perf(server): skip full-message reads while streaming - #9032
All clear
All clear
Details
Note
Your check run agent prompt is: .macroscope/check-run-agents/effect-service-conventions.md
More information about how Check Run Agents work can be found in our Help Center.
🔍 Agent activity — 4 tool calls
💭 I'm checking whether the standalone shape interface pattern is something the PR introduces or just extends from existing code, since conventions discourage flagging pre-existing patterns as new violations—only genuinely introduced issues within the changed scope should count.
1. browse_code — apps/server/src/persistence/Services/ProjectionThreadMessages.ts
2. browse_code — apps/server/src/persistence/Layers/ProjectionThreadMessages.ts
💭 I'm weighing whether the existing style violations in the module are pre-existing issues to leave alone versus something I should flag, since the PR is extending rather than refactoring the service, but it does still count as consuming/extending it per the conventions I need to apply. Requiring a full file-layout migration would be unrelated cleanup though, and the guidance says to only report concrete violations in the changed scope. I'm also checking whether
toPersistenceSqlErroris a trivial error-construction helper worth flagging, but it looks pre-existing and widely used, so probably not. Now I want to look at the attachments handling change in ProjectionPipeline.ts to check it against the conventions.
3. browse_code — apps/server/src/orchestration/Layers/ProjectionPipeline.ts
4. browse_code — apps/server/src/persistence/Errors.ts
💭 I'm checking that the new
appendStreamingusage oftoPersistenceSqlErroris fine since it's just reusing an existing legacy helper consistently with sibling methods, not introducing a new one-off error-wrapping helper. The namespace imports and schema struct usage look correct, and tests were added for the behavior change as required. The interface addition follows the existing pattern of defining the shape separately and exporting the layer, so nothing here breaks convention. I don't see any Effect-convention violations like misuse of Layer or unsafe runtime calls — the diff seems clean, so I'll conclude with "All clear," though I need to reconcile that with the JSON output format the system expects.
Agent Credits: 34 credits
Token usage: 6 uncached input · 58,237 cache-read input · 37,198 cache-write input · 2,459 output
Agent Credits may also include non-token charges from external tools such as web research.