feat(github-copilot): dynamic GitHub Copilot model discovery via API with github-models integration#1151
Conversation
…ded Copilot model options
…precated import to new import
…ultiple components for startup screen
There was a problem hiding this comment.
Pull request overview
Replaces the hardcoded GitHub Copilot model registry with dynamic discovery via the api.githubcopilot.com/models endpoint, wired into the existing integration discovery framework as a new github-models kind. Defaults are migrated from the legacy github:copilot alias to gpt-4o across provider profiles, onboarding, and fallback code paths, with a normalization step to translate persisted legacy values. The startup banner is updated to read the user's persisted model from settings and render its public display name.
Changes:
- Delete
copilotModels.tsand remove the static Copilot model dropdown; the model picker for GitHub now relies on the dynamic discovery catalog. - Add a new
'github-models'ModelDiscoveryKindwith hardcoded Copilot/VS Code editor headers, a 5s timeout, in-place label formatting, and silent failure handling. - Replace
'github:copilot'defaults with'gpt-4o'in launch env, provider config, profile builders, onboarding messaging, and provider summaries; normalize the legacy alias at settings read time. - Gate
reasoning_effortemission in the OpenAI shim onmodelSupportsEffort(resolvedModel). - Pre-select the user's saved main-loop model in the model picker via
getDefaultMainLoopModelSetting()fallback.
Reviewed changes
Copilot reviewed 14 out of 14 changed files in this pull request and generated 10 comments.
Show a summary per file
| File | Description |
|---|---|
| src/utils/providerProfile.ts | Default model in github profile env builder switched to gpt-4o. |
| src/utils/model/modelOptions.ts | Removed Copilot static option list; GitHub provider now exposes only the Default option. |
| src/utils/model/model.ts | Default fallbacks switched to gpt-4o; legacy alias github:copilot normalized in getUserSpecifiedModelSetting. |
| src/utils/model/copilotModels.ts | File deleted (was a 383-line static registry). |
| src/services/api/providerConfig.ts | GitHub fallback for resolved model changed to gpt-4o. |
| src/services/api/openaiShim.ts | reasoning_effort only emitted when the resolved model supports it. |
| src/integrations/gateways/github.ts | Catalog switched from static to dynamic using the new github-models kind. |
| src/integrations/discoveryService.ts | Implements the github-models discovery branch (auth header, version stamping, label formatting, dedup/filter, silent fail). |
| src/integrations/descriptors.ts | Adds 'github-models' to the ModelDiscoveryKind union. |
| src/components/StartupScreen.ts | Reads getInitialSettings() for the GitHub branch and returns a display name instead of the raw model ID. |
| src/components/ProviderManager.tsx | Updates GitHub provider default model constant to gpt-4o. |
| src/commands/provider/provider.tsx | Updates provider summary fallback model to gpt-4o. |
| src/commands/onboard-github/onboard-github.tsx | Updates default and user-facing instruction string to gpt-4o. |
| src/commands/model/model.tsx | Falls back to getDefaultMainLoopModelSetting() when mainLoopModel is unset so the picker pre-selects the saved model. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| const settings = getInitialSettings() || {} | ||
| const model = modelOverride || settings.model || process.env.OPENAI_MODEL || 'gpt-4o' | ||
| const displayName = model === 'github:copilot' ? 'GPT-4o' : (getPublicModelDisplayName(model) || model) | ||
| const baseUrl = | ||
| process.env.OPENAI_BASE_URL || 'https://api.githubcopilot.com' | ||
| return { name: 'GitHub Copilot', model, baseUrl, isLocal: false } | ||
| return { name: 'GitHub Copilot', model: displayName, baseUrl, isLocal: false } |
| import { getInitialSettings } from '../utils/settings/settings.js' | ||
