feat(ai-overage): Claude Team per-seat overage → class_ai_overage → cc_overage bullet - #1371
Merged
Conversation
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Run ID: 📥 CommitsReviewing files that changed from the base of the PR and between 7fc849e7e0eba768da9d00fa9b3dc5513d2aabad and 18f8eae. 📒 Files selected for processing (17)
✅ Files skipped from review due to trivial changes (5)
🚧 Files skipped from review as they are similar to previous changes (12)
📝 WalkthroughWalkthroughAdds end-to-end support for a new ChangesClaude Team cc_overage metric pipeline
Sequence Diagram(s)sequenceDiagram
participant Airbyte
participant Bronze as bronze_claude_team.claude_team_overage_spend
participant StagingModel as claude_team__ai_overage
participant Silver as silver.class_ai_overage
participant Gold as insight.ai_bullet_rows / ai_person_period
participant API as analytics-api metrics.query_ref
Airbyte->>Bronze: ingest overage_spend rows
Bronze->>StagingModel: dedup to latest snapshot per (account_uuid, month)
StagingModel->>StagingModel: compute overage_cents, is_over_limit, overage_metrics_json
StagingModel->>Silver: union_by_tag insert (tagged silver:class_ai_overage)
Silver->>Gold: ai_bullet_rows UNION branch emits metric_key=cc_overage
Gold->>Gold: ai_person_period classifies cc_overage as sum(metric_value)
Gold->>API: query_ref ARRAY JOIN includes (cc_overage, cc_overage_v)
API->>API: metric_catalog seeded with cc_overage thresholds
Estimated code review effort🎯 4 (Complex) | ⏱️ ~60 minutes Possibly related issues
Suggested reviewers
Poem
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Contributor
Author
|
Paired FE PR: constructorfabric/insight-front#167 (renders the |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
docs/components/connectors/ai/README.md (1)
201-213:⚠️ Potential issue | 🟡 Minor | ⚡ Quick winAdd language tags to fenced code blocks (Line 201 and Line 333).
These new fences are missing a language identifier and trigger
MD040.Proposed fix
-``` +```text silver/ ├── ai/ │ ├── class_ai_dev_usage (per-person-day code activity) │ ├── class_ai_assistant_usage (per-person-day chat/cowork/office/web) │ ├── class_ai_api_usage (per-key-or-project-day tokens + multi-modal API) │ ├── class_ai_cost (per-line-item-day financial — org/project total) │ ├── class_ai_overage (per-person-month spend-over-limit — BUILT: Claude Team) │ └── class_ai_audit_log (event-stream — admin actions, compliance) └── _shared/ ├── class_people (existing) — identity unification └── class_ai_directories? (optional — skills, connectors, projects metadata)```diff -``` +```text insight_tenant_id, source_id, unique_key, email, account_id period_month, tool, seat_tier, currency credit_limit_cents, used_amount_cents, overage_cents -- minor units (cents); overage = max(0, used − limit), honest-NULL if no limit is_over_limit, is_enabled, overage_metrics_json source, data_source, collected_at, _version</details> Also applies to: 333-339 <details> <summary>🤖 Prompt for AI Agents</summary>Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.In
@docs/components/connectors/ai/README.mdaround lines 201 - 213, Add the
language identifiertextto the two fenced code blocks that are missing
language tags. The first block contains the directory tree structure starting
with "silver/" and includes class definitions like class_ai_dev_usage and
class_ai_assistant_usage. The second block (also mentioned at line 333) contains
field names like insight_tenant_id, source_id, and period_month. Change both
opening fences from triple backticks with no language to triple backticks
followed bytextto resolve the MD040 linting error.</details> <!-- cr-comment:v1:d83b6511e9223367f898d08e --> _Source: Linters/SAST tools_ </blockquote></details> </blockquote></details>🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. Inline comments: In `@src/backend/services/analytics-api/src/migration/m20260618_000002_seed_claude_team_overage_catalog.rs`: - Around line 175-177: The assert_eq! macro call within the seed_count_is_one() test function violates rustfmt formatting standards due to line length. Reformat the assert_eq! macro invocation to comply with Rust's standard formatting guidelines by breaking it into multiple lines with proper indentation, or run cargo fmt --all to automatically apply the correct formatting to the entire file. In `@src/ingestion/connectors/ai/claude-team/dbt/claude_team__ai_overage.sql`: - Around line 42-53: The deduplication in the latest_per_seat_month CTE uses only account_uuid and the month from _airbyte_extracted_at in the LIMIT 1 BY clause, but this does not match the model's full grain which also includes tenant_id and source_id. This causes records to be incorrectly dropped when the same account_uuid appears across different tenants or sources within the same month. Modify the LIMIT 1 BY clause to include tenant_id and source_id alongside account_uuid and toStartOfMonth(_airbyte_extracted_at) to ensure deduplication respects the complete model grain and prevents cross-source row loss. --- Outside diff comments: In `@docs/components/connectors/ai/README.md`: - Around line 201-213: Add the language identifier `text` to the two fenced code blocks that are missing language tags. The first block contains the directory tree structure starting with "silver/" and includes class definitions like class_ai_dev_usage and class_ai_assistant_usage. The second block (also mentioned at line 333) contains field names like insight_tenant_id, source_id, and period_month. Change both opening fences from triple backticks with no language to triple backticks followed by `text` to resolve the MD040 linting error.🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID:
cf155300-c004-4199-b724-3147300dad74📥 Commits
Reviewing files that changed from the base of the PR and between c32821d and 6bb2c83e74df9459571d50c6974d2b637687812c.
