feat(pricing): add reviewable price refresh

[njbrake]

Description

Adds a reviewable Settings workflow for updating the genai-prices catalog. Accepted snapshots and pending reviews are persisted in the local database, custom prices remain higher-priority overrides, and the Models page is simplified with configurable 15, 25, or 50 row pagination.

PR Type

• New Feature
• Bug Fix
• Documentation

Relevant issues

Fixes #361

Checklist

• I understand the code I am submitting.
• I have added or updated tests that cover my change (tests/unit, tests/integration).
• I ran the Definition of Done checks locally (make lint, make typecheck, make test). Integration tests require Docker, which is unavailable in this environment.
• Documentation was updated where necessary.
• If the API contract changed, I regenerated the OpenAPI spec (uv run python scripts/generate_openapi.py).

AI Usage

• AI was used for drafting/refactoring.

AI Model/Tool used: Codex CLI

Any additional AI details you would like to share: All code and test changes were reviewed locally.

• I am an AI Agent filling out this form

Summary

• Added a Settings workflow to check for GenAI pricing catalog updates and review added/changed/removed rates before applying them.
• Persisted pending and accepted pricing snapshots in the local database, with startup loading and multi-worker convergence.
• Kept manually configured/custom model prices as higher-priority overrides so refreshes don’t overwrite them.
• Updated the Models page to simplify the UI (removing provider/context rendering and the “price an unlisted model” flow) and added configurable pagination (15/25/50 rows per page).
• Updated documentation, API examples, frontend types/hooks, and regenerated OpenAPI-generated assets; added/updated backend and integration tests.

Technical notes

• Implemented master-key-protected refresh endpoints for preview/confirm/reject, backed by a new DB table for pricing snapshots and a refresh service that computes diffs and enforces a pending/approved flow.
• Canceled background refresher task creation if log writer startup fails, and improved shutdown handling to avoid stopping an unstarted log writer.
• Definition of Done checks were not run because integration tests require Docker, which was unavailable.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 7981f217-c6a7-4037-9916-5509aa822ade

📥 Commits

Reviewing files that changed from the base of the PR and between 9af5c34 and 7d09220.

📒 Files selected for processing (1)

• src/gateway/main.py

🚧 Files skipped from review as they are similar to previous changes (1)

• src/gateway/main.py

---

Walkthrough

Changes

Pricing refresh workflow

Layer / File(s)	Summary
Snapshot persistence and refresh service alembic/versions/..., src/gateway/models/..., src/gateway/services/..., src/gateway/main.py, tests/unit/...	Adds persisted active and pending pricing snapshots, confirmation-gated refresh operations, startup hydration, worker convergence, and related tests.
Protected refresh API src/gateway/api/routes/pricing.py, tests/integration/...	Adds master-key-protected preview, confirm, and reject endpoints with response models and error handling.
Dashboard pricing review flow web/src/api/..., web/src/pages/SettingsPage.*, docs/...	Adds typed client mutations, review and accept/reject controls, UI tests, operator documentation, and Postman requests.

Models dashboard presentation

Layer / File(s)	Summary
Models table and detail refactor web/src/pages/ModelsPage.*	Removes provider and context presentation, removes unlisted-model pricing controls, and adds configurable pagination with updated tests.
Dashboard bundle regeneration src/gateway/static/dashboard/...	Updates dashboard entrypoints, compiled styles, runtime chunk mappings, and regenerated page assets.

Estimated code review effort: 4 (Complex) | ~60 minutes

Possibly related PRs

• mozilla-ai/otari#352: Both changes modify the models dashboard provider-handling and navigation flow.

Suggested reviewers: khaledosman, tbille

🚥 Pre-merge checks | ✅ 3 | ❌ 2

❌ Failed checks (2 warnings)

Check name	Status	Explanation	Resolution
Title check	⚠️ Warning	The title is related, but it uses feat(pricing): instead of the required Conventional Commit prefix format.	Change it to an allowed prefix, e.g. feat: add reviewable price refresh, and keep the main change in imperative mood.
Docstring Coverage	⚠️ Warning	Docstring coverage is 1.96% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (3 passed)

Check name	Status	Explanation
Description check	✅ Passed	The description includes the required sections and clearly covers the change, issue reference, checklist, and AI usage.
Linked Issues check	✅ Passed	The PR implements the refresh/review workflow, persists snapshots, and preserves custom-price precedence required by #361.
Out of Scope Changes check	✅ Passed	The changes stay aligned with the pricing refresh, persistence, dashboard review, and pagination objectives; no clear unrelated work stands out.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch feat/refresh-genai-prices

✨ Simplify code

• Create PR with simplified code
• Commit simplified code in branch feat/refresh-genai-prices

---

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.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 5

🤖 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 `@docs/configuration.md`:
- Around line 304-308: Update the configuration documentation near the
accepted-snapshot persistence description to state that snapshots are restored
across standalone restarts only. Preserve the existing precedence and refresh
behavior, while avoiding any implication that hybrid mode restores these
snapshots.

In `@src/gateway/models/entities.py`:
- Around line 211-217: Update the PricingSnapshot.snapshot field in the model to
declare its column using Text, matching the type created by the migration. Leave
source and updated_at unchanged so ORM metadata aligns with the existing schema.

In `@src/gateway/services/pricing_refresh_service.py`:
- Line 31: Replace the process-local _pending_snapshot state with persisted,
versioned pending-review storage so concurrent previews cannot overwrite one
another and preview/confirm requests work across workers. Update the preview
flow to issue a token or content hash, and require and validate that identifier
in the confirm and reject operations before applying or discarding the matching
snapshot; preserve the existing conflict response for missing or mismatched
versions.

