feat(temporal): opt-in continue-as-new for long-lived agent workflows

[danielmillerp]

Why

Long-lived chat/session agents run as a single Temporal workflow that stays open indefinitely. Two things killed them:

1. Event history grows until it hits Temporal's ~50k-event / 50MB limit, then the workflow stalls.
2. The workflow execution timeout (24h SDK default) terminated the whole chain — a user returning to an old chat hit a dead workflow and their message silently vanished.

This adds an opt-in continue-as-new pattern to BaseWorkflow so a session can stay open forever, and defaults the execution timeout to infinite so a long-lived chat isn't capped at 24h.

Design

Opt-in by adoption, no flag, no patch gate. An agent gets recycling only by calling run_until_complete from its @workflow.run instead of a bare wait_condition(timeout=None). Agents that keep the old wait are untouched. (There's no workflow.patched() gate: we have no in-flight long-running workflows to preserve, and per Temporal guidance the right tool for evolving these "pure entity" workflows later is Worker Versioning + upgrade-on-continue-as-new, not accumulating patches.)

BaseWorkflow helpers (src/agentex/lib/core/temporal/workflows/workflow.py):

• run_until_complete(*args, is_complete, can_recycle=True, timeout=None) — keeps the workflow open; recycles history when Temporal suggests it. timeout mirrors the old wait_condition(timeout=...) (durable idle bound); can_recycle lets an agent opt out when a recycle prerequisite is missing.
• should_continue_as_new() — recycle when workflow.info().is_continue_as_new_suggested() (Temporal owns the threshold).
• drain_and_continue_as_new() — waits all_handlers_finished (so an in-flight turn isn't lost) and re-checks completion before workflow.continue_as_new.
• is_continued_run() — the hook agents use to gate state rehydration after a recycle.

Execution timeout (environment_variables.py + temporal_task_service.py + temporal_client.py): WORKFLOW_EXECUTION_TIMEOUT_SECONDS now defaults to None = no execution timeout (None/0/negative → execution_timeout=None). The execution timeout is chain-wide (continue-as-new does NOT reset it), so capping it would still kill a forever-chat. To bound idle workflows, use a durable timer (run_until_complete's timeout), not this ceiling.

Scope (deliberately concise)

This PR is just the pattern + the timeout default. Restoring state after a recycle is framework-specific (rebuild from adk.messages, an adk.state snapshot, or a framework's own memory like a LangGraph checkpointer / Pydantic AI history) and is intentionally left to follow-up PRs, one per integration. The only example touched is 000_hello_acp, which keeps no cross-turn state and so proves the pattern with zero rehydration.

Verification

• Unit tests for the decision helpers (should_continue_as_new, is_continued_run) — tests/lib/core/temporal/test_base_workflow_continue_as_new.py.
• py_compile + ruff + pyright clean.
• Follow-up: replay/integration test of drain_and_continue_as_new against a Temporal test server.

🤖 Generated with Claude Code

Greptile Summary

This PR adds opt-in Temporal continue-as-new support for long-lived agent workflows. The main changes are:

• New BaseWorkflow helpers for recycle decisions, handler draining, continued-run detection, and wait-loop adoption.
• Default workflow execution timeout changed to unlimited, with positive env values still supported.
• The 000_hello_acp Temporal example now opts into the new wait helper.
• Unit tests cover the pure continue-as-new decision helpers.

Confidence Score: 4/5

The workflow infrastructure change is localized and tested, but the example opt-in path has a visible duplicate initialization-message regression when recycling.

The main helpers and timeout behavior are covered by focused tests, and the remaining issue is confined to the example workflow’s continued-run initialization path rather than the core helper implementation.

examples/tutorials/10_async/10_temporal/000_hello_acp/project/workflow.py

T-Rex Logs

What T-Rex did

• Reproduced the skip-continued prologue scenario by running a focused Python harness that imports the real hello ACP workflow class and mocks Temporal and ADK boundaries, then executed an initial run to call run_until_complete and a continued run with continued_run_id, observing two duplicate acknowledgement messages.
• Validated the execution_timeout handling by testing various before/after validation cases, confirming that both before and after validation runs exit with code 0 after dependencies are installed and showing the captured command, cwd, commit, scenario output, and exit code in the artifacts.
• Performed workflow recycling validation by comparing pre- and post-run logs and confirming a successful run (RESULT PASS) with timeout forwarded, no continue_as_new in some cases, and proper drain behavior, aided by the provided recycle script artifacts.

