trunk-merge/pr-89156/15cdeb55-b6d4-42fc-adbb-2b2605afdcda - #107630
trunk-io[bot] wants to merge 311 commits into
Conversation
…from the proposals api
…he workflow history
mypy rejected `**dict[str, object]` against fetch_app_metric_totals' typed keyword parameters. Annotate the parameterised filter mapping as dict[str, Any] so the unpack type-checks.
A double-click in the SQL editor selects one identifier, and the save dialogs treat any selection as the query to save. The dialog then sends that fragment to the API, which rejects it as an invalid query. The dialogs now parse a selected candidate before the person can submit, and hold the save with a message that names the selection as the cause. Only a selection is judged, because clearing it is always a way forward. The whole editor contents are left to the API. If the parser fails to load the check passes, so a missing parser cannot block a save. The candidate preview also moves above the name field, so the query being saved is visible before anything is typed. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The message told the person to clear the selection, which cannot be done while the dialog is open. It now names closing the dialog first. Adds a dialog-level test for the view, endpoint and metric dialogs. The helper tests cover the parse verdict but not the wiring, so a dropped saveTarget error would have let a selected identifier submit. Generated-By: PostHog Desktop Task-Id: 7b0812e1-1261-40dc-bdad-aad6a7172281
Generated-By: PostHog Desktop Task-Id: 7b0812e1-1261-40dc-bdad-aad6a7172281
In raw mode the text goes to the connection's own engine in its own dialect, so parsing it as HogQL refused valid SQL. The save dialogs now pass sendRawQueryEnabled and the check stands down. Also makes the refusal message name the last step, saving again. Generated-By: PostHog Desktop Task-Id: 7b0812e1-1261-40dc-bdad-aad6a7172281
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016F8jY3Du54fcQUdCQkr3ZE
…and validate daily Review round on the Temporal layer: - kickoff holds the pipeline row lock through run_training, as start_training does, and a failed launch rolls back the deduction and raises for the activity retry - discovery pauses a pipeline whose creator lost team access and skips one outside the flag rollout - validation runs on every sweep; scoring and kickoff run on a cadence day compared by calendar day, and kickoff starts after scoring ends - scoring and validation re-read the pipeline status - the queue defaults to the general-purpose fleet, a deploy keeps the schedule's paused state, and child timeouts cover every retry
… once per day Second review round on the Temporal layer: - kickoff applies the Desktop access and Tasks usage gates that /train applies, before the row lock - kickoff launches at most one run per UTC day, so a retry after a lost activity response cannot start a second paid run - scoring reads the current champion inside its activity, so a retry after a mid-scoring promotion scores with the new one - discovery passes the loaded organization id to the flag check - the general-purpose worker redeploys on autoresearch backend changes
…nator run - inference and validation heartbeat through HeartbeaterSync with a two-minute heartbeat timeout, so a lost worker is retried in minutes - the schedule gives the coordinator a twelve-hour execution timeout, so a stuck run ends before the next tick instead of making SKIP drop it - the temporal AGENTS.md says which logic stays in the sweep and which lives in the packages it calls
…unch Kickoff waits for scoring, hours after discovery checked the flag, so the launch gate checks the autoresearch rollout again.
Discovery leaves out a pipeline whose organization is deactivated (a null is_active included) or pending deletion, and the kickoff gate checks again before a paid launch.
…-templates-ai-first-new
|
Navigate logical layers of code changes, visualize relationships, and explore their blast radius. 📝 WalkthroughWalkthroughThe changes add HogQL traversal metadata to published catalogs, autoresearch Temporal workflows and scheduling, configurable calendar sync intervals, and HeyReach and MoEngage warehouse sources. They also change Hogland sandbox URL validation and request routing, add flag-based hiding and replacement guidance for MCP loop tools, and update worker deployment coverage. Priority: ➖ Normal Merge Risk: 🟡 Moderate · up to The new HeyReach source can skip earlier pages when a sync is retried after a completed run. The daily autoresearch job can time out as the number of pipelines grows. The loop-to-workflow guidance can send users to steps that fail for active loops or omit earlier run history. These issues should be fixed before merge. Security Architecture ReviewSecurity architecture risk: 🟡 Moderate · up to Scheduled inference can repeat externally visible prediction work after an interrupted attempt. The review found controls that limit cross-team execution and sandbox routing, but several integration boundaries remain only partially verified. Retained concerns
Security review detailsSecurity Blast Radius
Security Findings and Attack Paths
Trust Boundaries and Controls
Resilience and Maintainability Implications
Hardening Proposals
🚥 Pre-merge checks | ❌ 1❌ Failed checks (1 warning)
Full details: Description checkExplanation The description only documents Trunk Merge batch metadata, tested pull requests, and dependencies. It does not include the required Problem, Changes, testing, Release status, Automatic notifications, Docs update, or Agent context sections. Resolution Replace or supplement the Trunk Merge text with a standalone repository-formatted description. Explain the user problem, visible changes, tests and untested areas, select exactly one release-status option, state changelog and docs actions, and complete or remove the Agent context section as applicable.