In `@web/src/pages/SettingsPage.tsx`:
- Around line 609-611: Update the explanatory copy near the SettingsPage catalog
update controls to describe the fetched changes as a summary or preview, not a
complete review of every proposed price change. Account for truncated API
results and retain the existing guidance about accepting or rejecting changes,
stored genai-prices data, and custom-price precedence.
- Around line 587-636: Update PricingRefreshSection to remove the local preview
useState and use previewRefresh.data as the AlertDialog and PricingRefreshDialog
source. After successful confirmRefresh or rejectRefresh mutations, call
previewRefresh.reset() instead of clearing local state, and derive dialog
visibility from the mutation data.

🪄 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: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 20a2c4a3-d586-4272-8fe9-1b8b87cb7a03

📥 Commits

Reviewing files that changed from the base of the PR and between e114665 and 4411b96.

⛔ Files ignored due to path filters (2)

• docs/public/openapi.json is excluded by !docs/public/openapi.json
• uv.lock is excluded by !**/*.lock, !**/uv.lock

📒 Files selected for processing (21)

• alembic/versions/d5f7a9b1c3e5_add_pricing_snapshots.py
• docs/configuration.md
• pyproject.toml
• src/gateway/api/routes/pricing.py
• src/gateway/db/__init__.py
• src/gateway/main.py
• src/gateway/models/entities.py
• src/gateway/services/pricing_refresh_service.py
• src/gateway/static/dashboard/assets/index-CMLuwMLW.js
• src/gateway/static/dashboard/assets/index-DnD03gwo.js
• src/gateway/static/dashboard/assets/index-cUh--Y2D.css
• src/gateway/static/dashboard/index.html
• tests/conftest.py
• tests/integration/test_pricing_refresh_endpoint.py
• tests/unit/test_pricing_refresh_service.py
• web/src/api/hooks.ts
• web/src/api/types.ts
• web/src/pages/ModelsPage.test.tsx
• web/src/pages/ModelsPage.tsx
• web/src/pages/SettingsPage.test.tsx
• web/src/pages/SettingsPage.tsx

[Copilot]

Pull request overview

Adds a master-key-only, reviewable workflow to refresh the bundled genai-prices default pricing catalog, persisting both pending and accepted snapshots in the local database so accepted defaults survive restarts. This also streamlines the dashboard Models page table and introduces configurable page sizes.

Changes:

• Added backend pricing refresh service + persisted PricingSnapshot model/migration, plus master-key endpoints to preview/confirm/reject a refresh.
• Added Settings UI + hooks/types to review and apply/reject pricing refreshes, with associated unit/integration/web tests and docs updates.
• Simplified Models table UI and added configurable 15/25/50 row pagination (default 15).

Reviewed changes

Copilot reviewed 20 out of 24 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
web/src/pages/SettingsPage.tsx	Adds the Settings UI section + dialog to preview/accept/reject default pricing refreshes.
web/src/pages/SettingsPage.test.tsx	Adds UI tests for the pricing refresh accept/reject flows.
web/src/pages/ModelsPage.tsx	Simplifies the Models table and adds configurable page size pagination.
web/src/pages/ModelsPage.test.tsx	Updates tests to match the simplified table and new pagination behavior.
web/src/api/types.ts	Adds API types for pricing refresh preview payloads.
web/src/api/hooks.ts	Adds React Query mutations for preview/confirm/reject pricing refresh.
uv.lock	Adds the new backend dependency lock entry.
pyproject.toml	Adds the new backend dependency declaration.
src/gateway/services/pricing_refresh_service.py	Implements snapshot fetch/preview/persist/confirm/reject and startup restore.
src/gateway/models/entities.py	Adds the PricingSnapshot ORM entity.
alembic/versions/d5f7a9b1c3e5_add_pricing_snapshots.py	Adds the pricing_snapshots table migration.
src/gateway/main.py	Loads the persisted accepted snapshot during standalone startup.
src/gateway/db/init.py	Exposes PricingSnapshot in the DB module exports.
src/gateway/api/routes/pricing.py	Adds master-key endpoints for preview/confirm/reject refresh and response models.
tests/unit/test_pricing_refresh_service.py	Unit tests for pricing refresh snapshot persistence and activation behavior.
tests/integration/test_pricing_refresh_endpoint.py	Integration tests for the refresh endpoints (auth + conflict cases + protected count).
tests/conftest.py	Updates pricing reset fixture to clear refresh state between tests.
docs/configuration.md	Documents the operator workflow for refreshing bundled default pricing.
docs/public/openapi.json	Updates OpenAPI to include new pricing refresh endpoints/schemas.
docs/public/otari.postman_collection.json	Adds Postman requests for preview/confirm/reject refresh endpoints.
src/gateway/static/dashboard/index.html	Updates the committed dashboard asset hashes after the web build.

[khaledosman]

Approving to unblock. Solid change. All three endpoints are master-key gated (test confirms 401), the network fetch runs off the event loop via asyncio.to_thread, the DB commit precedes the in-memory set_custom_snapshot, persisted snapshots are re-validated on load, and the billable settlement path is untouched. httpx2 is a legitimate dep (it is genai-prices' own HTTP client). Migration chains correctly to head with a real downgrade().

Inline notes below. One product question: ModelsPage drops "Price a model not listed here" (PriceModelForm/ModelComboBox), so an operator can no longer price a brand-new or unconfigured-provider model from the UI (the /v1/pricing API still supports it). Reads as intentional per the description; just confirming it's the desired call.

Review created by Claude Code.

[khaledosman]

Multi-worker/replica divergence: this runs only in the worker that served the confirm; other workers/replicas keep the old snapshot until restart. Aliases and provider credentials have TTL refreshers for exactly this reason (documented in AGENTS.md); the accepted pricing snapshot has none, so an accepted price change applies inconsistently across a multi-worker deployment. Add a reload hook/refresher or document the restart requirement.