_{Ran code and verified through T-Rex}

Comments Outside Diff (2)

1. examples/tutorials/10_async/10_temporal/000_hello_acp/project/workflow.py, line 56-63 (link)

   Gate create acknowledgement

   This workflow now opts into continue-as-new, so @workflow.run executes from the top on every recycled run. Because the creation acknowledgement is emitted unconditionally before run_until_complete, each recycle writes another “You should only see this message once” message to the task history and UI. Gate this one-time prologue with is_continued_run() so continued runs do not repeat it.

   Artifacts

   
   

   Repro: focused workflow harness that executes the real on_task_create twice across original and continued runs

   • Contains supporting evidence from the run (text/x-python; charset=utf-8).

   Stack trace captured during the T-Rex run

   • Keeps the raw stack trace available without making the summary code-heavy.

   

   _{Ran code and verified through T-Rex}

   Prompt To Fix With AI

   This is a comment left during a code review.
   Path: examples/tutorials/10_async/10_temporal/000_hello_acp/project/workflow.py
   Line: 56-63
   
   Comment:
   **Gate create acknowledgement**
   
   This workflow now opts into continue-as-new, so `@workflow.run` executes from the top on every recycled run. Because the creation acknowledgement is emitted unconditionally before `run_until_complete`, each recycle writes another “You should only see this message once” message to the task history and UI. Gate this one-time prologue with `is_continued_run()` so continued runs do not repeat it.
   
   How can I resolve this? If you propose a fix, please make it concise.

   

2. examples/tutorials/10_async/10_temporal/000_hello_acp/project/workflow.py, line 57-63 (link)

   Skip continued prologue

   run_until_complete(params, ...) forwards the same params into every continued run, so @workflow.run starts over here after each recycle. Because this acknowledgement is emitted before the wait helper and is not gated by self.is_continued_run(), every continue-as-new appends another “You should only see this message once” message to the task history. This creates visible duplicate initialization messages for any long-lived chat that recycles.

   Artifacts

   
   

   Repro: generated workflow continue-as-new harness

   • Contains supporting evidence from the run (text/x-python; charset=utf-8).

   Repro: harness output showing duplicate acknowledgement messages

   • Keeps the command output available without making the summary code-heavy.

   

   _{Ran code and verified through T-Rex}

   Prompt To Fix With AI

   This is a comment left during a code review.
   Path: examples/tutorials/10_async/10_temporal/000_hello_acp/project/workflow.py
   Line: 57-63
   
   Comment:
   **Skip continued prologue**
   
   `run_until_complete(params, ...)` forwards the same `params` into every continued run, so `@workflow.run` starts over here after each recycle. Because this acknowledgement is emitted before the wait helper and is not gated by `self.is_continued_run()`, every continue-as-new appends another “You should only see this message once” message to the task history. This creates visible duplicate initialization messages for any long-lived chat that recycles.
   
   How can I resolve this? If you propose a fix, please make it concise.

   

Prompt To Fix All With AI

Fix the following 1 code review issue. Work through them one at a time, proposing concise fixes.

---

### Issue 1 of 1
examples/tutorials/10_async/10_temporal/000_hello_acp/project/workflow.py:57-63
**Skip continued prologue**

`run_until_complete(params, ...)` forwards the same `params` into every continued run, so `@workflow.run` starts over here after each recycle. Because this acknowledgement is emitted before the wait helper and is not gated by `self.is_continued_run()`, every continue-as-new appends another “You should only see this message once” message to the task history. This creates visible duplicate initialization messages for any long-lived chat that recycles.

_{Reviews (12): Last reviewed commit: "feat(temporal): opt-in continue-as-new f..." | Re-trigger Greptile}