Add memory.qmd.rerank as an opt-out for QMD query reranking when searchMode is query.
When set to false, direct QMD query calls pass --no-rerank and the mcporter unified query tool receives rerank:false. Search and vsearch modes keep their existing behavior.
Refs #61834.
Pre-compaction assistant messages carry thinkingSignature values bound to the
original conversation prefix. After compaction the prefix changes (summarized
content is replaced by the compaction summary), so Anthropic rejects those
signatures with "Invalid signature in thinking block", permanently stalling the
session through gateway restarts.
stripInvalidThinkingSignatures only catches absent/blank signatures; this adds
stripStaleThinkingSignaturesForCompactionReplay (thinking.ts) which identifies
pre-compaction assistant messages by timestamp comparison against the latest
compaction summary and strips their signature fields. Called in
sanitizeSessionHistory (replay-history.ts) before stripInvalidThinkingSignatures
for all signed-thinking providers (Anthropic, Bedrock, Vertex). Also fixes
buildSuccessorEntries (compaction-successor-transcript.ts) to strip only
pre-compaction kept entries when writing the rotation successor JSONL; uses
strict < timestamp boundary so same-instant post-compaction messages are not
affected.
Docs: update transcript-hygiene.md Anthropic and Bedrock sections.
Tests: 8 new cases for stripStaleThinkingSignaturesForCompactionReplay; 1 new
case for buildSuccessorEntries verifying pre/post-compaction signature boundary.
Fixes#90108
Enable Mistral prompt cache keys without long-retention forwarding. Update cached-read pricing and doctor migration for existing Mistral provider config. Fixes#83709.
Preserve OpenAI-compatible replay reasoning when the selected custom or self-hosted model already has reasoning metadata enabled.
The transcript policy now treats existing model metadata as the replay contract instead of requiring a new provider config knob, and the OpenAI-compatible serializer preserves reasoning_content for those routes while keeping stock OpenAI, Gemma 4, and known non-replayable OpenRouter safeguards.
Fixes#88068.
Replaces #88071.
* feat(browser): add optional vision understanding to screenshot tool
* fix(browser): wrap vision output as external content, enforce maxBytes, forward auth profiles
* fix(browser): remove no-op scope/attachments config, drop profile pass-through lacking runtime support
* feat(media-understanding): add profile/preferredProfile to DescribeImageFileWithModelParams and forward to describeImage
* style(browser): add curly braces to satisfy eslint curly rule
* fix(browser): correct tools.browser.enabled help text to match actual behavior
* fix(browser): thread agentDir/workspaceDir from plugin tool context into browser vision
* refactor(browser): move vision config from tools.browser to browser.models
The browser plugin's vision configuration now lives on the top-level
`browser` config namespace (browser.models, browser.visionEnabled,
browser.visionPrompt, etc.) instead of `tools.browser`. This aligns
with the plugin's existing config location and avoids confusion between
tool-level and plugin-level settings.
- Remove tools.browser from ToolsSchema and ToolsConfig
- Add models/vision* fields to BrowserConfig and its zod schema
- Update getBrowserVisionConfig to read from cfg.browser
- Update schema help, labels, and quality test
- Update vision.test.ts to use new config shape
* docs(browser): add screenshot vision configuration section
Document the new browser.models config for automatic screenshot
description via vision models, enabling text-only main models to
reason about web page content.
* fix(browser): remove deliverable media markers from vision result, drop unused import
P1: Vision-success path no longer exposes the raw screenshot as
deliverable media (removes MEDIA: line and details.media.mediaUrl).
This prevents channel delivery from auto-sending sensitive page content
when the intended output is a text description.
P2: Remove unused ToolsMediaUnderstandingSchema import that would fail
noUnusedLocals typecheck.
* fix(browser): add command/args fields to browser models schema
The browser vision model schema uses .strict(), so CLI-type entries
with command/args were rejected by TypeScript. Add these fields to
align with MediaUnderstandingModelSchema.