[khaledosman]

with_for_update() is a no-op on SQLite (the standalone default), so concurrent confirms/rejects aren't actually serialized there. Master-key-only admin path so low risk, but the lock gives false assurance on the default backend.

[coderabbitai]

Actionable comments posted: 7

🧹 Nitpick comments (3)

src/gateway/static/dashboard/assets/index-D5TTMN5D.css (1)

1-1: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win

Keep generated CSS out of source Stylelint checks.

Stylelint reports numerous errors in this generated Tailwind bundle, including intentional duplicate overrides such as transition-duration, outline-style, and line-height. Please exclude src/gateway/static/dashboard/assets/**/*.css from source linting and lint the authored CSS instead, or configure generated-asset validation separately; hand-editing this file would be overwritten on the next build.

🤖 Prompt for 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.

In `@src/gateway/static/dashboard/assets/index-D5TTMN5D.css` at line 1, Exclude
generated dashboard asset files matching
src/gateway/static/dashboard/assets/**/*.css from the source Stylelint
configuration, and ensure authored CSS remains covered by linting. Do not modify
the generated index-D5TTMN5D.css bundle or suppress its individual rules; use a
separate generated-asset validation path only if needed.

Source: Linters/SAST tools

src/gateway/static/dashboard/assets/ModelScopeControl-DfDDpsa6.js (1)

1-1: 🚀 Performance & Scalability | 🔵 Trivial | ⚡ Quick win

Use a Set for selected-model membership checks.

f.filter(s => !a.includes(s.id)) scans the selected list for every candidate. Build new Set(a) once with useMemo to keep filtering effectively linear as provider catalogs grow.

🤖 Prompt for 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.

In `@src/gateway/static/dashboard/assets/ModelScopeControl-DfDDpsa6.js` at line 1,
Update the selected-model filtering in component Q by creating a memoized Set
from a, then use set membership instead of a.includes(s.id) when building L.
Keep the existing search filtering and result limit unchanged, and ensure the
memoized Set recalculates when a changes.

tests/unit/test_pricing_refresh_service.py (1)

91-114: 🎯 Functional Correctness | 🔵 Trivial | ⚡ Quick win

Assert the propagated model price, not only provider identity.

The test computes model_key but only checks provider.id and _applied_snapshot_raw. A regression that loads the snapshot while leaving the model’s effective rate unchanged would still pass. Add an assertion for the changed model’s price.

🤖 Prompt for 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.

In `@tests/unit/test_pricing_refresh_service.py` around lines 91 - 114, Update
test_refresher_applies_a_snapshot_accepted_on_another_worker to assert the
effective price of the changed model identified by model_key after refresh. Keep
the existing provider identity and raw snapshot assertions, and ensure the
assertion verifies the expected changed rate rather than only successful
snapshot loading.

🤖 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/gateway/main.py`:
- Around line 139-142: Ensure the startup tasks created for alias_refresher,
provider_refresher, and price_refresher are covered by the same cleanup scope as
log_writer.start(). Move their creation inside the existing managed try/finally
block, or add startup-exception cleanup that cancels and awaits each task before
propagating the failure; preserve normal lifespan cleanup behavior.

In `@src/gateway/static/dashboard/assets/ActivityPage-CKgTCz2o.js`:
- Line 1: Sanitize provider diagnostics at the dashboard boundary: in
src/gateway/static/dashboard/assets/ActivityPage-CKgTCz2o.js:1-1, update Pe and
Te so entry.error_message is never rendered or copied verbatim, exposing only a
generic error message and opaque diagnostic ID; in
src/gateway/static/dashboard/assets/ProvidersPage-B1PWfyPm.js:1-1, apply the
same sanitization to connection-test and health error fields before rendering.
Preserve raw provider details only in protected server-side logs.

In `@src/gateway/static/dashboard/assets/BudgetsPage-BzUF6Aa4.js`:
- Line 1: Update the custom reset-period validation in component pe to require
Number.isInteger(o) in addition to the existing positive, finite checks. Reject
fractional day inputs by storing null and do not round or alter the entered
value; regenerate the BudgetsPage bundle from the updated source.

In `@src/gateway/static/dashboard/assets/ModelsPage-Cgm8YsX7.js`:
- Line 1: The model rows rendered by Mt are clickable but not keyboard
accessible. Update the row selection interaction around Mt and its model-cell
rendering to provide a focusable control or equivalent row focus and keyboard
handling, ensuring Enter/Space opens the same model details as clicking while
preserving access to the alias action.
- Line 1: The compiled ModelsPage asset is stale and should not be committed.
Update the source ModelsPage implementation to remove or adjust the provider and
context-window UI, then regenerate the dashboard bundle so the generated
ModelsPage asset matches the source; remove the checked-in compiled asset if
generated assets are excluded from the repository.

In `@src/gateway/static/dashboard/assets/ToolsGuardrailsPage-Dz4sDo0m.js`:
- Line 1: Update the URL test flow in component $ so its displayed l.data result
is associated with the URL submitted to l.mutate. Track the tested URL and
render success or failure details only when it matches the current trimmed
input, or disable editing while the test is pending; preserve the existing test
and save behavior.

In `@src/gateway/static/dashboard/assets/UsagePage-CFdEpDOA.js`:
- Line 1: Update the drill-down handler B to merge the currently active model
and user_id filters from the existing state with the clicked dimension before
constructing the activity URL. Preserve the clicked value as the selected
dimension while retaining any other active filter, and continue using the
existing start_date navigation behavior.

---

Nitpick comments:
In `@src/gateway/static/dashboard/assets/index-D5TTMN5D.css`:
- Line 1: Exclude generated dashboard asset files matching
src/gateway/static/dashboard/assets/**/*.css from the source Stylelint
configuration, and ensure authored CSS remains covered by linting. Do not modify
the generated index-D5TTMN5D.css bundle or suppress its individual rules; use a
separate generated-asset validation path only if needed.