📒 Files selected for processing (16)
docs/components/connectors/ai/README.mddocs/components/connectors/ai/claude-team/specs/DESIGN.mddocs/components/connectors/ai/claude-team/specs/FEATURE.mddocs/components/connectors/ai/claude-team/specs/PRD.mdsrc/backend/services/analytics-api/src/migration/m20260610_000001_fix_ai_label_drift.rssrc/backend/services/analytics-api/src/migration/m20260618_000001_ai_claude_team_overage_metric.rssrc/backend/services/analytics-api/src/migration/m20260618_000002_seed_claude_team_overage_catalog.rssrc/backend/services/analytics-api/src/migration/mod.rssrc/ingestion/connectors/ai/claude-team/connector.yamlsrc/ingestion/connectors/ai/claude-team/dbt/claude_team__ai_overage.sqlsrc/ingestion/connectors/ai/claude-team/dbt/claude_team__bronze_promoted.sqlsrc/ingestion/connectors/ai/claude-team/dbt/schema.ymlsrc/ingestion/connectors/ai/claude-team/descriptor.yamlsrc/ingestion/scripts/migrations/20260618000000_ai-claude-team-overage-gold.sqlsrc/ingestion/silver/ai/class_ai_overage.sqlsrc/ingestion/silver/ai/schema.yml
Contributor
Author
|
Addressed CodeRabbit review in df12eeb5:
Re-validated: |
Introduce class_ai_overage — a unified per-person per-month AI
spend-vs-limit (overage) Silver class — and wire Claude Team as its
first contributor. Designed homogeneous so OpenAI/ChatGPT seats can
join later by emitting the same 19-column positional contract under
tag silver:class_ai_overage (vendor extras → overage_metrics_json,
never new columns).
Claude Team (now unblocked: the proxy sessionKey has Owner role, so
/overage_spend_limits returns 200 with all 149 per-seat rows):
- bronze claude_team_overage_spend gains scalar fields currency /
out_of_credits / used_credits_basis / limit_type (confirmed live)
- promote claude_team_overage_spend to RMT (ADR-0002)
- claude_team__ai_overage.sql: bronze → Silver contributor. Grain =
latest snapshot per (seat, billing month); unique_key carries the
month so a new month never overwrites a prior month's closing value.
UNITS: used_credits / monthly_credit_limit are already cents (USD) —
mapped straight to *_cents with NO ×100 (unlike cost_cents in
claude_team__ai_dev_usage). overage_cents = max(0, used - limit),
honest-NULL when no limit is known.
- class_ai_overage.sql: union_by_tag('silver:class_ai_overage')
- schema.yml docs/tests for both layers
- descriptor 1.2.0 → 1.3.0 (additive; populates only with billing:view/Owner)
Validated: dbt parse + ls green; tag:claude-team+ now builds the
overage chain; tag:silver:class_ai_overage resolves to the contributor.
