[world] When resolveData='none', only strip eventData related to refs, keep other metadata

[karthikscale3]

Changes:

• Selective field stripping instead of full eventData removal — When resolveData='none', previously the entire eventData object was stripped. Now only known ref/payload fields (input, output, result, error, metadata, payload) are removed per event type, preserving metadata like stepName and workflowName. This allows the UI to display step and workflow names without requiring a full data resolve.

• Consistent implementation across all three world backends — The new stripEventDataRefs function (renamed from filterEventData) with an EVENT_DATA_REF_FIELDS lookup table is applied identically in world-local, world-postgres, and world-vercel.

• Updated test expectations — The world-local storage test now asserts that step_created events retain { stepName: 'test-step' } in eventData instead of eventData being undefined.

• web - Update web to fetch data withData: false

[VaguelySerious]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

[karthikscale3]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

Aligned! Will update.

[karthikscale3]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

Aligned! Will update.

Updated

[VaguelySerious]

LGTM, we do still need a tiny bit of test coverage, and maybe de-duplicate the data, please address those before merging

[karthikscale3]

@VaguelySerious - FYI I have updated web to fetch data withData: false as this change effectively fixes the core issue and I have verified it.

[karthikscale3]

Changes:

• Selective field stripping instead of full eventData removal — When resolveData='none', previously the entire eventData object was stripped. Now only known ref/payload fields (input, output, result, error, metadata, payload) are removed per event type, preserving metadata like stepName and workflowName. This allows the UI to display step and workflow names without requiring a full data resolve.

• Consistent implementation across all three world backends — The new stripEventDataRefs function (renamed from filterEventData) with an EVENT_DATA_REF_FIELDS lookup table is applied identically in world-local, world-postgres, and world-vercel.

• Updated test expectations — The world-local storage test now asserts that step_created events retain { stepName: 'test-step' } in eventData instead of eventData being undefined.

• web - Update web to fetch data withData: false

[VaguelySerious]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

[karthikscale3]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

Aligned! Will update.

[karthikscale3]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

Aligned! Will update.

Updated

[VaguelySerious]

LGTM, we do still need a tiny bit of test coverage, and maybe de-duplicate the data, please address those before merging

[karthikscale3]

@VaguelySerious - FYI I have updated web to fetch data withData: false as this change effectively fixes the core issue and I have verified it.

[karthikscale3]

Changes:

• Selective field stripping instead of full eventData removal — When resolveData='none', previously the entire eventData object was stripped. Now only known ref/payload fields (input, output, result, error, metadata, payload) are removed per event type, preserving metadata like stepName and workflowName. This allows the UI to display step and workflow names without requiring a full data resolve.

• Consistent implementation across all three world backends — The new stripEventDataRefs function (renamed from filterEventData) with an EVENT_DATA_REF_FIELDS lookup table is applied identically in world-local, world-postgres, and world-vercel.

• Updated test expectations — The world-local storage test now asserts that step_created events retain { stepName: 'test-step' } in eventData instead of eventData being undefined.

• web - Update web to fetch data withData: false

[VaguelySerious]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

[karthikscale3]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

Aligned! Will update.

[karthikscale3]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

Aligned! Will update.

Updated

[VaguelySerious]

LGTM, we do still need a tiny bit of test coverage, and maybe de-duplicate the data, please address those before merging

[karthikscale3]

@VaguelySerious - FYI I have updated web to fetch data withData: false as this change effectively fixes the core issue and I have verified it.

[karthikscale3]

Changes:

• Selective field stripping instead of full eventData removal — When resolveData='none', previously the entire eventData object was stripped. Now only known ref/payload fields (input, output, result, error, metadata, payload) are removed per event type, preserving metadata like stepName and workflowName. This allows the UI to display step and workflow names without requiring a full data resolve.

• Consistent implementation across all three world backends — The new stripEventDataRefs function (renamed from filterEventData) with an EVENT_DATA_REF_FIELDS lookup table is applied identically in world-local, world-postgres, and world-vercel.

• Updated test expectations — The world-local storage test now asserts that step_created events retain { stepName: 'test-step' } in eventData instead of eventData being undefined.

• web - Update web to fetch data withData: false

[VaguelySerious]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

[karthikscale3]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

Aligned! Will update.

[karthikscale3]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

Aligned! Will update.

Updated

[VaguelySerious]

LGTM, we do still need a tiny bit of test coverage, and maybe de-duplicate the data, please address those before merging

[karthikscale3]

@VaguelySerious - FYI I have updated web to fetch data withData: false as this change effectively fixes the core issue and I have verified it.