* chore(browser): remove debug console.log statements
* fix(browser): harden screenshot vision result against MEDIA: directive injection and restore image sanitization on failure fallback
ClawSweeper #84247 review round 2:
P1 (security, high): neutralize line-start MEDIA: directives in vision descriptions
before wrapping with wrapExternalContent. The agent media extractor scans every
browser tool-result text block via splitMediaFromOutput which treats line-start
MEDIA: as a trusted local-media delivery directive, and browser is on the
trusted-media allowlist. Without neutralization, page or vision-provider output
containing 'MEDIA:/tmp/secret.png' could synthesize a channel-deliverable media
artifact from untrusted content. wrapExternalContent itself does not strip
line-start directives. Introduce neutralizeMediaDirectives in vision.ts that
prepends '[neutralized] ' to any line whose trimStart() begins with MEDIA:
(case-insensitive), defanging the parser anchor while keeping the original
text human-readable.
P2 (compatibility): pass resolveRuntimeImageSanitization() to imageResultFromFile
in the vision-failure catch fallback. The non-vision screenshot path already
forwards this option (d5cc0d53b7) so configured agents.defaults.imageMaxDimensionPx
takes effect. Without this fix, any provider timeout/error silently bypasses the
sanitization guard and returns a raw full-resolution screenshot.
Regression coverage:
- vision.test.ts: 6 unit cases for neutralizeMediaDirectives (no-op fast path,
mid-line MEDIA: untouched, line-start defanged, leading-whitespace defanged,
case-insensitive, multiple directives per blob).
- browser-tool.test.ts: 2 integration cases that drive the full screenshot
tool execute path:
- 'neutralizes MEDIA: directives in vision text and does not attach media'
asserts no line matches /^\s*MEDIA:/i in returned text, secret path text
is preserved verbatim, details.media is absent, and imageResultFromFile
is not called on the success path.
- 'preserves screenshot image sanitization on vision failure fallback'
mocks describeImageFileWithModel to reject and asserts the fallback
imageResultFromFile call receives imageSanitization: {maxDimensionPx:1600}
plus the 'browser screenshot vision failed' extraText.
* fix(browser): apply clawsweeper fallback media fix from PR #84247
* refactor: reuse media image understanding for browser screenshots
* refactor: use structured media delivery
* test: update music completion media instruction expectation
* fix: trim buffered reply directive padding
* test: refresh codex prompt snapshots for message media aliases
---------
Co-authored-by: scotthuang <scotthuang@tencent.com>
Co-authored-by: Peter Steinberger <steipete@gmail.com>
Move OpenClaw skills and the routed workspace-memory pointer out of native Codex turn user input and into turn-scoped collaboration developer instructions.
Preserve full MEMORY.md fallback prompt injection, delivery-hint rewrapping, lightweight cron exact prompts, and trajectory reporting for the rendered developer surface.
Co-authored-by: Beru <beru@lastguru.lv>
* feat: add Claude Opus 4.8 support
* fix: omit Vertex Opus sampling overrides
* fix: preserve Opus adaptive thinking levels
* fix: clamp Anthropic max effort support
* fix: use sha256 for QA mock call ids
* fix: type Anthropic transport test model metadata
* test: update PDF model default for Opus 4.8
Compact promoted short-term memory snippets before writing them into MEMORY.md, while keeping the full rehydrated snippet in recall state for ranking/provenance. Adds the deep-dreaming config surface and docs, with the default promoted snippet cap set to 160 estimated tokens.
Verification:
- git diff --check
- fnm exec --using v24.13.0 node scripts/run-vitest.mjs run extensions/memory-core/src/short-term-promotion.test.ts extensions/memory-core/src/dreaming.test.ts src/memory-host-sdk/dreaming.test.ts
- GitHub CI run 26605272497
- CodeQL security run 26605272404
Co-authored-by: AMARA <amara@eyeinthesky.pl>
Summary:
- The PR moves the runtime `HEARTBEAT.md` bootstrap template into `src/agents/templates`, keeps docs templates ... or other workspace files, adds a legacy heartbeat-template doctor repair, and updates package guards/tests.
- PR surface: Source +281, Tests +283, Docs +11, Config +1, Other 0. Total +576 across 15 files.
- Reproducibility: yes. from source inspection: current main loads `HEARTBEAT.md` from the docs template, and ... pty heartbeat file non-empty to the runtime. I did not run a live heartbeat repro in this read-only review.
