feat(sdk,core): preserve chat.agent context after cancel / OOM / crash

[deepshekhardas]

When a \chat.agent\ run dies mid-stream (the user cancels, the worker OOMs, an unhandled exception kills the process), the next continuation run now reconstructs the conversation context automatically.

Changes

• Session recovery boot for chat.agent
• New \onRecoveryBoot\ hook for custom recovery policies
• Schema fix: \data\ is now \z.unknown()\ instead of \z.string()\
• Added recovery-boot.test.ts and replay tests

Test plan

• Build: \pnpm run build --filter @trigger.dev/sdk --filter @trigger.dev/core\
• SDK unit tests: \pnpm --filter @trigger.dev/sdk exec vitest run\

Closes triggerdotdev#3671

---

Summary by cubic

Preserves chat.agent context across cancel/OOM/crash so continuation runs resume the same conversation. Delivered via Sessions-backed recovery in @trigger.dev/sdk/@trigger.dev/core with a new boot hook; meets the recovery goals from Linear triggerdotdev#3671.

• New Features

  • Sessions-backed chat.agent runtime reconstructs conversation state on continuation runs.
  • onBoot lifecycle hook to rehydrate locals/state once per worker before handling messages.
  • chat.history helpers (getPendingToolCalls, extractNewToolResults, etc.) for HITL flows.
  • Stamp gen_ai.conversation.id on all spans/metrics inside chat runs for cross-run tracing.
  • mockChatAgent test harness to unit-test agents offline without the runtime.

• Bug Fixes

  • Cap idempotencyKey at 2048 chars with a structured 400 at the API boundary.
  • Retry SIGSEGV task crashes under the task’s existing retry policy.
  • Runs API: add region filter/field for list/retrieve.
  • Fix LocalsKey<T> type incompatibility across dual-package builds.

^{Written for commit 98f2903. Summary will update on new commits. Review in cubic}

[cubic-dev-ai]

4 issues found across 1513 files

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name=".changeset/runs-list-region-filter.md">

<violation number="1" location=".changeset/runs-list-region-filter.md:1">
P1: Changeset content does not match PR changes. The changeset describes 'Add region to the runs list / retrieve API' while the PR is about preserving chat.agent context after cancel/OOM/crash with session recovery. This will generate incorrect changelog entries for @trigger.dev/core and @trigger.dev/sdk.</violation>
</file>

<file name=".github/workflows/e2e-webapp.yml">

<violation number="1" location=".github/workflows/e2e-webapp.yml:67">
P2: Gate DockerHub login on both username and token; checking only username can run the step with missing credentials and fail the workflow.</violation>
</file>

<file name=".github/workflows/publish-worker-v4.yml">

<violation number="1" location=".github/workflows/publish-worker-v4.yml:69">
P2: Semver releases no longer publish the `v4-beta` alias tag, so consumers tracking that channel will stop receiving updates.</violation>
</file>

<file name=".github/workflows/claude-md-audit.yml">

<violation number="1" location=".github/workflows/claude-md-audit.yml:52">
P2: Use the PR's base branch dynamically instead of hardcoding `origin/main`</violation>
</file>

<sub>Note: This PR contains a large number of files. cubic only reviews up to 100 files per PR, so some files may not have been reviewed. cubic prioritizes the most important files to review.
On a pro plan you can use ultrareview for larger PRs.

Re-trigger cubic</sub>

[cubic-dev-ai]

P1: Changeset content does not match PR changes. The changeset describes 'Add region to the runs list / retrieve API' while the PR is about preserving chat.agent context after cancel/OOM/crash with session recovery. This will generate incorrect changelog entries for @trigger.dev/core and @trigger.dev/sdk.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At .changeset/runs-list-region-filter.md, line 1:

<comment>Changeset content does not match PR changes. The changeset describes 'Add region to the runs list / retrieve API' while the PR is about preserving chat.agent context after cancel/OOM/crash with session recovery. This will generate incorrect changelog entries for @trigger.dev/core and @trigger.dev/sdk.</comment>

<file context>
@@ -0,0 +1,6 @@
+---
+"@trigger.dev/core": patch
+"@trigger.dev/sdk": patch
</file context>

[cubic-dev-ai]

P2: Gate DockerHub login on both username and token; checking only username can run the step with missing credentials and fail the workflow.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At .github/workflows/e2e-webapp.yml, line 67:

<comment>Gate DockerHub login on both username and token; checking only username can run the step with missing credentials and fail the workflow.</comment>

<file context>
@@ -0,0 +1,97 @@
+
+      # ..to avoid rate limits when pulling images
+      - name: 🐳 Login to DockerHub
+        if: ${{ env.DOCKERHUB_USERNAME }}
+        uses: docker/login-action@4907a6ddec9925e35a0a9e82d7399ccc52663121 # v4.1.0
+        with:
</file context>

Suggested change

	if: ${{ env.DOCKERHUB_USERNAME }}
	if: ${{ secrets.DOCKERHUB_USERNAME && secrets.DOCKERHUB_TOKEN }}

[cubic-dev-ai]

P2: Semver releases no longer publish the v4-beta alias tag, so consumers tracking that channel will stop receiving updates.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At .github/workflows/publish-worker-v4.yml, line 69:

<comment>Semver releases no longer publish the `v4-beta` alias tag, so consumers tracking that channel will stop receiving updates.</comment>

<file context>
@@ -62,26 +65,24 @@ jobs:
-            image_tags=$image_tags,$ref_without_tag:v4-beta
-          fi
+          ref_without_tag=ghcr.io/triggerdotdev/${STEPS_GET_REPOSITORY_OUTPUTS_REPO}
+          image_tags=$ref_without_tag:${STEPS_GET_TAG_OUTPUTS_TAG}
 
           echo "image_tags=${image_tags}" >> "$GITHUB_OUTPUT"
</file context>

Suggested change

	image_tags=$ref_without_tag:${STEPS_GET_TAG_OUTPUTS_TAG}
	image_tags=$ref_without_tag:${STEPS_GET_TAG_OUTPUTS_TAG}
	if [[ "$STEPS_GET_TAG_OUTPUTS_IS_SEMVER" == true ]]; then
	image_tags=$image_tags,$ref_without_tag:v4-beta
	fi

[cubic-dev-ai]

P2: Use the PR's base branch dynamically instead of hardcoding origin/main

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At .github/workflows/claude-md-audit.yml, line 52:

<comment>Use the PR's base branch dynamically instead of hardcoding `origin/main`</comment>

<file context>
@@ -0,0 +1,73 @@
+
+            ## Your task
+
+            1. Run `git diff origin/main...HEAD --name-only` to see which files changed in this PR.
+            2. For each changed directory, check if there's a CLAUDE.md in that directory or a parent directory.
+            3. Determine if any CLAUDE.md or .claude/rules/ file should be updated based on the changes. Consider:
</file context>