In `@src/gateway/static/dashboard/assets/ModelScopeControl-DfDDpsa6.js`:
- Line 1: Update the selected-model filtering in component Q by creating a
memoized Set from a, then use set membership instead of a.includes(s.id) when
building L. Keep the existing search filtering and result limit unchanged, and
ensure the memoized Set recalculates when a changes.

In `@tests/unit/test_pricing_refresh_service.py`:
- Around line 91-114: Update
test_refresher_applies_a_snapshot_accepted_on_another_worker to assert the
effective price of the changed model identified by model_key after refresh. Keep
the existing provider identity and raw snapshot assertions, and ensure the
assertion verifies the expected changed rate rather than only successful
snapshot loading.

🪄 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: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: b5b0ebe6-7828-4854-b283-f17248129feb

📥 Commits

Reviewing files that changed from the base of the PR and between 532d0de and 9af5c34.

⛔ Files ignored due to path filters (1)

• uv.lock is excluded by !**/*.lock, !**/uv.lock

📒 Files selected for processing (26)

• docs/configuration.md
• docs/public/otari.postman_collection.json
• src/gateway/main.py
• src/gateway/models/entities.py
• src/gateway/services/pricing_refresh_service.py
• src/gateway/static/dashboard/assets/ActivityPage-CKgTCz2o.js
• src/gateway/static/dashboard/assets/AliasesPage-B5qObkia.js
• src/gateway/static/dashboard/assets/BudgetsPage-BzUF6Aa4.js
• src/gateway/static/dashboard/assets/KeysPage-BeOrjVIO.js
• src/gateway/static/dashboard/assets/ModelScopeControl-DfDDpsa6.js
• src/gateway/static/dashboard/assets/ModelsPage-Cgm8YsX7.js
• src/gateway/static/dashboard/assets/OverviewPage-C5UlyWbe.js
• src/gateway/static/dashboard/assets/ProvidersPage-B1PWfyPm.js
• src/gateway/static/dashboard/assets/SettingsPage-CopNjETz.js
• src/gateway/static/dashboard/assets/ToolsGuardrailsPage-Dz4sDo0m.js
• src/gateway/static/dashboard/assets/UsagePage-CFdEpDOA.js
• src/gateway/static/dashboard/assets/UsersPage-Bqu5E2p8.js
• src/gateway/static/dashboard/assets/index-D5TTMN5D.css
• src/gateway/static/dashboard/assets/index-DvNpC9jo.js
• src/gateway/static/dashboard/index.html
• tests/unit/test_pricing_refresh_service.py
• web/src/api/hooks.ts
• web/src/pages/ModelsPage.test.tsx
• web/src/pages/ModelsPage.tsx
• web/src/pages/SettingsPage.test.tsx
• web/src/pages/SettingsPage.tsx

🚧 Files skipped from review as they are similar to previous changes (8)

• src/gateway/static/dashboard/index.html
• web/src/pages/SettingsPage.test.tsx
• web/src/api/hooks.ts
• docs/configuration.md
• src/gateway/models/entities.py
• web/src/pages/SettingsPage.tsx
• web/src/pages/ModelsPage.test.tsx
• web/src/pages/ModelsPage.tsx

[coderabbitai]

🔒 Security & Privacy | 🟠 Major | ⚡ Quick win

Sanitize provider diagnostics before they cross the UI boundary.

Both pages expose provider-supplied error text verbatim. Return only a generic message and opaque diagnostic ID to the dashboard, while retaining raw details in protected server-side logs.

• src/gateway/static/dashboard/assets/ActivityPage-CKgTCz2o.js#L1-L1: sanitize entry.error_message before rendering or copying it.
• src/gateway/static/dashboard/assets/ProvidersPage-B1PWfyPm.js#L1-L1: sanitize connection-test and health error fields before rendering them.

📍 Affects 2 files

• src/gateway/static/dashboard/assets/ActivityPage-CKgTCz2o.js#L1-L1 (this comment)
• src/gateway/static/dashboard/assets/ProvidersPage-B1PWfyPm.js#L1-L1

🤖 Prompt for 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.

In `@src/gateway/static/dashboard/assets/ActivityPage-CKgTCz2o.js` at line 1,
Sanitize provider diagnostics at the dashboard boundary: in
src/gateway/static/dashboard/assets/ActivityPage-CKgTCz2o.js:1-1, update Pe and
Te so entry.error_message is never rendered or copied verbatim, exposing only a
generic error message and opaque diagnostic ID; in
src/gateway/static/dashboard/assets/ProvidersPage-B1PWfyPm.js:1-1, apply the
same sanitization to connection-test and health error fields before rendering.
Preserve raw provider details only in protected server-side logs.

Source: Coding guidelines

[coderabbitai]

🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win

Reject fractional reset periods instead of rounding them.

The field promises whole days, but Math.round silently converts input such as 1.5 into two days. Validate Number.isInteger(o) and reject the value, rather than changing the operator’s input.

Suggested validation

-const o=Number(d.trim()),c=Math.round(o)*j;
+const o=Number(d.trim()),c=o*j;

-r(d.trim()===""||!Number.isFinite(o)||o<=0||!Number.isFinite(c)||c<=0?null:c)
+r(d.trim()===""||!Number.isInteger(o)||o<=0||!Number.isSafeInteger(c)?null:c)

Regenerate this bundle from the source after applying the fix.