| import { getPublicModelDisplayName, parseUserSpecifiedModel } from '../utils/model/model.js' |
| const apiKey = getRouteDiscoveryApiKey(routeId, options) | ||
| const headers: Record<string, string> = { | ||
| Authorization: `Bearer ${apiKey}`, | ||
| 'Editor-Version': 'vscode/1.96.0', | ||
| 'Editor-Plugin-Version': 'copilot/1.250.0', | ||
| 'User-Agent': 'GitHubCopilot/1.250.0', | ||
| Accept: 'application/json', | ||
| } |
| if (base.startsWith('gpt-')) { | ||
| const rest = base.slice(4) | ||
| const named = rest | ||
| .replace(/^4o-mini/, '4o Mini') | ||
| .replace(/^4o/, '4o') | ||
| .replace(/^4-o/, '4o') | ||
| .replace(/^4\.1/, '4.1') | ||
| .replace(/^4/, '4') | ||
| .replace(/^3\.5-turbo/, '3.5 Turbo') | ||
| .replace(/^5\.5-mini/, '5.5 Mini') | ||
| .replace(/^5\.5/, '5.5') | ||
| .replace(/^5\.4-mini/, '5.4 Mini') | ||
| .replace(/^5\.4/, '5.4') | ||
| .replace(/^5\.3-codex/, '5.3 Codex') | ||
| .replace(/^5\.2-codex/, '5.2 Codex') | ||
| .replace(/^5\.2/, '5.2') | ||
| .replace(/^5\.1-codex/, '5.1 Codex') | ||
| .replace(/^5-mini/, '5 Mini') | ||
| .replace(/^5/, '5') | ||
| return `GPT-${named}` | ||
| } | ||
| if (base.startsWith('gemini-')) return base.replace('gemini-', 'Gemini ') | ||
| if (base.startsWith('grok-')) { | ||
| const v = base.replace('grok-', '') | ||
| if (v === 'code-fast-1') return 'Grok Code Fast 1' | ||
| return `Grok ${v}` | ||
| } | ||
| return rawId | ||
| } |
| const headers: Record<string, string> = { | ||
| Authorization: `Bearer ${apiKey}`, | ||
| 'Editor-Version': 'vscode/1.96.0', | ||
| 'Editor-Plugin-Version': 'copilot/1.250.0', | ||
| 'User-Agent': 'GitHubCopilot/1.250.0', | ||
| Accept: 'application/json', | ||
| } | ||
|
|
||
| function formatModelLabel(rawId: string): string { | ||
| const base = rawId.replace(/-\d{4}-\d{2}-\d{2}$/, '').replace(/-\d{4}$/, '') | ||
| if (base.startsWith('claude-')) { | ||
| const parts = base.replace('claude-', '').split('-') | ||
| if (parts.length >= 2) return `Claude ${parts[0].charAt(0).toUpperCase() + parts[0].slice(1)} ${parts.slice(1).join('.')}` | ||
| } | ||
| if (base.startsWith('gpt-')) { | ||
| const rest = base.slice(4) | ||
| const named = rest | ||
| .replace(/^4o-mini/, '4o Mini') | ||
| .replace(/^4o/, '4o') | ||
| .replace(/^4-o/, '4o') | ||
| .replace(/^4\.1/, '4.1') | ||
| .replace(/^4/, '4') | ||
| .replace(/^3\.5-turbo/, '3.5 Turbo') | ||
| .replace(/^5\.5-mini/, '5.5 Mini') | ||
| .replace(/^5\.5/, '5.5') | ||
| .replace(/^5\.4-mini/, '5.4 Mini') | ||
| .replace(/^5\.4/, '5.4') | ||
| .replace(/^5\.3-codex/, '5.3 Codex') | ||
| .replace(/^5\.2-codex/, '5.2 Codex') | ||
| .replace(/^5\.2/, '5.2') | ||
| .replace(/^5\.1-codex/, '5.1 Codex') | ||
| .replace(/^5-mini/, '5 Mini') | ||
| .replace(/^5/, '5') | ||
| return `GPT-${named}` | ||
| } | ||
| if (base.startsWith('gemini-')) return base.replace('gemini-', 'Gemini ') | ||
| if (base.startsWith('grok-')) { | ||
| const v = base.replace('grok-', '') | ||
| if (v === 'code-fast-1') return 'Grok Code Fast 1' | ||
| return `Grok ${v}` | ||
| } | ||
| return rawId | ||
| } | ||
|
|
||
| try { | ||
| const response = await fetch('https://api.githubcopilot.com/models', { | ||
| headers, | ||
| signal: AbortSignal.timeout(5000), | ||
| }) | ||
| if (!response.ok) return null | ||
|
|
||
| const body = (await response.json()) as { data?: Array<{ id?: string }> } | ||
| if (!body.data || !Array.isArray(body.data)) return null | ||
|
|
||
| const seen = new Set<string>() | ||
| const models = body.data | ||
| .map(item => { | ||
| const id = item.id?.trim() | ||
| if (!id) return null | ||
| if (id.startsWith('accounts/') || id.startsWith('oswe-') || id.includes('text-embedding')) return null | ||
| if (seen.has(id)) return null | ||
| seen.add(id) | ||
| return { id, apiName: id, label: formatModelLabel(id) } as ModelCatalogEntry | ||
| }) | ||
| .filter((m): m is ModelCatalogEntry => m !== null) | ||
|
|
||
| return models.length > 0 ? models : null | ||