Automerge notes:
- PR branch already contained follow-up commit before automerge: fix(doctor): recognize heartbeat docs boilerplate
- PR branch already contained follow-up commit before automerge: fix(agents): update heartbeat workspace test
- PR branch already contained follow-up commit before automerge: fix(doctor): tighten heartbeat template repair
Validation:
- ClawSweeper review passed for head e34e85864c.
- Required merge gates passed before the squash merge.
Prepared head SHA: e34e85864c
Review: https://github.com/openclaw/openclaw/pull/85416#issuecomment-4519851630
Co-authored-by: Mason Huang <masonxhuang@tencent.com>
Co-authored-by: clawsweeper[bot] <274271284+clawsweeper[bot]@users.noreply.github.com>
Approved-by: hxy91819
Co-authored-by: hxy91819 <8814856+hxy91819@users.noreply.github.com>
Carry over #82973 and fix#81281 by preserving explicit cacheRetention for OpenAI-compatible completions providers that opt into prompt-cache-key support.
The change keeps explicit cacheRetention suppressed for OpenAI-compatible providers without compat.supportsPromptCacheKey, adds regression coverage for both paths, and updates prompt-caching docs for prompt_cache_key / prompt_cache_retention behavior.
Fixes#81281.
Supersedes #82973.
Co-authored-by: lonexreb <reach2shubhankar@gmail.com>
scripts/docs-spellcheck.sh uses set -u and constructs args=( ... "${write_flag[@]}" ), where write_flag may be an empty array. On bash 3.2 (still the default /bin/bash on macOS), referencing an empty array under set -u raises an unbound variable error. Newer bash (>= 4.4) handles this expression correctly, which is why the script ships green on Linux CI runners.
Switch to the bash 3.2-safe parameter expansion ${write_flag[@]+"${write_flag[@]}"}: it expands to nothing when the array is empty and to the array contents otherwise, preserving --write behavior unchanged.
Also fixes overrideable -> overridable in docs/reference/test.md, which the now-running spellcheck surfaces.
Repro:
bash scripts/docs-spellcheck.sh # was: write_flag[@]: unbound variable, exit 1
bash scripts/docs-spellcheck.sh # now: codespell runs to completion
* fix: harden package URL downloads
Guard package acceptance URL downloads with HTTPS-only validation, no embedded credentials, private/special-use DNS and IP rejection, manual redirect checks, bounded timeout/size limits, pinned lookup, and atomic temp-file writes. Add tooling tests for unsafe URLs, redirect validation, size limits, and successful writes.
* fix: cancel redirect response bodies before closing dispatcher
ClawSweeper P2: the redirect branch in openPackageDownloadResponse cleared
the timeout and awaited dispatcher.close() without first cancelling
response.body. Undici's close() is graceful — it waits for in-flight
requests to complete — so a malicious redirect with a slow/never-ending
body could hang the hardened downloader.
Fix: call response.body?.cancel() before dispatcher.close() to abort the
redirect body immediately.
Test: add a regression test that uses a ReadableStream with an indefinite
interval to simulate a hanging body, and asserts cancel() was called.
Refs: clawsweeper review on PR #85512
* test: harden redirect body cancellation race in regression test
Guard the ReadableStream controller.enqueue() call with a cancelled
flag and try/catch to prevent ERR_INVALID_STATE when the interval
fires after cancel() closes the controller.
* fix: cancel final response body before closing dispatcher in downloadUrl
ClawSweeper P2: the HTTP-error and declared-oversize early-exit paths
in downloadUrl threw before consuming or canceling response.body. The
finally block then cleared the timeout and awaited graceful
dispatcher.close() with the body still open, allowing a slow/never-ending
response to hang release tooling.
Fix: add response.body?.cancel() in the finally block before
dispatcher.close().
Tests: add two regressions:
- HTTP 500 with slow body: asserts cancel() called before dispatcher close
- Declared content-length oversize with slow body: same assertion
* fix: add trusted package URL source policy
* fix: keep package URL resolver dependency-free
* test: cover encoded IPv6 package URL bypasses
* docs: sync package acceptance source overview
* docs: restore release doc formatting
* docs: sync package acceptance trusted-url source
* test: cover dotted IPv4 embedded IPv6 package URLs
* fix: parse dotted IPv4 embedded in IPv6 package URLs
* test: isolate anthropic pruning defaults
* test: move anthropic dated model coverage
---------
Co-authored-by: Peter Steinberger <steipete@gmail.com>