Will populate on the next claude-team sync (Bronze table created with
0 rows when the session lacked billing:view).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: Sergei Mozhaev <mozhaev.dev@gmail.com>
Address non-blocker review notes on the class_ai_overage introduction: - docs: add class_ai_overage to the AI coverage matrix (docs/components/connectors/ai/README.md) — coverage-by-class row, silver-layer split tree, and a dedicated subsection next to class_ai_cost explaining the distinction (cost = org/project total financial; overage = per-person spend above their own monthly limit), the 19-col contract, units, month grain, and per-provider feeding (Claude Team BUILT; OpenAI/Windsurf future). Notes Claude Team is the first live source despite predating the matrix's column set. - tests: currency gains accepted_values ['USD'] on both staging and silver (extend when a non-USD billing tenant appears). - tests: used_amount_cents gains not_null on silver (always populated; mirrors staging). email stays nullable on silver by design — future OpenAI seats attribute via account_id without email (same precedent as class_ai_dev_usage silver email). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Signed-off-by: Sergei Mozhaev <mozhaev.dev@gmail.com>
Take Claude Team overage from Silver up to the metric API so the frontend can render it. Adds metric_key `cc_overage` (per-seat spend above the monthly credit limit, in cents). - Gold: 20260618000000_ai-claude-team-overage-gold.sql — DROP+CREATE insight.ai_bullet_rows with a new Branch 6 reading silver.class_ai_overage (source='claude_team'): emits cc_overage = overage_cents, dated at toDate(collected_at) so the current month's snapshot lands inside the dashboard's rolling window (period_month's 1st would miss short windows). Branches 1–5 unchanged. honest-NULL: only rows with a computable overage (overage_cents NOT NULL) are emitted; within-limit seats emit a real 0. - analytics-api m20260618_000001: extend Team/IC Bullet AI query_ref with cc_overage_v — countIf-guarded sumIf (honest-NULL like prs), NOT in ACTIVE_LIST → averaged per person over the period like cc_cost. 22 → 23 metric_keys. Shape tests added. - analytics-api m20260618_000002: seed metric_catalog + product-default threshold for cc_overage — label "Claude Overage", unit ¢, higher_is_better=false, good ≤ 0¢ / warn ≤ 5000¢. Tests added. - Register both migrations in migration/mod.rs. Validated: cargo test migration → 85 passed, 0 failed. FE (cyber-insight-front bullet-layout-groups) is the paired change. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Signed-off-by: Sergei Mozhaev <mozhaev.dev@gmail.com>
Review gap: cc_overage was added to the query_ref rollup (Path A) but not to the parallel ai_person_period period-rollup (Path B), so it silently defaulted to avg() — the exact issue #1286 defect class. For a monthly overage snapshot, avg divides by each person's active-day count (understated), while a per-period spend counter must sum. - Gold migration 20260618000000 now also DROP+CREATEs insight.ai_person_period with cc_overage added to the sum branch (twin of cc_cost). Counters→sum, active→max, ratios→avg preserved. - Guard test m20260610_000001: add cc_overage to SUM_KEYS and to the emitted-key list BULLET_ROWS_KEYS (branch 6) so every_bullet_key_is_classified_not_defaulting_to_avg actually covers it; add a focused cc_overage_sums_like_cc_cost spot-check. Documented that the emitted-key list is hand-maintained (false-green risk for any future gold branch key not added here). cargo test migration → 86 passed, 0 failed. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Signed-off-by: Sergei Mozhaev <mozhaev.dev@gmail.com>
The PRD/DESIGN/FEATURE described claude-team as Bronze-only (`dbt_select: ''`, "No Silver, no Gold") but Silver has since landed: `claude_team__ai_dev_usage` → class_ai_dev_usage (INSIGHT-458) and now `claude_team__ai_overage` → class_ai_overage (Gold bullet `cc_overage`, descriptor 1.3.0). Update DESIGN §4.4, PRD §3.2 non-goals, and FEATURE §1 to state Silver landed and point to the cross-connector contract. Verified end-to-end locally on kind-insight with real overage data (149 seats from the dev proxy, Owner role): bronze 149 → silver 149 (formula invariant 0 violations, 33 dbt tests) → gold 148 cc_overage (honest-NULL) → ai_person_period sum-classified (proven 314 vs avg 157) → analytics-api migrations applied to MariaDB (probes green) → stored query_ref run against CH returns cc_overage=2.317¢, 23 keys, no regression. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Signed-off-by: Sergei Mozhaev <mozhaev.dev@gmail.com>
