Bump inspect-ai from 0.3.235 to 0.3.238

[dependabot]

Bumps inspect-ai from 0.3.235 to 0.3.238.

Changelog

Sourced from inspect-ai's changelog.

0.3.238 (08 June 2026)

• Transcript: Continue using the realtime sample buffer database when WAL journal mode cannot be enabled.
• ACP: Render tool calls for agent-bridge agents (e.g. claude_code, codex_cli).
• ACP: Substitute {{param}} placeholders in tool-call views.
• ACP: Keep the optimistic user · queued chip visible when agents are idle mid-turn.
• Limits: Fix a spurious, mislabeled working limit event recorded alongside the real one when a message/token/cost limit is hit inside a sandboxed agent bridge (e.g. claude_code).
• Inspect View: Improve model event rendering - add INFO tab (usage + stop reason) and a Stop Reason display
• Inspect View: Fix: stop VirtualList from fighting deep-link scroll in WebKit
• Inspect View: Fix: collapse the timeline by default when only main + scoring lanes exist
• Inspect View: Fix: make collapse-all include model events (symmetry with expand-all)
• Inspect View: Improved sample message/transcript placeholders + steady streaming
• Inspect View: Color tool calls in model-event views like the messages tab
• Inspect View: Improve log list responsiveness during bulk sync

0.3.237 (07 June 2026)

• Model API: ChatCompletionChoice.stop_details (StopDetails/StopCategory) surfaces a model's refusal/safety category and explanation when available.
• Transcript: transcript().events now resolves content attachments (large text, images) instead of returning bare attachment:// references when reads are served from the bounded-history provider.
• Transcript: Use WAL journal mode for the realtime sample buffer database so concurrent reads and writes no longer raise OperationalError: database is locked.
• Remove ACP patch for connection initialization order issue (resolved in ACP 0.10.1).

0.3.236 (06 June 2026)

• Model API: Add ModelInfo.family and ModelAPI.model_family(). Provider capability and request-shape checks now consult a registered ModelInfo.family before falling back to model-name matching, while preserving the configured model name for provider requests.
• Together: Support the stream model arg (e.g. -M stream=true) to stream completions. A length-truncated streaming response (with structured output or tools) now degrades gracefully to stop_reason="max_tokens" instead of raising.
• Groq: Map tool_choice="any" to "required" for forced tool calls.
• Bedrock: Support response_schema (structured output) for Claude models via output_config.format.
• Deep Agent: Subagent submit() calls are retained in the subagent transcript (previously stripped) and rendered as markdown, so a submit is distinguishable from a normal assistant message. The parent's result is unchanged.
• Sandbox: Allow sandbox_service() instances running as different users in the same sandbox to share /var/tmp/sandbox-services.
• Sandbox: Read sandbox_service() request payloads with a higher output limit (150 MiB) than the default exec output cap.
• Agent Bridge: By default the bridge no longer forwards client generation-tuning parameters (e.g. max_tokens, temperature, reasoning effort/tokens) to the resolved Inspect model, leaving these parameters entirely determined by the evaluation config. Pass forward_generation_config=True to agent_bridge()/sandbox_agent_bridge() to restore previous behavior.
• Agent Intervention: Support connecting to all samples (disabling interruption and user messages if the agent doesn't explicitly support ACP).
• Docker Compose: accept platform, extra_hosts, cap_add, cap_drop, security_opt, and tmpfs in ComposeService.
• Docker Sandbox: SandboxTimeoutError now carries truncated_output with the partial command output captured before a timeout (surfaced to tool callers), instead of discarding it.
• Transcript: Bound resident memory for long-running samples by evicting older events to a history provider (opt-in via the INSPECT_TRANSCRIPT_BOUNDED environment variable). transcript().events remains a full, compatible view; use transcript().history for memory-aware access.
• Inspect View: New ViewerConfig (passed via Task(viewer=...)) lets eval authors customize how a task's sample list, score panel, and scanner results render in the log viewer — including sample-list columns, default sort, score labels, and color scales. See Custom Views.
• Transcript: Revert disabling of buffer history database when running tests.
• Inspect View: Dark mode, event and message color support.
• Inspect View: Migrated the sample transcript to a virtualized list for smoother rendering of long transcripts.
• Inspect View: Collapse same-name nested solver/agent spans in transcripts.
• Inspect View: Fix truncation of long transcript outlines when scrolling.
• Inspect View: Switch to dev dependencies for scout and inspect apps.
• Inspect View: Fixes to live sample following behavior.
• Inspect View: Properly clear log/sample before paint when viewing log (#284).
• Inspect View: don't render primitive solvers as sub-agents (#287).
• Bugfix: Inspect View sample-list columns now expand to fill the available width.
• Bugfix: Avoid emitting empty assistant output messages when converting Chat Completions tool-call with reasoning into Responses API input items.
• Bugfix: Preserve OpenAI Responses API encrypted reasoning through agent bridge round-trips and replay reasoning input items with empty content to avoid server validation errors.
• Bugfix: Agent bridge checks for google.genai more defensively (ensure that module not found is raised).

Commits

• 18f833a update changelog for release
• d84ad8f misc fixes discovered during claude_code acp testing (#4182)
• f49aa06 Remove duplicate changelog entry for realtime sample buffer
• 66e7bf5 Add /app-config endpoint to view server (#4142)
• ab8658f Bump to latest (#4176)
• 43e96a1 viewer config: use covariant Mapping/Sequence for union-valued fields (#4175)
• 53d3413 ACP: Render tool calls for agent-bridge agents (e.g. claude_code, `codex_cl...
• c3492fa Transcript: tolerate unavailable WAL journal mode (#4174)
• 23da047 update changelog for release
• b9b788b don't use resident events for acp snapshot
• Additional commits viewable in compare view

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

---

Dependabot commands and options

You can trigger Dependabot actions by commenting on this PR:

• @dependabot rebase will rebase this PR
• @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
• @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
• @dependabot ignore this major version will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this minor version will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this dependency will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)