky <[email protected]> Reviewed-by: @mbelinky * Update CONTRIBUTING.md * Add Robin Waslander to maintainers * Update CONTRIBUTING.md * fix(acp): map error states to end_turn instead of unconditional refusal (openclaw#41187) * fix(acp): map error states to end_turn instead of unconditional refusal * fix: map ACP error stop reason to end_turn (openclaw#41187) (thanks @pejmanjohn) --------- Co-authored-by: Pejman Pour-Moezzi <[email protected]> Co-authored-by: Onur <[email protected]> * fix(acp): propagate setSessionMode gateway errors to client (openclaw#41185) * fix(acp): propagate setSessionMode gateway errors to client * fix: add changelog entry for ACP setSessionMode propagation (openclaw#41185) (thanks @pejmanjohn) --------- Co-authored-by: Pejman Pour-Moezzi <[email protected]> Co-authored-by: Onur <[email protected]> * plugins: harden global hook runner state (openclaw#40184) * fix(telegram): bridge direct delivery to internal message:sent hooks (openclaw#40185) * telegram: bridge direct delivery message hooks * telegram: align sent hooks with command session * Cron: enforce cron-owned delivery contract (openclaw#40998) Merged via squash. Prepared head SHA: 5877389 Co-authored-by: mbelinky <[email protected]> Co-authored-by: mbelinky <[email protected]> Reviewed-by: @mbelinky * Agents: add embedded error observations (openclaw#41336) Merged via squash. Prepared head SHA: 4900042 Co-authored-by: altaywtf <[email protected]> Co-authored-by: altaywtf <[email protected]> Reviewed-by: @altaywtf * Doctor: fix non-interactive cron repair gating (openclaw#41386) * iOS: reconnect gateway on foreground return (openclaw#41384) Merged via squash. Prepared head SHA: 0e2e0dc Co-authored-by: mbelinky <[email protected]> Co-authored-by: mbelinky <[email protected]> Reviewed-by: @mbelinky * fix(cron): do not misclassify empty/NO_REPLY as interim acknowledgement (openclaw#41401) * fix(cron): do not misclassify empty/NO_REPLY as interim acknowledgement When a cron task's agent returns NO_REPLY, the payload filter strips the silent token, leaving an empty text string. isLikelyInterimCronMessage() previously returned true for empty input, causing the cron runner to inject a forced rerun prompt ('Your previous response was only an acknowledgement...'). Change the empty-string branch to return false: empty text after payload filtering means the agent deliberately chose silent completion, not that it sent an interim 'on it' message. Fixes openclaw#41246 * fix(cron): do not misclassify empty/NO_REPLY as interim acknowledgement Fixes openclaw#41246. (openclaw#41383) thanks @jackal092927. --------- Co-authored-by: xaeon2026 <[email protected]> * fix(auth): reset cooldown error counters on expiry to prevent infinite escalation (openclaw#41028) Merged via squash. Prepared head SHA: 89bd83f Co-authored-by: zerone0x <[email protected]> Co-authored-by: altaywtf <[email protected]> Reviewed-by: @altaywtf * Gateway: add pending node work primitives (openclaw#41409) Merged via squash. Prepared head SHA: a6d7ca9 Co-authored-by: mbelinky <[email protected]> Co-authored-by: mbelinky <[email protected]> Reviewed-by: @mbelinky * Gateway: tighten node pending drain semantics (openclaw#41429) Merged via squash. Prepared head SHA: 361c2eb Co-authored-by: mbelinky <[email protected]> Co-authored-by: mbelinky <[email protected]> Reviewed-by: @mbelinky * acp: fail honestly in bridge mode (openclaw#41424) Merged via squash. Prepared head SHA: b5e6e13 Co-authored-by: mbelinky <[email protected]> Co-authored-by: mbelinky <[email protected]> Reviewed-by: @mbelinky * acp: restore session context and controls (openclaw#41425) Merged via squash. Prepared head SHA: fcabdf7 Co-authored-by: mbelinky <[email protected]> Co-authored-by: mbelinky <[email protected]> Reviewed-by: @mbelinky * Sandbox: import STATE_DIR from paths directly (openclaw#41439) * acp: enrich streaming updates for ide clients (openclaw#41442) Merged via squash. Prepared head SHA: 0764368 Co-authored-by: mbelinky <[email protected]> Co-authored-by: mbelinky <[email protected]> Reviewed-by: @mbelinky * acp: forward attachments into ACP runtime sessions (openclaw#41427) Merged via squash. Prepared head SHA: f2ac51d Co-authored-by: mbelinky <[email protected]> Co-authored-by: mbelinky <[email protected]> Reviewed-by: @mbelinky * acp: add regression coverage and smoke-test docs (openclaw#41456) Merged via squash. Prepared head SHA: 514d587 Co-authored-by: mbelinky <[email protected]> Co-authored-by: mbelinky <[email protected]> Reviewed-by: @mbelinky * fix(agents): probe single-provider billing cooldowns (openclaw#41422) Merged via squash. Prepared head SHA: bbc4254 Co-authored-by: altaywtf <[email protected]> Co-authored-by: altaywtf <[email protected]> Reviewed-by: @altaywtf * acp: harden follow-up reliability and attachments (openclaw#41464) Merged via squash. Prepared head SHA: 7d167df Co-authored-by: mbelinky <[email protected]> Co-authored-by: mbelinky <[email protected]> Reviewed-by: @mbelinky * Agents: add fallback error observations (openclaw#41337) Merged via squash. Prepared head SHA: 852469c Co-authored-by: altaywtf <[email protected]> Co-authored-by: altaywtf <[email protected]> Reviewed-by: @altaywtf * build(protocol): regenerate Swift models after pending node work schemas (openclaw#41477) Merged via squash. Prepared head SHA: cae0aaf Co-authored-by: mbelinky <[email protected]> Co-authored-by: mbelinky <[email protected]> Reviewed-by: @mbelinky * fix(discord): apply effective maxLinesPerMessage in live replies (openclaw#40133) Merged via squash. Prepared head SHA: 031d032 Co-authored-by: rbutera <[email protected]> Co-authored-by: altaywtf <[email protected]> Reviewed-by: @altaywtf * Logging: harden probe suppression for observations (openclaw#41338) Merged via squash. Prepared head SHA: d18356c Co-authored-by: altaywtf <[email protected]> Co-authored-by: altaywtf <[email protected]> Reviewed-by: @altaywtf * ci(sre:PLA-760): fix smoke workflow fallbacks * test(sre:PLA-760): allowlist secret-scan false positives * test(sre:PLA-760): refresh secret baseline for upstream sync * test(sre:PLA-760): update detect-secrets baseline * ci(sre:PLA-760): exclude auto-response from zizmor * fix(ci:PLA-760): unblock audit and bun test lane * ci(sre:PLA-760): run linux-only ci --------- Signed-off-by: sallyom <[email protected]> Co-authored-by: Peter Steinberger <[email protected]> Co-authored-by: Tak Hoffman <[email protected]> Co-authored-by: Ayaan Zaidi <[email protected]> Co-authored-by: darkamenosa <[email protected]> Co-authored-by: Hermione <[email protected]> Co-authored-by: rbutera <[email protected]> Co-authored-by: altaywtf <[email protected]> Co-authored-by: Altay <[email protected]> Co-authored-by: echoVic <[email protected]> Co-authored-by: echoVic <[email protected]> Co-authored-by: Ayaan Zaidi <[email protected]> Co-authored-by: Vincent Koc <[email protected]> Co-authored-by: GitBuck <[email protected]> Co-authored-by: starbuck100 <[email protected]> Co-authored-by: jalehman <[email protected]> Co-authored-by: yuweuii <[email protected]> Co-authored-by: Rémi <[email protected]> Co-authored-by: remusao <[email protected]> Co-authored-by: gumadeiras <[email protected]> Co-authored-by: Mariano <[email protected]> Co-authored-by: Shrey Pandya <[email protected]> Co-authored-by: Claude Opus 4.6 (1M context) <[email protected]> Co-authored-by: Josh Lehman <[email protected]> Co-authored-by: Joe Harouni <[email protected]> Co-authored-by: AaronWander <[email protected]> Co-authored-by: Matt Van Horn <[email protected]> Co-authored-by: Charles Dusek <[email protected]> Co-authored-by: Nimrod Gutman <[email protected]> Co-authored-by: shichangs <[email protected]> Co-authored-by: Gustavo Madeira Santana <[email protected]> Co-authored-by: langdon <[email protected]> Co-authored-by: sallyom <[email protected]> Co-authored-by: Tyler Yust <[email protected]> Co-authored-by: langdon <[email protected]> Co-authored-by: ademczuk <[email protected]> Co-authored-by: ademczuk <[email protected]> Co-authored-by: Doruk Ardahan <[email protected]> Co-authored-by: 0xsline <[email protected]> Co-authored-by: bbblending <[email protected]> Co-authored-by: Radek Sienkiewicz <[email protected]> Co-authored-by: velvet-shark <[email protected]> Co-authored-by: Tyson Cung <[email protected]> Co-authored-by: obviyus <[email protected]> Co-authored-by: Mariano <[email protected]> Co-authored-by: Kesku <[email protected]> Co-authored-by: Kyle <[email protected]> Co-authored-by: Kyle <[email protected]> Co-authored-by: Bronko <[email protected]> Co-authored-by: GazeKingNuWu <[email protected]> Co-authored-by: GazeKingNuWu <[email protected]> Co-authored-by: scoootscooob <[email protected]> Co-authored-by: Daniel dos Santos Reis <[email protected]> Co-authored-by: DevMac <[email protected]> Co-authored-by: merlin <[email protected]> Co-authored-by: dimatu <[email protected]> Co-authored-by: George Kalogirou <[email protected]> Co-authored-by: rexlunae <[email protected]> Co-authored-by: opriz <[email protected]> Co-authored-by: opriz <[email protected]> Co-authored-by: Joshua Lelon Mitchell <[email protected]> Co-authored-by: dsantoreis <[email protected]> Co-authored-by: Charles Dusek <[email protected]> Co-authored-by: xaeon2026 <[email protected]> Co-authored-by: xaeon2026 <[email protected]> Co-authored-by: Robin Waslander <[email protected]> Co-authored-by: Pejman Pour-Moezzi <[email protected]> Co-authored-by: Pejman Pour-Moezzi <[email protected]> Co-authored-by: Onur <[email protected]> Co-authored-by: zerone0x <[email protected]> Co-authored-by: zerone0x <[email protected]>