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

import{j as e}from"./tanstack-query-W9y7rsMr.js";import{r as u}from"./react-q-ooZ0ti.js";import{j as Y,u as G,k as K,l as Q,m as X,n as J,P as Z,E as R,I as ee,o as se}from"./index-DvNpC9jo.js";import{F as L}from"./Field-gj3-ox4q.js";import{T as te,a as re,b as S,L as ae,c as ne,d as ie,e as C}from"./Table-DEsIhjZo.js";import{C as B,I as le,a as de,b as oe,B as x,c as A,S as ue}from"./heroui-CewI8xK4.js";const ce=50;function me({value:s,onChange:r,users:a,label:i,description:l}){const[g,d]=u.useState(""),o=u.useMemo(()=>a.filter(t=>!t.user_id.startsWith("apikey-")).map(t=>({id:t.user_id,label:t.alias?`${t.user_id} (${t.alias})`:t.user_id})),[a]),c=u.useMemo(()=>{const t=g.trim().toLowerCase();return o.filter(m=>!s.includes(m.id)).filter(m=>!t||m.id.toLowerCase().includes(t)||m.label.toLowerCase().includes(t)).slice(0,ce)},[o,s,g]),h=t=>{s.includes(t)||r([...s,t]),d("")},b=t=>r(s.filter(m=>m!==t));return e.jsxs("div",{className:"flex flex-col gap-2",children:[e.jsxs("div",{children:[e.jsx("span",{className:"text-sm font-medium text-[var(--otari-ink)]",children:i}),l?e.jsx("p",{className:"text-xs text-[var(--otari-muted)]",children:l}):null]}),s.length>0?e.jsx("div",{className:"flex flex-wrap gap-1.5",children:s.map(t=>e.jsxs("span",{className:"inline-flex items-center gap-1 rounded-full bg-[var(--otari-brand-tint)] px-2.5 py-1 font-mono text-xs text-[var(--otari-brand-dark)]",children:[t,e.jsx("button",{type:"button","aria-label":`Remove ${t}`,onClick:()=>b(t),className:"text-[var(--otari-brand-dark)] hover:text-red-700",children:"×"})]},t))}):null,o.length===0?e.jsx("span",{className:"text-xs text-[var(--otari-muted)]",children:"No users yet. Create users first, then assign them here or from the Users page."}):e.jsxs(B.Root,{allowsEmptyCollection:!0,menuTrigger:"input",inputValue:g,onInputChange:d,selectedKey:null,onSelectionChange:t=>{t!=null&&h(String(t))},className:"flex flex-col gap-1",children:[e.jsxs(B.InputGroup,{children:[e.jsx(le,{"aria-label":"Add a user",placeholder:"Search users…",autoComplete:"off"}),e.jsx(B.Trigger,{})]}),e.jsx(B.Popover,{children:e.jsx(de,{items:c,className:"max-h-72 overflow-auto",children:t=>e.jsx(oe,{id:t.id,textValue:t.label,children:t.label})})})]})]})}const xe=new Intl.NumberFormat(void 0,{style:"currency",currency:"USD",maximumFractionDigits:2});function w(s){return xe.format(s)}const j=86400,F=3600,I=[{label:"No reset",seconds:null},{label:"Daily",seconds:j},{label:"Weekly",seconds:7*j},{label:"Monthly",seconds:30*j}];function ge(s){if(s===null)return"No reset";const r=I.find(a=>a.seconds===s);return r?r.label:s%j===0?`Every ${s/j} days`:s%F===0?`Every ${s/F} hours`:`Every ${s}s`}function $(s){if(!s)return"—";const r=new Date(s);return Number.isNaN(r.getTime())?"—":r.toLocaleString()}function he(s){const r=s.trim();if(r==="")return{value:null,valid:!0};const a=Number(r);return!Number.isFinite(a)||a<0?{value:null,valid:!1}:{value:a,valid:!0}}function pe({value:s,onChange:r}){const a=I.some(d=>d.seconds===s),[i,l]=u.useState(!a),g=s!==null&&s%j===0?String(s/j):"";return e.jsxs("div",{className:"flex flex-col gap-2",children:[e.jsx("span",{className:"text-sm font-medium text-[var(--otari-ink)]",children:"Reset period"}),e.jsxs("div",{className:"flex flex-wrap gap-2",children:[I.map(d=>e.jsx(x,{size:"sm",variant:!i&&s===d.seconds?"primary":"outline",onPress:()=>{l(!1),r(d.seconds)},children:d.label},d.label)),e.jsx(x,{size:"sm",variant:i?"primary":"outline",onPress:()=>l(!0),children:"Custom"})]}),i?e.jsx("div",{className:"flex items-end gap-2",children:e.jsx(L,{label:"Every N days",value:g,onChange:d=>{const o=Number(d.trim()),c=Math.round(o)*j;r(d.trim()===""||!Number.isFinite(o)||o<=0||!Number.isFinite(c)||c<=0?null:c)},placeholder:"14",description:"Whole days between resets."})}):null,e.jsx("span",{className:"text-xs text-[var(--otari-muted)]",children:"Spend returns to zero each period. A user’s clock starts when the budget is assigned to them."})]})}function O({title:s,submitLabel:r,initial:a,error:i,isPending:l,onSubmit:g,onClose:d,assignUsers:o}){const[c,h]=u.useState(a.name??""),[b,t]=u.useState(a.max_budget===null?"":String(a.max_budget)),[m,f]=u.useState(a.budget_duration_sec),[v,N]=u.useState([]),_=he(b),P=()=>{l||!_.valid||g({name:c.trim()||null,max_budget:_.value,budget_duration_sec:m},v)};return e.jsx(A,{children:e.jsxs(A.Content,{className:"flex flex-col gap-4 p-5",children:[e.jsx("div",{className:"text-sm font-semibold text-[var(--otari-ink)]",children:s}),e.jsx(R,{error:i}),e.jsx(L,{label:"Name (optional)",value:c,onChange:h,autoFocus:!0,placeholder:"team-free-tier",description:"A label to recognize this budget later."}),e.jsx(L,{label:"Spending limit (USD)",value:b,onChange:t,placeholder:"100.00",description:_.valid?"The most a single user on this budget may spend per period. Leave blank for no limit.":e.jsx("span",{className:"text-red-700",children:"Enter a non-negative number, or leave blank for no limit."})}),e.jsx(pe,{value:m,onChange:f}),o?e.jsx(me,{label:"Assign to users (optional)",description:"Attach this budget to existing users now. You can also manage assignments later on the Users page.",value:v,onChange:N,users:o}):null,e.jsxs("div",{className:"flex gap-2",children:[e.jsx(x,{variant:"primary",isDisabled:l||!_.valid,onPress:P,children:l?"Saving…":r}),e.jsx(x,{variant:"ghost",isDisabled:l,onPress:d,children:"Cancel"})]})]})})}function fe({budget:s}){if(s.user_count===0)return e.jsx("span",{className:"text-xs text-[var(--otari-muted)]",children:"No users assigned"});const r=s.total_spend;if(s.max_budget===null)return e.jsxs("span",{className:"text-xs text-[var(--otari-ink)]",children:[w(r)," spent",e.jsx("span",{className:"text-[var(--otari-muted)]",children:" · no limit"})]});const a=s.max_budget*s.user_count,i=a>0?Math.min(100,r/a*100):0,l=r>a;return e.jsxs("div",{className:"flex min-w-[140px] flex-col gap-1",children:[e.jsxs("div",{className:"flex items-baseline justify-between gap-2 text-xs",children:[e.jsx("span",{className:"text-[var(--otari-ink)]",children:w(r)}),e.jsxs("span",{className:"text-[var(--otari-muted)]",children:["of ",w(a)]})]}),e.jsx("div",{className:"h-1.5 w-full overflow-hidden rounded-full bg-[var(--otari-line)]",role:"progressbar","aria-valuenow":Math.round(i),"aria-valuemin":0,"aria-valuemax":100,"aria-label":"Aggregate spend against total allocation",children:e.jsx("div",{className:`h-full rounded-full ${l?"bg-red-500":"bg-[var(--otari-brand)]"}`,style:{width:`${Math.max(i,l?100:2)}%`}})})]})}function je({budgetId:s}){const r=se(s);if(r.isLoading)return e.jsxs("div",{className:"flex items-center gap-2 px-4 py-4 text-sm text-[var(--otari-muted)]",children:[e.jsx(ue,{size:"sm"})," Loading reset history…"]});if(r.error)return e.jsx("div",{className:"px-4 py-4",children:e.jsx(R,{error:r.error})});const a=r.data??