* fix(channels): resolve bundled channel catalog from dist/extensions/ in published installs
* refactor(channels): delegate bundled channel catalog loader to resolveBundledPluginsDir
---------
Co-authored-by: Claude <noreply@anthropic.com>
* fix(openrouter): handle reasoning_details field in Qwen3 stream parsing
Add support for the reasoning_details field returned by OpenRouter/Qwen3
models. Previously this field was not recognized, causing payloads=0 and
incomplete turn errors.
- Add reasoning_details handling in processOpenAICompletionsStream
- Extract text from reasoning_details array items with type reasoning.text
- Treat as thinking content, similar to other reasoning fields
- Add test case for reasoning_details handling
Fixes#66833
* fix(openrouter): keep tool calls with reasoning_details
* fix: handle OpenRouter Qwen3 reasoning_details streams (#66905) (thanks @bladin)
* fix: preserve streamed tool calls with reasoning deltas (#66905) (thanks @bladin)
---------
Co-authored-by: bladin <bladin@users.noreply.github.com>
Co-authored-by: Ayaan Zaidi <hi@obviy.us>
* fix(audio): restore allowPrivateNetwork for self-hosted STT endpoints
resolveProviderExecutionContext built the request object passed to
transcribeAudio using only sanitizeConfiguredProviderRequest on the
tool-level config and entry — which strips allowPrivateNetwork. The
provider-level request config (models.providers.*.request) was never
included in the merge, so allowPrivateNetwork:true was silently dropped.
Additionally, resolveProviderRequestPolicyConfig only read allowPrivate
Network from params.allowPrivateNetwork (a direct parameter) and ignored
params.request?.allowPrivateNetwork even when it was present.
Fix both gaps:
- runner.entries.ts: use mergeModelProviderRequestOverrides with
sanitizeConfiguredModelProviderRequest(providerConfig?.request) so
models.providers.*.request.allowPrivateNetwork flows through to the
media execution context
- provider-request-config.ts: fall back to params.request?.allowPrivate
Network when params.allowPrivateNetwork is undefined
Fixes#66691. Regression introduced in v2026.4.14.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* test(media-understanding): assert allowPrivateNetwork flows through resolveProviderExecutionContext
Regression test for the bug where providerConfig.request.allowPrivateNetwork
was dropped when building the AudioTranscriptionRequest passed to media
providers. Verifies that setting allowPrivateNetwork in the provider config
reaches the provider's request object after the fix to use
mergeModelProviderRequestOverrides + sanitizeConfiguredModelProviderRequest.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* test(media-understanding): tighten allowPrivateNetwork regression types
* fix: restore allowPrivateNetwork for self-hosted STT endpoints (#66692) (thanks @jhsmith409)
---------
Co-authored-by: Jim Smith <jhsmith0@me.com>
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: Ayaan Zaidi <hi@obviy.us>
Remove the old qa-lab-runtime shim now that qa-runtime is the only live
consumer seam. This leaves one tiny shared runtime facade instead of two
parallel names for the same private helper surface.
Introduce a tiny generic qa-runtime seam for shared live-lane helpers and
repoint qa-matrix to it. This keeps the qa-lab host split while removing
the host-owned runtime name from runner code.
Drop the old qa-lab-runtime shim/export now that nothing consumes it and
keep the plugin-sdk surface aligned with the new seam.
* fix(context-engine): pass deferred maintenance token budget
Thread tokenBudget through the after-turn runtime context so background context-engine maintenance reuses the real model context window instead of falling back to 128k. Also pass through a best-effort currentTokenCount from the latest call total and make the runtime context type explicit about both fields.
Regeneration-Prompt: |
OpenClaw already passed the real context token budget into direct context-engine calls like afterTurn and assemble, but deferred maintain() reused only the runtimeContext object and that object did not carry tokenBudget. Lossless Claw therefore fell back to 128k during background maintenance, which made budget-trigger fire much more aggressively than the live model context warranted. Thread the real contextTokenBudget into buildAfterTurnRuntimeContext so deferred maintenance receives the same budget, and pass a straightforward best-effort currentTokenCount from the latest call total while the relevant data is already in scope. Keep the change additive, update the runtime-context type, and cover the background maintenance/runtime-context behavior with focused tests.
