Repository navigation
Conversation
Streams store records as temporal.sdk.streams.v1.StreamRecord. Only the Python SDK reads it, so it is generated here rather than in the API repo.
Records map to the envelope proto in one place, and the body helpers hash the converted payload before the codec so a retry deduplicates.
The generated v1 modules are rewritten on every proto run, so the warning goes on the hand-written package above them.
Only provider code uses them, so exporting them would commit the package to an interface no application needs.
A stored kind at 3 would decode as SUPERSEDED, which readers refuse. The format test now pins field types and labels too, since a retyped field breaks stored records.
A resumed read started with no attempts seen, so a retry right after the cursor reached the reader as plain DATA with no SUPERSEDED.
Core now owns the envelope proto and builds every stored record, so lang keeps conversion and the two hashes that need the plaintext. The content hash is raw bytes for Core, and the batch digest feeds the store's dedupe.
Deterministic serialization differs across protobuf implementations, so a retry through another SDK wouldn't dedupe. A publishing task does add one marker event, so the package shouldn't promise none.
moetemp
marked this pull request as draft
October 12, 2026 06:58
Owner
Author
|
Replaced by Brook #150. Brook was split in two: Brook now has streams with Activity and client publish and outside readers, and Ford has the Workflow Task publish path. The code is the same, re-cut into new branches. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What changed?
This PR adds
StreamRecordand the hashes lang takes before Core stores the records.StreamRecord[T],RecordKind,SupersessionandCursorgo intemporalio.contrib.streams._body.pytakes the plaintextcontent_hashand the batch digest, both before the codec. Core stores them unchanged._wire.pyturns a record of Core's read answer into aStreamRecord. A supersession that Core synthesized becomes aSUPERSEDEDrecord.Why?
Only lang has the plaintext. With both hashes taken before the codec, a retry through a codec with a fresh nonce still dedupes (DD-11).
How did you test it?
Link to a test plan if any -
tests/contrib/streams/test_records_wire.pyalso checks the digest against the vector Core's tests pin.