| } catch { | ||
| return null | ||
| } |
| try { | ||
| const response = await fetch('https://api.githubcopilot.com/models', { | ||
| headers, | ||
| signal: AbortSignal.timeout(5000), | ||
| }) | ||
| if (!response.ok) return null | ||
|
|
||
| const body = (await response.json()) as { data?: Array<{ id?: string }> } | ||
| if (!body.data || !Array.isArray(body.data)) return null | ||
|
|
||
| const seen = new Set<string>() | ||
| const models = body.data | ||
| .map(item => { | ||
| const id = item.id?.trim() | ||
| if (!id) return null | ||
| if (id.startsWith('accounts/') || id.startsWith('oswe-') || id.includes('text-embedding')) return null | ||
| if (seen.has(id)) return null | ||
| seen.add(id) | ||
| return { id, apiName: id, label: formatModelLabel(id) } as ModelCatalogEntry | ||
| }) | ||
| .filter((m): m is ModelCatalogEntry => m !== null) | ||
|
|
||
| return models.length > 0 ? models : null | ||
| } catch { | ||
| return null | ||
| } |
| source: 'dynamic', | ||
| discovery: { kind: 'github-models', requiresAuth: true }, | ||
| discoveryCacheTtl: '1h', | ||
| discoveryRefreshMode: 'background-if-stale', | ||
| allowManualRefresh: true, |
| // Normalize legacy GitHub Copilot alias to concrete model ID | ||
| if (specifiedModel === 'github:copilot') { | ||
| specifiedModel = 'gpt-4o' | ||
| } |
| function getModelOptionsBase(fastMode = false): ModelOption[] { | ||
| if (getAPIProvider() === 'github') { | ||
| return [getDefaultOptionForUser(fastMode), ...getCopilotModelOptions()] | ||
| return [getDefaultOptionForUser(fastMode)] | ||
| } |
| Accept: 'application/json', | ||
| } | ||
|
|
||
| function formatModelLabel(rawId: string): string { |
…eout handling for test
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 16 out of 16 changed files in this pull request and generated 5 comments.
Comments suppressed due to low confidence (6)
src/integrations/discoveryService.ts:283
formatModelLabelcontains many redundant/no-op replacements that obscure intent. For example,.replace(/^4o/, '4o'),.replace(/^4/, '4'), and.replace(/^4o/, '4o')after.replace(/^4o-mini/, '4o Mini')(the latter already replaced the prefix so the^4orule cannot meaningfully change anything in the dash-mini case; for plain4oit’s a no-op). Several other rules (/^5\.5/,/^5\.4/, etc.) similarly map a string to itself. This makes the formatting table look exhaustive but the only effective transformations are the ones that change spacing (e.g.4o-mini→4o Mini,5.5-mini→5.5 Mini,5.3-codex→5.3 Codex, …). Consider removing the no-op replacements and documenting that the goal is purely to insert spaces and capitalize sub-tokens, otherwise future maintainers will think these lines do something they don’t.
if (base.startsWith('gpt-')) {
const rest = base.slice(4)
const named = rest
.replace(/^4o-mini/, '4o Mini')
.replace(/^4o/, '4o')
.replace(/^4-o/, '4o')
.replace(/^4\.1/, '4.1')
.replace(/^4/, '4')
.replace(/^3\.5-turbo/, '3.5 Turbo')
.replace(/^5\.5-mini/, '5.5 Mini')
.replace(/^5\.5/, '5.5')
.replace(/^5\.4-mini/, '5.4 Mini')
.replace(/^5\.4/, '5.4')
.replace(/^5\.3-codex/, '5.3 Codex')
.replace(/^5\.2-codex/, '5.2 Codex')
.replace(/^5\.2/, '5.2')
.replace(/^5\.1-codex/, '5.1 Codex')
.replace(/^5-mini/, '5 Mini')
.replace(/^5/, '5')
return `GPT-${named}`
}
src/integrations/discoveryService.ts:262
- Label formatting for Claude models is fragile. Stripping a trailing
-YYYY-MM-DDand then splitting on-works forclaude-opus-4-6(→Claude Opus 4.6) but produces awkward results for ids likeclaude-3-5-sonnet-20241022, which becomesClaude 3 5.sonnet, orclaude-3-7-sonnet-thinking→Claude 3 7.sonnet.thinking. Since the GitHub Copilot catalog historically includes such IDs, the picker will show confusing labels. Consider a more structured parser (recognize the family token explicitly) or fall back to a title-cased raw id when the pattern doesn’t match the expectedclaude-<family>-<major>-<minor>shape.
if (base.startsWith('claude-')) {
const parts = base.replace('claude-', '').split('-')
if (parts.length >= 2) return `Claude ${parts[0].charAt(0).toUpperCase() + parts[0].slice(1)} ${parts.slice(1).join('.')}`
}
src/integrations/discoveryService.ts:315
- The model
id/apiNameis preserved verbatim, butformatModelLabelstrips dated suffixes (e.g.gpt-4o-2024-08-06andgpt-4o-2024-11-20both label asGPT-4o). Deduplication is performed onid, not onlabel, so the picker can end up with multiple visually identical entries that resolve to different concrete models. Users will not be able to distinguish them. Either include the date suffix in the label when there is a collision, or dedupe by label and pick a canonical id.
const seen = new Set<string>()
const models = body.data
.map(item => {
const id = item.id?.trim()
if (!id) return null
if (id.startsWith('accounts/') || id.startsWith('oswe-') || id.includes('text-embedding')) return null
if (seen.has(id)) return null
seen.add(id)
return { id, apiName: id, label: formatModelLabel(id) } as ModelCatalogEntry
})
.filter((m): m is ModelCatalogEntry => m !== null)
src/integrations/discoveryService.ts:322
- The
Editor-Version,Editor-Plugin-Version, andUser-Agentheaders are hardcoded to specific VS Code/Copilot versions. The PR description acknowledges this header "may need periodic updates", but if GitHub's API tightens validation, every Claude Code user will silently lose Copilot model discovery (the function returnsnullon!response.ok, which then masks the issue). At minimum, log a warning when the response is non-OK (status, body) so this is diagnosable in production, and consider centralizing this version string so future bumps touch one place. Also, no telemetry/logging on the catch handler means any network/parse error is also silently swallowed.
const headers: Record<string, string> = {
Authorization: `Bearer ${apiKey}`,
'Editor-Version': 'vscode/1.96.0',
'Editor-Plugin-Version': 'copilot/1.250.0',
'User-Agent': 'GitHubCopilot/1.250.0',
Accept: 'application/json',
}
function formatModelLabel(rawId: string): string {
const base = rawId.replace(/-\d{4}-\d{2}-\d{2}$/, '').replace(/-\d{4}$/, '')
if (base.startsWith('claude-')) {
const parts = base.replace('claude-', '').split('-')
if (parts.length >= 2) return `Claude ${parts[0].charAt(0).toUpperCase() + parts[0].slice(1)} ${parts.slice(1).join('.')}`
}
if (base.startsWith('gpt-')) {
const rest = base.slice(4)
const named = rest
.replace(/^4o-mini/, '4o Mini')
.replace(/^4o/, '4o')
.replace(/^4-o/, '4o')
.replace(/^4\.1/, '4.1')
.replace(/^4/, '4')
.replace(/^3\.5-turbo/, '3.5 Turbo')
.replace(/^5\.5-mini/, '5.5 Mini')
.replace(/^5\.5/, '5.5')
.replace(/^5\.4-mini/, '5.4 Mini')
.replace(/^5\.4/, '5.4')
.replace(/^5\.3-codex/, '5.3 Codex')
.replace(/^5\.2-codex/, '5.2 Codex')
.replace(/^5\.2/, '5.2')
.replace(/^5\.1-codex/, '5.1 Codex')
.replace(/^5-mini/, '5 Mini')
.replace(/^5/, '5')
return `GPT-${named}`
}
if (base.startsWith('gemini-')) return base.replace('gemini-', 'Gemini ')
if (base.startsWith('grok-')) {
const v = base.replace('grok-', '')
if (v === 'code-fast-1') return 'Grok Code Fast 1'
return `Grok ${v}`
}
return rawId
}
const controller = new AbortController()
const timeoutId = setTimeout(() => controller.abort(), 5000)
try {
const response = await fetch('https://api.githubcopilot.com/models', {
headers,
signal: controller.signal,
})
if (!response.ok) return null
const body = (await response.json()) as { data?: Array<{ id?: string }> }
if (!body.data || !Array.isArray(body.data)) return null
const seen = new Set<string>()
const models = body.data
.map(item => {
const id = item.id?.trim()
if (!id) return null
if (id.startsWith('accounts/') || id.startsWith('oswe-') || id.includes('text-embedding')) return null
if (seen.has(id)) return null
seen.add(id)
return { id, apiName: id, label: formatModelLabel(id) } as ModelCatalogEntry
})
.filter((m): m is ModelCatalogEntry => m !== null)
return models.length > 0 ? models : null
} catch {
return null
} finally {
clearTimeout(timeoutId)
}
src/integrations/discoveryService.ts:291
formatModelLabelis declared inside theswitchcaseblock using a function declaration. Function declarations inside blocks have implementation-defined hoisting behavior in strict mode; while it works in modern V8/Node, prefer either lifting this helper to module scope (it's pure) or usingconst formatModelLabel = (rawId: string) => { ... }. Hoisting it also lets it be unit-tested directly, which is desirable given its complexity.
function formatModelLabel(rawId: string): string {
const base = rawId.replace(/-\d{4}-\d{2}-\d{2}$/, '').replace(/-\d{4}$/, '')
if (base.startsWith('claude-')) {
const parts = base.replace('claude-', '').split('-')
if (parts.length >= 2) return `Claude ${parts[0].charAt(0).toUpperCase() + parts[0].slice(1)} ${parts.slice(1).join('.')}`
}
if (base.startsWith('gpt-')) {
const rest = base.slice(4)
const named = rest
.replace(/^4o-mini/, '4o Mini')
.replace(/^4o/, '4o')
.replace(/^4-o/, '4o')
.replace(/^4\.1/, '4.1')
.replace(/^4/, '4')
.replace(/^3\.5-turbo/, '3.5 Turbo')
.replace(/^5\.5-mini/, '5.5 Mini')
.replace(/^5\.5/, '5.5')
.replace(/^5\.4-mini/, '5.4 Mini')
.replace(/^5\.4/, '5.4')
.replace(/^5\.3-codex/, '5.3 Codex')
.replace(/^5\.2-codex/, '5.2 Codex')
.replace(/^5\.2/, '5.2')
.replace(/^5\.1-codex/, '5.1 Codex')
.replace(/^5-mini/, '5 Mini')
.replace(/^5/, '5')
return `GPT-${named}`
}
if (base.startsWith('gemini-')) return base.replace('gemini-', 'Gemini ')
if (base.startsWith('grok-')) {
const v = base.replace('grok-', '')
if (v === 'code-fast-1') return 'Grok Code Fast 1'
return `Grok ${v}`
}
return rawId
}
src/integrations/discoveryService.ts:310
- The filter
id.startsWith('accounts/') || id.startsWith('oswe-') || id.includes('text-embedding')is an ad-hoc denylist for non-chat models. Other irrelevant model classes (e.g.*-search, image generation, moderation models) may be added to the Copilot catalog and would pass through here, appearing in the model picker for chat use. A more robust approach is to inspect thecapabilities/type/objectfields returned by the API (the Copilot/modelsresponse includes per-model capability metadata) and accept only chat-completion-capable models.
if (id.startsWith('accounts/') || id.startsWith('oswe-') || id.includes('text-embedding')) return null
| const apiKey = getRouteDiscoveryApiKey(routeId, options) | ||
| const headers: Record<string, string> = { | ||
| Authorization: `Bearer ${apiKey}`, | ||
| 'Editor-Version': 'vscode/1.96.0', | ||
| 'Editor-Plugin-Version': 'copilot/1.250.0', | ||
| 'User-Agent': 'GitHubCopilot/1.250.0', | ||
| Accept: 'application/json', | ||
| } |
| // Normalize legacy GitHub Copilot alias to concrete model ID | ||
| if (specifiedModel === 'github:copilot') { | ||
| specifiedModel = 'gpt-4o' | ||
| } |
| function getModelOptionsBase(fastMode = false): ModelOption[] { | ||
| if (getAPIProvider() === 'github') { | ||
| return [getDefaultOptionForUser(fastMode), ...getCopilotModelOptions()] | ||
| return [getDefaultOptionForUser(fastMode)] |
| onDone: (result?: string, options?: { display?: CommandResultDisplay }) => void | ||
| }) { | ||
| const mainLoopModel = useAppState((s: AppState) => s.mainLoopModel) | ||
| const mainLoopModel = useAppState((s: AppState) => s.mainLoopModel) ?? getDefaultMainLoopModelSetting() |
| const settings = getInitialSettings() || {} | ||
| const model = modelOverride || settings.model || process.env.OPENAI_MODEL || 'gpt-4o' | ||
| const displayName = model === 'github:copilot' ? 'GPT-4o' : (getPublicModelDisplayName(model) || model) | ||
| const baseUrl = | ||
| process.env.OPENAI_BASE_URL || 'https://api.githubcopilot.com' | ||
| return { name: 'GitHub Copilot', model, baseUrl, isLocal: false } | ||
| return { name: 'GitHub Copilot', model: displayName, baseUrl, isLocal: false } |
jatmn
left a comment
jatmn
left a comment
There was a problem hiding this comment.
Findings
-
[P2] Keep curated GitHub models as a hybrid fallback
src/integrations/gateways/github.ts:45
The integration docs already support the catalog behavior this PR is trying to build out:docs/integrations/how-to/add-gateway.mddocumentshybridcatalogs as curated entries plus discovered models, anddocs/integrations/reference-samples.mdcalls the hosted hybrid gateway shape the recommended pattern for discovery-heavy gateways. The implementation already follows that contract too:discoverModelsForRoute()merges static entries with cached/network discovery results and falls back to static entries when discovery cannot run. I also could not find official GitHub documentation forhttps://api.githubcopilot.com/modelsor for requiring VS Code/CopilotEditor-Versionheaders as a supported integration contract. The documented GitHub Models catalog endpoint ishttps://models.github.ai/catalog/modelswith normal GitHub REST headers, while the Copilot docs publish supported model lists and plan/client availability rather than a Copilot model-discovery API. So the newgithub-modelsdiscovery kind is not justified by the public docs as written, and this PR goes well beyond the existing architecture by changing GitHub from a static catalog to a purely dynamic one with nomodelsfallback. If nonessential traffic is disabled, the user is offline, auth is unavailable, or the first refresh fails before a cache exists,discoverModelsForRoute()returns onlystaticEntries, which is now an empty array. That means/modelregresses from exposing the known Copilot models to effectively only the current/default model until network discovery succeeds. Please narrow the PR to a documented/proven gap: keep the existing curated GitHub/Copilot models as ahybridcatalog, remove the undocumented dynamic-only replacement, and only add provider-specific discovery code if there is a documented API contract or a demonstrated reason the existing descriptor/generic discovery hooks cannot cover it. -
[P2] Fix the startup banner regression and failing CI
src/components/StartupScreen.ts:98
The PR says the startup banner now shows the saved/current GitHub model display name, but CI is failing ondetectProvider — modelOverride from --model flag > modelOverride works for GitHub provider:detectProvider('gpt-4o')still returns"gpt-4o"instead of"GPT-4o". This is in the exact path changed by the PR, so the test update is currently proving the implementation does not do what it claims. Please fix the display-name resolution here and keep the test passing before merge.
Recommendation
Request changes. The shared integration infrastructure already supports dynamic discovery, caching, refresh, and hybrid fallback. The documented hybrid approach appears to provide the expected GitHub model-picker behavior with much less churn, and the new Copilot-specific discovery endpoint/headers are not backed by the official docs I could find. Please remove the redundant overreach around replacing curated model support with a dynamic-only catalog and keep the existing model metadata as the fallback. The currently failing startup test also needs to be fixed.
Blockers
Non-Blocking
Looks Good
Verdict: Changes Requested — use hybrid catalog, fix failing test. |
|
Confirmed @jatmn's findings, and flagging this against our provider/fork guidelines since it sits right on them. Three concerns, all verified in the diff:
Routing through the documented endpoint with a |
|
Summary
Screenshot
export-1778679132329.mp4
export-1778678732601.mp4
Impact
Testing
Notes