- claude_team__ai_overage: LIMIT 1 BY now includes tenant_id + source_id (was account_uuid + month only) so a multi-tenant / multi-instance bronze_claude_team doesn't drop same-account_uuid rows across tenants /sources as false duplicates — matches the model's full grain. No behaviour change for single-tenant (local e2e: 149 → 149). - docs README: add `text` language to two bare code fences (MD040). - cargo fmt the two touched migration files. Re-validated: cargo test migration 86 passed; fmt --check clean; dbt re-run on local CH still 149 silver rows. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Signed-off-by: Sergei Mozhaev <mozhaev.dev@gmail.com>
…d view The ai_bullet_rows cc_overage branch (20260618000000_ai-claude-team-overage-gold.sql) does CREATE VIEW ... FROM silver.class_ai_overage. CH 24.x validates a view's source tables at CREATE time, so in any env where migrations run before dbt builds the silver model (E2E, fresh deploy) the view creation fails with Code 60 UNKNOWN_TABLE and aborts the whole migration step. Mirrors the existing class_ai_dev_usage / class_support_activity placeholders: an empty ReplacingMergeTree(_version) table tagged INSIGHT_PLACEHOLDER_v1 (dropped and rebuilt by dbt via drop_silver_placeholders_at_start). Columns match the claude_team__ai_overage staging output. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Signed-off-by: Sergei Mozhaev <mozhaev.dev@gmail.com>
Backtick the bare identifiers (claude_team, ComingSoon, cc_overage, cc_cost, silver.class_ai_overage) in the m20260618_* overage migration doc comments and the m20260610_000001 entries touched by this PR. `cargo clippy --all-targets` runs with -D warnings, so these tripped doc_markdown and failed Lint & Test. Doc-comment-only change (clippy --fix); 97 migration tests pass. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Signed-off-by: Sergei Mozhaev <mozhaev.dev@gmail.com>
mozhaev-dev
force-pushed
the
feat/ai-overage-silver
branch
from
June 18, 2026 14:53
7fc849e to
e104962
Compare
cyberantonz
approved these changes
Jun 19, 2026
8 tasks
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
2 participants
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
Introduces AI overage end-to-end for Anthropic / Claude Team: a new unified Silver class
class_ai_overage(per-person-per-month spend over the monthly credit limit) surfaced all the way to a Gold AI bulletcc_overage.Designed homogeneous so OpenAI/ChatGPT can join later by emitting the same contract (deferred by PM for now).
Layers
class_ai_overage(19-col positional contract) + first contributorclaude_team__ai_overagefrom the existingclaude_team_overage_spendstream (now unblocked: the proxy sessionKey has Owner/billing:view, so/overage_spend_limitsreturns all 149 seats).currency/out_of_credits/used_credits_basis/limit_type; promoted to RMT.used_credits/monthly_credit_limitare already cents (USD) → mapped straight to*_cents, no ×100 (unlikecost_cents).overage_cents = max(0, used − limit), honest-NULL when no limit.unique_keypreserves history.ai_bullet_rowsBranch 6 →cc_overage; classifiedsumin both rollups (query_refPath A andai_person_periodPath B — the AI bullet metrics: ai_person_period rolls up Codex/cc counters as daily-avg (and PRs as fake zeros) — wrong period aggregation #1286 defect class).query_refextended (22→23 keys, honest-NULL guarded, averaged per person likecc_cost) +metric_catalogseed (label "Claude Overage", unit ¢, lower-is-better, good ≤ 0¢ / warn ≤ 5000¢). Guard test updated to cover the new key.class_ai_overageadded to the AI coverage matrix; claude-team PRD/DESIGN/FEATURE reconciled (were Bronze-only); descriptor 1.2.0 → 1.3.0.Validation
cargo test migration→ 86 passed, 0 failed.cc_overage(honest-NULL) →ai_person_periodsum-classification functionally proven (314 vs avg-bug 157) → analytics-apimigrateon MariaDB clean (probes green) → exact storedquery_refvs ClickHouse returnscc_overage = 2.317¢, 23 keys, no regression.Paired FE
Renders the
cc_overagebullet (catalog-driven): constructorfabric/insight-frontfeat/ai-overage-frontend.🤖 Generated with Claude Code
Summary by CodeRabbit
cc_overagemetric (USD, cents)cc_overageto enable it in analytics viewscc_overageaggregates as a summing counter (not average)cc_overageis mapped end-to-end and classified correctly