[];return a.length===0?e.jsx("div",{className:"px-4 py-4 text-sm text-[var(--otari-muted)]",children:"No resets recorded yet for this budget."}):e.jsx("div",{className:"overflow-x-auto px-4 py-3",children:e.jsxs("table",{className:"w-full border-collapse text-xs",children:[e.jsx("thead",{className:"text-left text-[var(--otari-muted)]",children:e.jsxs("tr",{children:[e.jsx("th",{className:"py-1.5 pr-4 font-medium",children:"User"}),e.jsx("th",{className:"py-1.5 pr-4 font-medium",children:"Spend cleared"}),e.jsx("th",{className:"py-1.5 pr-4 font-medium",children:"Reset at"}),e.jsx("th",{className:"py-1.5 font-medium",children:"Next reset"})]})}),e.jsx("tbody",{children:a.map(i=>e.jsxs("tr",{className:"border-t border-[var(--otari-line)]",children:[e.jsx("td",{className:"py-1.5 pr-4",children:e.jsx("code",{children:i.user_id??"—"})}),e.jsx("td",{className:"py-1.5 pr-4 text-[var(--otari-ink)]",children:w(i.previous_spend)}),e.jsx("td",{className:"py-1.5 pr-4 text-[var(--otari-muted)]",children:$(i.reset_at)}),e.jsx("td",{className:"py-1.5 text-[var(--otari-muted)]",children:$(i.next_reset_at)})]},i.id))})]})})}function be({onCreate:s}){return e.jsx(A,{children:e.jsxs(A.Content,{className:"flex flex-col gap-4 p-6",children:[e.jsxs("div",{children:[e.jsx("h2",{className:"text-lg font-semibold text-[var(--otari-ink)]",children:"No budgets yet"}),e.jsx("p",{className:"mt-1 text-sm text-[var(--otari-muted)]",children:"A budget caps how much a user may spend and, optionally, resets that spend on a schedule. Create one, then assign it to users to enforce a limit."})]}),e.jsx("div",{children:e.jsx(x,{variant:"primary",onPress:s,children:"Create your first budget"})})]})})}function ve({label:s,isPending:r,onConfirm:a}){const[i,l]=u.useState(!1);return i?e.jsxs("div",{className:"flex flex-col items-end gap-1.5 rounded-lg border border-amber-200 bg-amber-50 p-2 text-right",children:[e.jsxs("span",{className:"max-w-xs text-xs text-amber-800",children:["Delete ",e.jsx("strong",{children:s}),"? Users keep their spend but lose this limit. Cannot be undone."]}),e.jsxs("span",{className:"inline-flex gap-1",children:[e.jsx(x,{size:"sm",variant:"danger",isDisabled:r,onPress:a,children:"Delete permanently"}),e.jsx(x,{size:"sm",variant:"ghost",isDisabled:r,onPress:()=>l(!1),children:"Cancel"})]})]}):e.jsx(x,{size:"sm",variant:"danger-soft",onPress:()=>l(!0),children:"Delete"})}function V(s){return s.split("-")[0]}function H(s){return s.name??V(s.budget_id)}function Pe(){const s=Y(),r=G(),a=K(),i=Q(),l=X(),g=J(),[d,o]=u.useState(!1),[c,h]=u.useState(null),[b,t]=u.useState(null),[m,f]=u.useState(null),[v,N]=u.useState(null),[_,P]=u.useState(!1),U=s.data??[],T=s.isLoading,y=U.find(n=>n.budget_id===c)??null,M=!T&&U.length===0&&!d,z=async(n,p)=>{P(!0),f(null);const D=await Promise.allSettled(p.map(k=>g.mutateAsync({id:k,body:{budget_id:n}})));P(!1);const E=D.flatMap((k,q)=>k.status==="rejected"?[p[q]]:[]);if(E.length>0){N({budgetId:n,userIds:E}),f(new Error(`Budget created, but could not assign it to: ${E.join(", ")}. Retry to try again.`));return}N(null),o(!1)},W=(n,p)=>{if(v){z(v.budgetId,v.userIds);return}f(null),a.mutate(n,{onSuccess:async D=>{if(p.length>0){await z(D.budget_id,p);return}o(!1)}})};return e.jsxs("div",{className:"flex flex-col gap-6",children:[e.jsx(Z,{title:"Budgets",description:"Define spending limits and reset schedules. Assign a budget to users to enforce it.",action:d||M?null:e.jsx(x,{variant:"primary",onPress:()=>{h(null),f(null),N(null),o(!0)},children:"Create budget"})}),e.jsx(R,{error:s.error??a.error??i.error??l.error??g.error}),e.jsx(ee,{children:"Assign a budget to users when you create it, or later from the Users page. Each row’s usage aggregates the spend of the users currently on that budget."}),M?e.jsx(be,{onCreate:()=>{h(null),f(null),N(null),o(!0)}}):null,d?e.jsx(O,{title:"Create budget",submitLabel:v?"Retry assignments":"Create budget",initial:{name:null,max_budget:null,budget_duration_sec:null},error:a.error??m,isPending:a.isPending||_,assignUsers:r.data??[],onSubmit:W,onClose:()=>{f(null),N(null),o(!1)}}):null,y?e.jsx(O,{title:`Edit budget ${H(y)}`,submitLabel:"Save changes",initial:{name:y.name,max_budget:y.max_budget,budget_duration_sec:y.budget_duration_sec},error:i.error,isPending:i.isPending,onSubmit:n=>i.mutate({id:y.budget_id,body:n},{onSuccess:()=>h(null)}),onClose:()=>h(null)},y.budget_id):null,e.jsxs(te,{children:[e.jsx(re,{children:e.jsxs("tr",{children:[e.jsx(S,{children:"Budget"}),e.jsx(S,{children:"Limit (per user)"}),e.jsx(S,{children:"Reset"}),e.jsx(S,{children:"Users"}),e.jsx(S,{children:"Usage"}),e.jsx(S,{className:"text-right",children:"Actions"})]})}),e.jsx("tbody",{children:T?e.jsx(ae,{colSpan:6}):U.length===0?e.jsx(ne,{colSpan:6,children:"No budgets yet. Create one to cap spending."}):U.map(n=>e.jsxs(u.Fragment,{children:[e.jsxs(ie,{selected:c===n.budget_id,onClick:()=>{o(!1),h(n.budget_id)},children:[e.jsx(C,{className:"font-medium text-[var(--otari-ink)]",children:e.jsxs("div",{className:"flex flex-col gap-0.5",children:[e.jsx("span",{children:n.name??e.jsx("span",{className:"text-[var(--otari-muted)]",children:"(unnamed)"})}),e.jsx("code",{className:"text-[11px] text-[var(--otari-muted)]",title:n.budget_id,children:V(n.budget_id)})]})}),e.jsx(C,{children:n.max_budget===null?e.jsx("span",{className:"text-[var(--otari-muted)]",children:"Unlimited"}):w(n.max_budget)}),e.jsx(C,{className:"text-[var(--otari-muted)]",children:ge(n.budget_duration_sec)}),e.jsx(C,{className:"text-[var(--otari-muted)]",children:n.user_count}),e.jsx(C,{children:e.jsx(fe,{budget:n})}),e.jsx(C,{children:e.jsxs("div",{className:"flex items-center justify-end gap-1.5",onClick:p=>p.stopPropagation(),children:[e.jsx(x,{size:"sm",variant:"ghost",onPress:()=>t(p=>p===n.budget_id?null:n.budget_id),children:b===n.budget_id?"Hide history":"History"}),e.jsx(x,{size:"sm",variant:"ghost",onPress:()=>{o(!1),h(n.budget_id)},children:"Edit"}),e.jsx(ve,{label:H(n),isPending:l.isPending,onConfirm:()=>l.mutate(n.budget_id)})]})})]}),b===n.budget_id?e.jsx("tr",{className:"border-b border-[var(--otari-line)] bg-[var(--otari-bg)]",children:e.jsx("td",{colSpan:6,children:e.jsx(je,{budgetId:n.budget_id})})}):null]},n.budget_id))})]})]})}export{Pe as BudgetsPage};

