Repository navigation
fix(pi): count standalone usage entries - #1484
Conversation
intent(pi-usage): include standard Pi usage records for model calls outside assistant replies. decision(pi-parser): accept arbitrary usage kinds through the shared Pi-format parser and invalidate cached assistant-only results. rejected(pi-usage): provider-specific parsing would tie a standard session feature to one router implementation. constraint(prime-accounting): standalone usage must not enter assistant completion matching, and cached replay must retain positional alignment.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
![cubic-dev-ai[bot]](https://avatars.githubusercontent.com/in/1082092?s=60&v=4){kind=small}
There was a problem hiding this comment.
All reported issues were addressed across 4 files
Reply to a comment to ask cubic a question or push back. It learns from your replies.
View guided diff | Re-trigger cubic
intent(pi-parser): retain assistant usage and agent attribution when other entry types reuse the new standalone field names. decision(pi-parser): tolerate invalid optional standalone fields and infer the provider when its value contains only whitespace. constraint(prime-accounting): a same-ID background usage entry must not replace an assistant attribution target in fresh or cached accounting. decision(pi-docs): mirror native usage coverage in the three maintained README translations.
There was a problem hiding this comment.
All reported issues were addressed across 5 files (changes from recent commits).
Reply with feedback, questions, or to request a fix.
View guided diff | Re-trigger cubic
intent(pi): Resolve PR junhoyeo#1484 merge conflicts by merging upstream main without rewriting branch history. decision(pi): Retain upstream billed operations, reported costs, and legacy model fallback while preserving standalone usage for Prime Agent. constraint(prime-agent): Standalone usage must keep distinct deduplication identities and stay outside assistant completion matching on fresh and cached reads.
|
The parser review found no blocking issue. I pushed the small documentation correction in |
junhoyeo
left a comment
junhoyeo
left a comment
There was a problem hiding this comment.
Reviewed ab0a8c184a3805a77977bfbe55f069d1f2365869.
The standalone usage handling preserves distinct accounting identities, tolerant optional fields, reported costs, and the shared Pi/Prime cache invalidation. The corrected examples include valid parent links in all four languages. Focused tests passed (50 Pi, 36 Prime, 3 cache), JSONL parse/equality checks passed, and formatting/diff checks passed. Fresh native/platform/lint CI is green on the corrected head, and all review threads are resolved.
No blocking findings remain in this revision.
|
Thank you! |
Why
Pi's standalone
usageentries are a standard, documented Pi session feature. They record model usage outside assistant replies. Pi itself uses them for cache warming throughSessionManager.appendUsage().References:
TokScale currently reads assistant messages but skips these standalone records, so their tokens are missing from totals. This change reads the existing standard format. It does not introduce a custom record type, a provider integration, or Jev-specific parsing.
Human-written explanation
I'm doing this because Pi recently added support for Jev, a decision model. Jev can be used as a model router, and redirect the user message to model A or B (like sol or luna). Very useful to redirect to a smaller model when the user's input is simple.
But tokscale only track the usage of the real model (sol, luna, etc), not Jev. With this PR, Pi extensions can write usage from Jev into the session, and tokscale can read the usage. Voilà!
Changes
usage,model,provider, and timestamp fields for standalone usage entries. Accept arbitrarykindvalues.Validation
Seven new regressions cover the shared Pi-compatible clients, arbitrary kinds, malformed records, attribution, fork copies, identity collisions, Prime accounting, and cache invalidation.
cargo fmt --all -- --checkpasses.cargo clippy --locked --workspace --all-features -- -D warningspasses.cargo test --locked --workspace --all-featurespasses: 4,515 tests pass and five existing tests remain ignored.cargo build --locked --release -p tokscale-cli --bin tokscalepasses.Source-reported costs keep the existing Pi parser behavior. TokScale still uses its own price resolver, rather than treating Pi catalog estimates as provider-billed charges.
How it looks like
Summary by cubic
Counts Pi's standalone
usagesession entries so tokens from background model calls (like cache warming) are included in totals instead of being skipped. Also retains the upstream accounting for the other billed operations and their reported costs after merging main.usage,model,provider, and timestamp fields; accepts anykind, skips malformed records, and ignores unrelated metadata that reuses those field names.Written for commit 03ee902. Summary will update on new commits.