* fix(context-engine): use prompt usage for deferred maintenance
* Docs: add Anthropic max_tokens investigation memo
Regeneration-Prompt: |
Investigate the reported OpenClaw cron isolated-agent failure where an
Anthropic Haiku run returned "max_tokens: must be greater than or equal to 1".
Do not implement a fix yet. Inspect the cron isolated-agent execution path,
the embedded runner, extra param plumbing, Anthropic transport code, and any
model-selection or token-budget logic that could synthesize maxTokens = 0.
Produce a concise maintainer memo with concrete file references, explain why
cron itself is not the component setting maxTokens, identify the most likely
root cause, describe the smallest repro shape, and recommend the cleanest fix.
* openclaw-e82: guard Anthropic Messages maxTokens
Regeneration-Prompt: |
Fix the Anthropic Messages path so OpenClaw never sends max_tokens <= 0
to Anthropic. Match the positive-number guard already used by the
Anthropic Vertex transport, but keep the change scoped: validate token
limits in src/agents/anthropic-transport-stream.ts where transport
options are resolved and where the final payload is assembled, fall back
to the model limit when a runtime override is zero, fail locally when no
positive token budget exists, and drop non-positive maxTokens from
src/agents/pi-embedded-runner/extra-params.ts so hidden config params do
not leak through. Add focused regression coverage for both the transport
and extra-param forwarding path, and remove the earlier investigation memo
from the branch so the PR diff only contains the fix.
* fix: scope Anthropic max token guard
* fix: document Anthropic max token guard
* fix: floor Anthropic max token overrides
Drop the generated qa-runner catalog and the missing/install placeholder
path for repo-private QA runners. The host should discover bundled QA
commands from manifest plus runtime surface only.
Also trim stale qa-matrix install docs and package metadata so the
source-only QA policy stays consistent.
* fix(mcp): harden loopback request guards
* fix(commit): block staged user log
* Revert pre-commit USER.md guard from this PR
Out of scope for the MCP loopback hardening — keep this PR
focused on the loopback request gate and the bearer-comparison
fix. The pre-commit worklog guard can land separately if
maintainers want it.
* changelog: note MCP loopback constant-time + Origin guard (#66665)
* fix(mcp): allow loopback flows that browsers flag as cross-site
The previous Sec-Fetch-Site early-return rejected legit local
browser callers like a UI hosted on http://localhost:<ui-port>
talking to MCP on http://127.0.0.1:<mcp-port> — browsers report
that host mismatch as cross-site even though both ends are
loopback. checkBrowserOrigin already authorizes those via its
local-loopback matcher (loopback peer + loopback Origin host),
so route every Origin-bearing request through that helper and
let it decide. Native MCP clients (no Origin header) continue to
short-circuit through to the bearer check unchanged.
Adds a regression test asserting that
origin: http://localhost:43123, sec-fetch-site: cross-site
from a loopback peer is accepted with a valid bearer.
---------
Co-authored-by: Devin Robison <drobison@nvidia.com>
* fix(agents): tighten workspace file opens
* fix(agents): clarify symlink rejection tests
* fix(agents): surface unsafe identity reads
* fix(agents): use non-blocking opens for identity reads and write-mode probes
* fix(fssafe): restore symlink read identity check
* fix(worklog): append comment resolution status
* fix(fssafe): close afterOpen handle leaks
* fix(worklog): append comment resolution follow-up
* fix(worklog): drop internal user file
* fix(agents): rethrow unexpected errors in agents.files.get
* changelog: note agents.files fs-safe routing + fd-first realpath (#66636)
* fix(agents): rethrow unexpected errors in agents.files.set too
Match the narrow-SafeOpenError catch pattern that agents.files.get
(commit 633b8f92) and writeWorkspaceFileOrRespond already use, so a
real OS error (ENOSPC, EACCES, EBUSY, ...) surfaces through normal
gateway error handling instead of being masked as
'unsafe workspace file'.
* test(agents): match fsStat/fsLstat mock signatures
The mock functions are declared as
vi.fn(async (..._args: unknown[]) => Stats | null)
so mockImplementation callbacks must accept ...unknown[], not a
narrowed (filePath: string) argument. The narrower signature
works at runtime but trips tsgo's strict type check; switch to
args[0] unpacking so the callbacks match the hoisted mock shape.
---------
Co-authored-by: Devin Robison <drobison@nvidia.com>