function pe({value:s,onChange:r}){const a=I.some(d=>d.seconds===s),[i,l]=u.useState(!a),g=s!==null&&s%j===0?String(s/j):"";return e.jsxs("div",{className:"flex flex-col gap-2",children:[e.jsx("span",{className:"text-sm font-medium text-[var(--otari-ink)]",children:"Reset period"}),e.jsxs("div",{className:"flex flex-wrap gap-2",children:[I.map(d=>e.jsx(x,{size:"sm",variant:!i&&s===d.seconds?"primary":"outline",onPress:()=>{l(!1),r(d.seconds)},children:d.label},d.label)),e.jsx(x,{size:"sm",variant:i?"primary":"outline",onPress:()=>l(!0),children:"Custom"})]}),i?e.jsx("div",{className:"flex items-end gap-2",children:e.jsx(L,{label:"Every N days",value:g,onChange:d=>{const o=Number(d.trim()),c=o*j;r(d.trim()===""||!Number.isInteger(o)||o<=0||!Number.isSafeInteger(c)?null:c)},placeholder:"14",description:"Whole days between resets."})}):null,e.jsx("span",{className:"text-xs text-[var(--otari-muted)]",children:"Spend returns to zero each period. A user’s clock starts when the budget is assigned to them."})]})}

🤖 Prompt for 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.