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 9
Note
Quiet mode is enabled, so only the most important comments were posted inline. Other review comments are grouped below.
🟡 Other comments (3)
posthog/hogql/catalog_traversal.py-186-204 (1)
186-204: 🚀 Performance & Scalability | 🟡 Minor | ⚡ Quick winCache the negative result in
_canonical_name.The early return on Line 190-191 skips the cache. Line 199-201 also returns without writing to the cache. When serialization raises, each later lookup of the same target runs
serialize_fieldsagain and incrementscanonical_unserializableagain. The per-path checks at Line 176 and Line 247 repeat this work, and the omission counts grow on every repeat. StoreNonein the cache before returning.Proposed fix
except Exception: self.omissions["canonical_unserializable"] += 1 + self.canonical_name_cache[id(target)] = None return Noneproducts/tasks/backend/presentation/views/api.py-3469-3469 (1)
3469-3469: 🩺 Stability & Availability | 🟡 Minor | ⚡ Quick winReject malformed ports as invalid sandbox URLs. If
connection.sandbox_urlcontains a nonnumeric port, the delegated check readstarget.portoutside itstryblock and raisesValueError. The command endpoint then returns a server error instead of its invalid-URL response. Catch port-parsing errors inproducts/tasks/backend/logic/services/agent_command.pyand returnFalse.products/customer_analytics/frontend/scenes/CustomerAnalyticsConfigurationScene/calendar/calendarSyncLogic.ts-263-263 (1)
263-263: 🎯 Functional Correctness | 🟡 Minor | ⚡ Quick winKeep the saved interval visible when status refresh fails. After a successful POST,
loadStatuses()starts an asynchronous refresh. Its error handler returns the previous statuses, whilefinallyclears the saving state. If that refresh fails for an idle integration, the selector shows the old interval again until another load occurs. Update the displayed interval after the successful POST, and let a later refresh reconcile it. (keajs.org)
🧹 Nitpick comments (2)
posthog/hogql/catalog_traversal.py (1)
249-249: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueHandle a missing chain component explicitly.
When a
FieldTraversernames a field that does not exist,current.fields[str(component)]raisesKeyError. The broadexcept Exceptionin_relation_for_fieldcatches that error, so the catalog still builds. The problem is that this relies on the broad handler. RaiseResolutionErrordirectly so the handler's intent stays clear and a real bug does not get recorded astarget_unresolvable.Proposed fix
- next_value = current.fields[str(component)] + next_value = current.fields.get(str(component)) + if next_value is None: + raise ResolutionError("catalog traversal reached an unknown field")products/warehouse_sources/backend/temporal/data_imports/sources/heyreach/heyreach.py (1)
109-115: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueAdd the missing return annotation on
_top_level_rows.The function has no return type. The repo guideline says "Annotate every signature". Use
-> Resource, or the type thatrest_api_resourcereturns.Source: Coding guidelines
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository: PostHog/posthog/.coderabbit.yaml
Review profile: QUIET
Plan: Enterprise
Run ID: 6de4135d-bf00-4698-b23d-f361e6cdf548
⛔ Files ignored due to path filters (3)
products/customer_analytics/frontend/generated/api.schemas.tsis excluded by!**/generated/**products/customer_analytics/frontend/generated/api.tsis excluded by!**/generated/**products/customer_analytics/frontend/generated/api.zod.tsis excluded by!**/generated/**
📒 Files selected for processing (59)
.github/workflows/container-images-cd.ymldocs/internal/hogql-language-service.mdposthog/api/services/query.pyposthog/api/test/test_query_service.pyposthog/hogql/catalog_traversal.pyposthog/hogql/language_service.pyposthog/hogql/test/test_language_service.pyposthog/management/commands/start_temporal_worker.pyposthog/settings/temporal.pyposthog/temporal/schedule.pyproducts/autoresearch/backend/dataset/AGENTS.mdproducts/autoresearch/backend/evaluation/AGENTS.mdproducts/autoresearch/backend/facade/temporal.pyproducts/autoresearch/backend/inference/AGENTS.mdproducts/autoresearch/backend/temporal/AGENTS.mdproducts/autoresearch/backend/temporal/CLAUDE.mdproducts/autoresearch/backend/temporal/__init__.pyproducts/autoresearch/backend/temporal/schedule.pyproducts/autoresearch/backend/temporal/test_workflows.pyproducts/autoresearch/backend/temporal/workflows.pyproducts/autoresearch/backend/training/AGENTS.mdproducts/customer_analytics/backend/facade/api.pyproducts/customer_analytics/backend/facade/contracts.pyproducts/customer_analytics/backend/logic/calendar_sync.pyproducts/customer_analytics/backend/presentation/views/serializers.pyproducts/customer_analytics/backend/presentation/views/views.pyproducts/customer_analytics/backend/temporal/calendar_sync.pyproducts/customer_analytics/backend/test/test_calendar_sync.pyproducts/customer_analytics/backend/test/test_views.pyproducts/customer_analytics/frontend/scenes/CustomerAnalyticsConfigurationScene/calendar/CalendarSyncConfig.tsxproducts/customer_analytics/frontend/scenes/CustomerAnalyticsConfigurationScene/calendar/calendarSyncLogic.tsproducts/customer_analytics/mcp/tools.yamlproducts/tasks/backend/facade/api.pyproducts/tasks/backend/presentation/views/api.pyproducts/tasks/backend/tests/test_api.pyproducts/tasks/backend/tests/test_sandbox_url_validation.pyproducts/tasks/mcp/tools.yamlproducts/warehouse_sources/backend/temporal/data_imports/sources/SOURCES.mdproducts/warehouse_sources/backend/temporal/data_imports/sources/generated_configs/heyreach.pyproducts/warehouse_sources/backend/temporal/data_imports/sources/generated_configs/moengage.pyproducts/warehouse_sources/backend/temporal/data_imports/sources/heyreach/canonical_descriptions.pyproducts/warehouse_sources/backend/temporal/data_imports/sources/heyreach/heyreach.pyproducts/warehouse_sources/backend/temporal/data_imports/sources/heyreach/settings.pyproducts/warehouse_sources/backend/temporal/data_imports/sources/heyreach/source.pyproducts/warehouse_sources/backend/temporal/data_imports/sources/heyreach/tests/test_heyreach.pyproducts/warehouse_sources/backend/temporal/data_imports/sources/heyreach/tests/test_heyreach_source.pyproducts/warehouse_sources/backend/temporal/data_imports/sources/moengage/canonical_descriptions.pyproducts/warehouse_sources/backend/temporal/data_imports/sources/moengage/moengage.pyproducts/warehouse_sources/backend/temporal/data_imports/sources/moengage/settings.pyproducts/warehouse_sources/backend/temporal/data_imports/sources/moengage/source.pyproducts/warehouse_sources/backend/temporal/data_imports/sources/moengage/tests/test_moengage.pyproducts/warehouse_sources/backend/temporal/data_imports/sources/moengage/tests/test_moengage_source.pyservices/mcp/schema/generated-tool-definitions.jsonservices/mcp/schema/tool-definitions-all.jsonservices/mcp/schema/tool-definitions.jsonservices/mcp/scripts/generate-tools.tsservices/mcp/scripts/yaml-config-schema.tsservices/mcp/src/api/generated.tsservices/mcp/tests/unit/tool-filtering.test.ts
💤 Files with no reviewable changes (4)
- products/autoresearch/backend/evaluation/AGENTS.md
- products/autoresearch/backend/training/AGENTS.md
- products/autoresearch/backend/inference/AGENTS.md
- products/autoresearch/backend/dataset/AGENTS.md
Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 4 remain after this review.
| active = await workflow.execute_activity( | ||
| activity_load_active_pipelines, | ||
| LoadActivePipelinesInput(), | ||
| start_to_close_timeout=timedelta(minutes=2), | ||
| retry_policy=_COORDINATOR_RETRY, | ||
| ) |
There was a problem hiding this comment.
🩺 Stability & Availability | 🟠 Major | ⚡ Quick win
Size the discovery activity for per-pipeline network calls.
activity_load_active_pipelines loops over every live pipeline across all teams. For each pipeline, it runs _resolve_acting_user, which is a DB query. It also runs has_autoresearch_access, which can make a remote flag evaluation through posthog_feature_flag_value or feature_enabled(only_evaluate_locally=False).
The activity has a 2-minute start_to_close_timeout and no heartbeat. When the pipeline count grows or flag latency rises, the activity times out, and each of the 3 retries repeats the whole sweep. If all retries time out, the daily tick fails and no pipeline is validated or scored.
To fix this, either raise the timeout and add HeartbeaterSync() with a heartbeat_timeout, or split discovery into bounded pages.
Proposed fix
active = await workflow.execute_activity(
activity_load_active_pipelines,
LoadActivePipelinesInput(),
- start_to_close_timeout=timedelta(minutes=2),
+ start_to_close_timeout=timedelta(minutes=15),
+ heartbeat_timeout=_HEARTBEAT_TIMEOUT,
retry_policy=_COORDINATOR_RETRY,
)Also wrap the activity body in with HeartbeaterSync():.
📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| active = await workflow.execute_activity( | |
| activity_load_active_pipelines, | |
| LoadActivePipelinesInput(), | |
| start_to_close_timeout=timedelta(minutes=2), | |
| retry_policy=_COORDINATOR_RETRY, | |
| ) | |
| active = await workflow.execute_activity( | |
| activity_load_active_pipelines, | |
| LoadActivePipelinesInput(), | |
| start_to_close_timeout=timedelta(minutes=15), | |
| heartbeat_timeout=_HEARTBEAT_TIMEOUT, | |
| retry_policy=_COORDINATOR_RETRY, | |
| ) |
| counts: CalendarSyncCounts | ||
|
|
||
|
|
||
| def get_calendar_sync_interval(config: dict) -> int: |
There was a problem hiding this comment.
📐 Maintainability & Code Quality | 🟠 Major | ⚡ Quick win
Specify the calendar config types.
The bare dict annotation leaves key and value types implicit in this new signature. Use a typed mapping, such as dict[str, object], and keep the exact-integer check when narrowing the stored value. As per coding guidelines, “Write as if mypy --strict were on. Annotate every signature, avoid Any.”
Source: Coding guidelines
| sync_interval_minutes = serializers.IntegerField(help_text="Minutes between scheduled syncs: 5, 15, 30, or 60.") | ||
|
|
||
| def validate_sync_interval_minutes(self, value: int) -> int: | ||
| if value not in (5, 15, 30, 60): |
There was a problem hiding this comment.
🗄️ Data Integrity & Integration | 🟠 Major | ⚡ Quick win
Use one set of allowed intervals.
This tuple duplicates ALLOWED_SYNC_INTERVALS in products/customer_analytics/backend/logic/calendar_sync.py. If the two lists diverge, the API can save an interval that the status getter and scheduler treat as 60 minutes. Move the choices to a shared contract and use them in both places. As per path instructions, code should “say everything once and only once.”
Source: Path instructions
| summary="Set Google account sync interval", | ||
| ) | ||
| @action(methods=["POST"], detail=False, url_path="interval") | ||
| def interval(self, request: ValidatedRequest, *args, **kwargs) -> Response: |
There was a problem hiding this comment.
📐 Maintainability & Code Quality | 🟠 Major | ⚡ Quick win
Annotate the new action’s variadic parameters.
interval leaves *args and **kwargs unannotated. Add parameter types to this new signature. As per coding guidelines, “Write as if mypy --strict were on. Annotate every signature, avoid Any.”
Source: Coding guidelines
| }, | ||
| ) | ||
|
|
||
| def test_sync_interval_defaults_and_is_scoped_to_one_account(self): |
There was a problem hiding this comment.
📐 Maintainability & Code Quality | 🟠 Major | ⚡ Quick win
Annotate the new test signatures. Add -> None to all three new test methods. As per coding guidelines, “Write as if mypy --strict were on. Annotate every signature.”
Also applies to: 3545-3545, 3562-3562
Source: Coding guidelines
| "https://hogland.prod-us.posthog.dev/v1/hogboxes/box-1/proxy/8080/command", | ||
| ) | ||
|
|
||
| def test_command_blocks_hogland_sandbox_url_when_unconfigured(self): |
There was a problem hiding this comment.
📐 Maintainability & Code Quality | 🟠 Major | ⚡ Quick win
Annotate the new test signatures. The new tests omit parameter or return annotations.
products/tasks/backend/tests/test_api.py#L14326-L14326: add-> None.products/tasks/backend/tests/test_api.py#L14299-L14299: annotate both mock parameters asMagicMockand add-> None.products/tasks/backend/tests/test_api.py#L14349-L14349: annotate both mock parameters asMagicMockand add-> None.products/tasks/backend/tests/test_sandbox_url_validation.py#L44-L44: annotateurl: str,expected: bool, and-> None.
As per coding guidelines, “Write as if mypy --strict were on. Annotate every signature.”
📍 Affects 2 files
products/tasks/backend/tests/test_api.py#L14326-L14326(this comment)products/tasks/backend/tests/test_api.py#L14299-L14299products/tasks/backend/tests/test_api.py#L14349-L14349products/tasks/backend/tests/test_sandbox_url_validation.py#L44-L44
Source: Coding guidelines
| superseded_by: | ||
| - workflows-archive |
There was a problem hiding this comment.
🎯 Functional Correctness | 🟠 Major | ⚡ Quick win
Provide the required disable step before archiving an active loop.
If an active loop receives a loops-destroy request, the redirect names workflows-archive. The supplied workflows-archive definition requires the workflow to be draft or disabled first. Directly following this redirect fails for an active loop. Name the disable step and its tool before workflows-archive, or provide a replacement that accepts active loops.
| superseded_by: | ||
| - tasks-list | ||
| - tasks-runs-retrieve | ||
| - workflows-list-invocations | ||
| redirect_hint: | ||
| 'Loops in this project run on workflows: read the building-loops skill, then use the workflows tools with | ||
| origin_product "loops". For run history, call tasks-list with the hog_flow_id of the loop and archived set | ||
| to all: it lists the tasks the loop created, newest first, each with the status and error of its latest run. | ||
| Then call tasks-runs-retrieve on a run for its branch and output, including any PR URL. |
There was a problem hiding this comment.
🎯 Functional Correctness | 🟠 Major | ⚡ Quick win
Include tasks-runs-list in the run-history redirect.
If a loop-created task has multiple runs, tasks-list exposes only its latest run. The hint then directs the caller to tasks-runs-retrieve, but gives no way to find earlier run IDs. Add tasks-runs-list to the successors and direct callers to list each task's runs before retrieving run details.
| def save_checkpoint(state: Optional[dict[str, Any]]) -> None: | ||
| if state and state.get("offset"): | ||
| resumable_source_manager.save_state(HeyReachResumeConfig(offset=int(state["offset"]))) |
There was a problem hiding this comment.
🩺 Stability & Availability | 🟠 Major | ⚡ Quick win
🔎 Supported by static analysis
🏁 Script executed:
#!/bin/bash
rg -nP '\bclear_state\s*\(' products/warehouse_sources/backend/temporal -g '!**/tests/**' -C3 | head -80Repository: PostHog/posthog
Length of output: 9616
🏁 Script executed:
#!/bin/bash
set -e
printf '%s\n' '--- HeyReach source outline and relevant source ---'
ast-grep outline products/warehouse_sources/backend/temporal/data_imports/sources/heyreach/heyreach.py
sed -n '1,190p' products/warehouse_sources/backend/temporal/data_imports/sources/heyreach/heyreach.py
printf '%s\n' '--- ResumableSourceManager definitions and references ---'
rg -n -C4 'class ResumableSourceManager|def (save_state|clear_state|can_resume)|ResumableSourceManager' products/warehouse_sources/backend/temporal -g '*.py' | head -240
printf '%s\n' '--- SourceResponse consumption and success cleanup candidates ---'
rg -n -C4 'SourceResponse|clear_state\(\)|save_checkpoint' products/warehouse_sources/backend/temporal -g '*.py' | head -300Repository: PostHog/posthog
Length of output: 41700
🏁 Script executed:
#!/bin/bash
set -e
printf '%s\n' '--- locate relevant definitions ---'
rg -l 'class ResumableSourceManager|class SourceResponse|def _run' products/warehouse_sources/backend/temporal -g '*.py'
printf '%s\n' '--- manager implementation ---'
manager=$(rg -l 'class ResumableSourceManager' products/warehouse_sources/backend/temporal -g '*.py' | head -1)
cat -n "$manager"
printf '%s\n' '--- source response and workflow definitions ---'
for file in $(rg -l 'class SourceResponse|def _run' products/warehouse_sources/backend/temporal -g '*.py'); do
printf '\n--- %s ---\n' "$file"
rg -n -C8 'class SourceResponse|def _run' "$file"
doneRepository: PostHog/posthog
Length of output: 42400
🏁 Script executed:
#!/bin/bash
set -e
printf '%s\n' '--- resumable manager ---'
sed -n '1,230p' products/warehouse_sources/backend/temporal/data_imports/sources/common/resumable.py
printf '%s\n' '--- shared import workflow ---'
sed -n '900,1065p' products/warehouse_sources/backend/temporal/data_imports/workflow_activities/import_data_sync.py
printf '%s\n' '--- REST resource and paginator definitions ---'
rg -l 'class OffsetPaginator|def rest_api_resource|resume_hook|initial_paginator_state' products/warehouse_sources/backend/temporal/data_imports/sources/common -g '*.py'
for file in $(rg -l 'class OffsetPaginator|def rest_api_resource|resume_hook|initial_paginator_state' products/warehouse_sources/backend/temporal/data_imports/sources/common -g '*.py'); do
printf '\n--- %s ---\n' "$file"
rg -n -C12 'class OffsetPaginator|def rest_api_resource|resume_hook|initial_paginator_state' "$file"
doneRepository: PostHog/posthog
Length of output: 42553
🏁 Script executed:
#!/bin/bash
set -e
printf '%s\n' '--- cleanup calls in workflow and pipelines ---'
rg -n -C5 'clear_state\s*\(' products/warehouse_sources/backend/temporal/data_imports/workflow_activities products/warehouse_sources/backend/temporal/data_imports/pipelines -g '*.py' || true
printf '%s\n' '--- pipeline run definitions and resumable manager handling ---'
rg -n -C12 'class PipelineNonDLT|class PipelineV3|def run\(|resumable_source_manager' products/warehouse_sources/backend/temporal/data_imports/pipelines -g '*.py' | head -260Repository: PostHog/posthog
Length of output: 35378
Clear the resume state after the final page.
RESTClient.paginate calls save_checkpoint with None after the terminal page. The callback ignores that value. The previous offset can remain in Redis for 24 hours, so a later attempt with the same job ID can skip earlier pages.
🐛 Suggested fix
def save_checkpoint(state: Optional[dict[str, Any]]) -> None:
+ if state is None:
+ resumable_source_manager.clear_state()
+ return
if state and state.get("offset"):
resumable_source_manager.save_state(HeyReachResumeConfig(offset=int(state["offset"])))📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| def save_checkpoint(state: Optional[dict[str, Any]]) -> None: | |
| if state and state.get("offset"): | |
| resumable_source_manager.save_state(HeyReachResumeConfig(offset=int(state["offset"]))) | |
| def save_checkpoint(state: Optional[dict[str, Any]]) -> None: | |
| if state is None: | |
| resumable_source_manager.clear_state() | |
| return | |
| if state and state.get("offset"): | |
| resumable_source_manager.save_state(HeyReachResumeConfig(offset=int(state["offset"]))) |
|
|
15 participants
This pull request was created and is being managed by Trunk Merge.
This pull request is based on the master branch at SHA cd7f5b6241b6f6b661536a2ff70aeb958598c8ae.
See more details about each PR in the batch here:
When CI completes, this pull request will be closed automatically.
Pull Requests Being Tested
This pull request is testing a batch with the changes from pull requests 89156, 107255, and 107252 - batching documentation.
Dependencies
This pull request depends on the changes from pull requests 107389, 102336, 107079, and 104978.