[karthikscale3]

Changes:

• Selective field stripping instead of full eventData removal — When resolveData='none', previously the entire eventData object was stripped. Now only known ref/payload fields (input, output, result, error, metadata, payload) are removed per event type, preserving metadata like stepName and workflowName. This allows the UI to display step and workflow names without requiring a full data resolve.

• Consistent implementation across all three world backends — The new stripEventDataRefs function (renamed from filterEventData) with an EVENT_DATA_REF_FIELDS lookup table is applied identically in world-local, world-postgres, and world-vercel.

• Updated test expectations — The world-local storage test now asserts that step_created events retain { stepName: 'test-step' } in eventData instead of eventData being undefined.

• web - Update web to fetch data withData: false

[VaguelySerious]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

[karthikscale3]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

Aligned! Will update.

[karthikscale3]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

Aligned! Will update.

Updated

[VaguelySerious]

LGTM, we do still need a tiny bit of test coverage, and maybe de-duplicate the data, please address those before merging

[karthikscale3]

@VaguelySerious - FYI I have updated web to fetch data withData: false as this change effectively fixes the core issue and I have verified it.

[karthikscale3]

Changes:

• Selective field stripping instead of full eventData removal — When resolveData='none', previously the entire eventData object was stripped. Now only known ref/payload fields (input, output, result, error, metadata, payload) are removed per event type, preserving metadata like stepName and workflowName. This allows the UI to display step and workflow names without requiring a full data resolve.

• Consistent implementation across all three world backends — The new stripEventDataRefs function (renamed from filterEventData) with an EVENT_DATA_REF_FIELDS lookup table is applied identically in world-local, world-postgres, and world-vercel.

• Updated test expectations — The world-local storage test now asserts that step_created events retain { stepName: 'test-step' } in eventData instead of eventData being undefined.

• web - Update web to fetch data withData: false

[VaguelySerious]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

[karthikscale3]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

Aligned! Will update.

[karthikscale3]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

Aligned! Will update.

Updated

[VaguelySerious]

LGTM, we do still need a tiny bit of test coverage, and maybe de-duplicate the data, please address those before merging

[karthikscale3]

@VaguelySerious - FYI I have updated web to fetch data withData: false as this change effectively fixes the core issue and I have verified it.

[karthikscale3]

Changes:

• Selective field stripping instead of full eventData removal — When resolveData='none', previously the entire eventData object was stripped. Now only known ref/payload fields (input, output, result, error, metadata, payload) are removed per event type, preserving metadata like stepName and workflowName. This allows the UI to display step and workflow names without requiring a full data resolve.

• Consistent implementation across all three world backends — The new stripEventDataRefs function (renamed from filterEventData) with an EVENT_DATA_REF_FIELDS lookup table is applied identically in world-local, world-postgres, and world-vercel.

• Updated test expectations — The world-local storage test now asserts that step_created events retain { stepName: 'test-step' } in eventData instead of eventData being undefined.

• web - Update web to fetch data withData: false

[VaguelySerious]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

[karthikscale3]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

Aligned! Will update.

[karthikscale3]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

Aligned! Will update.

Updated

[VaguelySerious]

LGTM, we do still need a tiny bit of test coverage, and maybe de-duplicate the data, please address those before merging

[karthikscale3]

@VaguelySerious - FYI I have updated web to fetch data withData: false as this change effectively fixes the core issue and I have verified it.

[karthikscale3]

Changes:

• Selective field stripping instead of full eventData removal — When resolveData='none', previously the entire eventData object was stripped. Now only known ref/payload fields (input, output, result, error, metadata, payload) are removed per event type, preserving metadata like stepName and workflowName. This allows the UI to display step and workflow names without requiring a full data resolve.

• Consistent implementation across all three world backends — The new stripEventDataRefs function (renamed from filterEventData) with an EVENT_DATA_REF_FIELDS lookup table is applied identically in world-local, world-postgres, and world-vercel.

• Updated test expectations — The world-local storage test now asserts that step_created events retain { stepName: 'test-step' } in eventData instead of eventData being undefined.

• web - Update web to fetch data withData: false

[VaguelySerious]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

[karthikscale3]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

Aligned! Will update.

[karthikscale3]

@karthikscale3 I think the right solution here is to generally specifically strip out all refs, and no other metadata, instead of special casing properties we want. That'll be more generic and future-proof. We can call all the filter functions stripEventDataRefs or similar for clarity. I renamed the PR to that affect, leaving implementation to you

Aligned! Will update.

Updated

[VaguelySerious]

LGTM, we do still need a tiny bit of test coverage, and maybe de-duplicate the data, please address those before merging

[karthikscale3]

@VaguelySerious - FYI I have updated web to fetch data withData: false as this change effectively fixes the core issue and I have verified it.