In `@src/gateway/static/dashboard/assets/BudgetsPage-BzUF6Aa4.js` at line 1,
Update the custom reset-period validation in component pe to require
Number.isInteger(o) in addition to the existing positive, finite checks. Reject
fractional day inputs by storing null and do not round or alter the entered
value; regenerate the BudgetsPage bundle from the updated source.

[coderabbitai]

🎯 Functional Correctness | 🟠 Major | ⚡ Quick win

Make model-row selection keyboard accessible.

Mt attaches onClick to <tr> without a focus target or keyboard handler, so keyboard users cannot open model details or reach the alias action through row selection. Use a focusable control in the model cell, or add appropriate focus/keyboard behavior to the row.

🤖 Prompt for 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.

In `@src/gateway/static/dashboard/assets/ModelsPage-Cgm8YsX7.js` at line 1, The
model rows rendered by Mt are clickable but not keyboard accessible. Update the
row selection interaction around Mt and its model-cell rendering to provide a
focusable control or equivalent row focus and keyboard handling, ensuring
Enter/Space opens the same model details as clicking while preserving access to
the alias action.

---

🎯 Functional Correctness | 🟠 Major | ⚡ Quick win

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
set -euo pipefail

echo "== repo files matching ModelsPage =="
git ls-files | rg '(^|/)ModelsPage\.(tsx|jsx|ts|js)$|src/gateway/static/dashboard/assets/ModelsPage-.*\.js$' || true

echo
echo "== source if present =="
if [ -f web/src/pages/ModelsPage.tsx ]; then
  wc -l web/src/pages/ModelsPage.tsx
  rg -n "Provider|provider|Context window|context|context_window|inputPrice|outputPrice|contextWindow" web/src/pages/ModelsPage.tsx || true
else
  echo "web/src/pages/ModelsPage.tsx: not present"
fi

echo
echo "== target bundle relevant snippets =="
python3 - <<'PY'
from pathlib import Path
p = Path("src/gateway/static/dashboard/assets/ModelsPage-Cgm8YsX7.js")
text = p.read_text()
for needle in ['"Provider"', "'Provider'", 'contextWindow', '"Context window"', "'Context window'", "inputPrice", "outputPrice"]:
    print(f"\n-- {needle!r} occurrences --")
    for i,line in enumerate(text.splitlines(),1):
        if needle in line:
            start=max(1,i-3); end=min(200, i+2000)  # compact for minified
            lines=line.strip().split("import{j as e}from")[0]+line
            print(f"{i}: {line[:600]}")
PY

Repository: mozilla-ai/otari

Length of output: 8783

---

🏁 Script executed:

#!/bin/bash
set -euo pipefail

echo "== locate and inspect model-related UI definitions =="
rg -n "ModelsPage|mt=|Mt=|jt=|contextWindow|provider|Provider|context_window|input_price|output_price|context window" web/src src/gateway/static/dashboard/assets || true

Repository: mozilla-ai/otari

Length of output: 50373

---

Regenerate the dashboard bundle for the ModelsPage changes.

web/src/pages/ModelsPage.tsx still contains provider and context-window UI, and src/gateway/static/dashboard/assets/ModelsPage-Cgm8YsX7.js is a compiled asset that shouldn’t live in the repo. Rebuild after removing/adjusting those UI fields so the generated bundle doesn’t diverge from the source.

🤖 Prompt for 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.

In `@src/gateway/static/dashboard/assets/ModelsPage-Cgm8YsX7.js` at line 1, The
compiled ModelsPage asset is stale and should not be committed. Update the
source ModelsPage implementation to remove or adjust the provider and
context-window UI, then regenerate the dashboard bundle so the generated
ModelsPage asset matches the source; remove the checked-in compiled asset if
generated assets are excluded from the repository.

[coderabbitai]

🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win

Prevent stale test results from being shown for a new URL.

The field remains editable while l.mutate(...) is pending, but the returned l.data is displayed without the URL it belongs to. If the operator edits the value before the request completes, the old result can be shown for the new URL. Track the tested URL and only render matching results, or disable the input while testing.

🤖 Prompt for 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.

In `@src/gateway/static/dashboard/assets/ToolsGuardrailsPage-Dz4sDo0m.js` at line
1, Update the URL test flow in component $ so its displayed l.data result is
associated with the URL submitted to l.mutate. Track the tested URL and render
success or failure details only when it matches the current trimmed input, or
disable editing while the test is pending; preserve the existing test and save
behavior.

[coderabbitai]

🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win

Preserve the other active filter during drill-down.

B drops the current model or user_id filter and only adds the clicked dimension. For example, drilling into a model while filtering by user loses the user constraint and shows unrelated activity. Merge the existing filters with the clicked value before navigating.

🤖 Prompt for 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.

In `@src/gateway/static/dashboard/assets/UsagePage-CFdEpDOA.js` at line 1, Update
the drill-down handler B to merge the currently active model and user_id filters
from the existing state with the clicked dimension before constructing the
activity URL. Preserve the clicked value as the selected dimension while
retaining any other active filter, and continue using the existing start_date
navigation behavior.