openclaw-barnacle bot added the size: S label Mar 6, 2026

greptile-apps bot reviewed Mar 6, 2026

View reviewed changes

jalehman force-pushed the codex/fix-35372-contexttokens-switch branch from ff835f9 to cfcac27 Compare March 8, 2026 17:50

openclaw-barnacle bot added cli CLI command changes scripts Repository scripts agents Agent runtime and tooling labels Mar 8, 2026

ljg and others added 7 commits March 8, 2026 10:51

fix(sessions): clear stale context tokens on model switch

1ede8b9

test(sessions): assert aligned context tokens are preserved

2c30f21

fix: clear stale contextTokens on model switch

3f4bfe9

test: format daemon lifecycle test

652d962

test: reset git commit test module mocks

fc08abd

test: isolate git commit ref suite

bbb3a16

test: isolate child adapter service env

bac2df4

jalehman force-pushed the codex/fix-35372-contexttokens-switch branch from cfcac27 to bac2df4 Compare March 8, 2026 17:53

openclaw-barnacle bot removed the agents Agent runtime and tooling label Mar 8, 2026

jalehman merged commit 6c9b49a into openclaw:main Mar 8, 2026
26 of 28 checks passed

github-actions bot mentioned this pull request Mar 8, 2026

📡 Upstream Digest — 2026-03-08 18:22 UTC curtismercier/openclaw-mods#214

Open

github-actions bot mentioned this pull request Mar 9, 2026

上游更新: v2026.3.8 — 4 P0 + 48 P1 待合并 jiulingyun/openclaw-cn#488

Open

alexey-pelykh mentioned this pull request Mar 10, 2026

Cherry-pick: upstream security/infra/gateway updates (8 commits) remoteclaw/remoteclaw#903

Closed

4 tasks

MoerAI mentioned this pull request Mar 18, 2026

fix(sessions): clear stale context/token cache on model switch #44236

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix(sessions): clear stale contextTokens on model switch#38044

fix(sessions): clear stale contextTokens on model switch#38044
jalehman merged 7 commits intoopenclaw:mainfrom
yuweuii:codex/fix-35372-contexttokens-switch

yuweuii commented Mar 6, 2026

Uh oh!

greptile-apps bot commented Mar 6, 2026

Uh oh!

greptile-apps bot Mar 6, 2026

Uh oh!

yuweuii commented Mar 6, 2026

Uh oh!

Uh oh!

jalehman commented Mar 8, 2026

Uh oh!

yuweuii commented Mar 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

+  it("retains aligned runtime model fields when selection and runtime already match", () => {
+    const before = Date.now() - 5_000;
+    const entry: SessionEntry = {
+      sessionId: "sess-3",
+      updatedAt: before,
+      modelProvider: "openai",
+      model: "gpt-5.2",
+      providerOverride: "openai",
+      modelOverride: "gpt-5.2",
+      contextTokens: 65_536,
+    };
+    const result = applyModelOverrideToSessionEntry({
+      entry,
+      selection: {
+        provider: "openai",
+        model: "gpt-5.2",
+      },
+    });
+    expect(result.updated).toBe(false);
+    expect(entry.modelProvider).toBe("openai");
+    expect(entry.model).toBe("gpt-5.2");
+    expect(entry.contextTokens).toBe(65_536);
+    expect(entry.updatedAt).toBe(before);
+  });

Uh oh!

Conversation

yuweuii commented Mar 6, 2026

Summary

Change Type (select all)

Scope (select all touched areas)

Linked Issue/PR

User-visible / Behavior Changes

Security Impact (required)

Repro + Verification

Environment

Steps

Expected

Actual

Evidence

Human Verification (required)

Compatibility / Migration

Failure Recovery (if this breaks)

Risks and Mitigations

Uh oh!

greptile-apps bot commented Mar 6, 2026

Greptile Summary

Confidence Score: 4/5

Uh oh!

greptile-apps bot Mar 6, 2026

Choose a reason for hiding this comment

Uh oh!

yuweuii commented Mar 6, 2026

Uh oh!

Uh oh!

jalehman commented Mar 8, 2026

Uh oh!

yuweuii commented Mar 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants