Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
34 changes: 25 additions & 9 deletions platform-api/src/api/generated.go

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

1 change: 1 addition & 0 deletions platform-api/src/internal/model/llm.go
Original file line number Diff line number Diff line change
Expand Up @@ -88,6 +88,7 @@ type CostRateLimit struct {
}

type RateLimitingResourceLimit struct {
Methods []string `json:"methods" db:"-"`
Resource string `json:"resource" db:"-"`
Limit RateLimitingLimitConfig `json:"limit" db:"-"`
}
Expand Down
15 changes: 14 additions & 1 deletion platform-api/src/internal/service/llm.go
Original file line number Diff line number Diff line change
Expand Up @@ -1202,7 +1202,12 @@ func mapResourceWiseRateLimitingAPIToModel(in *api.ResourceWiseRateLimitingConfi
}
resources := make([]model.RateLimitingResourceLimit, 0, len(in.Resources))
for _, r := range in.Resources {
methods := make([]string, 0, len(r.Methods))
for _, m := range r.Methods {
methods = append(methods, string(m))
}
resources = append(resources, model.RateLimitingResourceLimit{
Methods: methods,
Resource: r.Resource,
Limit: *mapRateLimitingLimitAPIToModel(&r.Limit),
})
Expand Down Expand Up @@ -1516,7 +1521,15 @@ func mapResourceWiseRateLimitingModelToAPI(in *model.ResourceWiseRateLimitingCon
}
resources := make([]api.RateLimitingResourceLimit, 0, len(in.Resources))
for _, r := range in.Resources {
resources = append(resources, api.RateLimitingResourceLimit{Resource: r.Resource, Limit: mapRateLimitingLimitModelToAPIValue(&r.Limit)})
methods := make([]api.RateLimitingResourceLimitMethods, 0, len(r.Methods))
if len(r.Methods) == 0 {
methods = append(methods, api.RateLimitingResourceLimitMethodsAsterisk)
} else {
for _, m := range r.Methods {
methods = append(methods, api.RateLimitingResourceLimitMethods(m))
}
}
resources = append(resources, api.RateLimitingResourceLimit{Resource: r.Resource, Methods: methods, Limit: mapRateLimitingLimitModelToAPIValue(&r.Limit)})
}
return &api.ResourceWiseRateLimitingConfig{
Default: mapRateLimitingLimitModelToAPIValue(&in.Default),
Expand Down
22 changes: 18 additions & 4 deletions platform-api/src/internal/service/llm_deployment.go
Original file line number Diff line number Diff line change
Expand Up @@ -709,10 +709,17 @@ func generateLLMProviderDeploymentYAML(provider *model.LLMProvider, templateHand
if err != nil {
return "", fmt.Errorf("invalid token reset window for resource %s: %w", r.Resource, err)
}
// TODO: the methods should be coming as input
policyMethods := make([]api.LLMPolicyPathMethods, 0, len(r.Methods))
if len(r.Methods) == 0 {
policyMethods = append(policyMethods, "*")
} else {
for _, m := range r.Methods {
policyMethods = append(policyMethods, api.LLMPolicyPathMethods(m))
}
}
addOrAppendPolicyPath(&policies, tokenBasedRateLimitPolicyName, rateLimitPolicyVersion, api.LLMPolicyPath{
Path: r.Resource,
Methods: []api.LLMPolicyPathMethods{"*"},
Methods: policyMethods,
Params: map[string]interface{}{
"totalTokenLimits": []map[string]interface{}{
{
Expand All @@ -729,10 +736,17 @@ func generateLLMProviderDeploymentYAML(provider *model.LLMProvider, templateHand
if err != nil {
return "", fmt.Errorf("invalid request reset window for resource %s: %w", r.Resource, err)
}
// TODO: the methods should be coming as input
policyMethods := make([]api.LLMPolicyPathMethods, 0, len(r.Methods))
if len(r.Methods) == 0 {
policyMethods = append(policyMethods, "*")
} else {
for _, m := range r.Methods {
policyMethods = append(policyMethods, api.LLMPolicyPathMethods(m))
}
}
addOrAppendPolicyPath(&policies, advancedRateLimitPolicyName, rateLimitPolicyVersion, api.LLMPolicyPath{
Path: r.Resource,
Methods: []api.LLMPolicyPathMethods{"*"},
Methods: policyMethods,
Params: map[string]interface{}{
"quotas": []map[string]interface{}{
{
Expand Down
6 changes: 6 additions & 0 deletions platform-api/src/resources/openapi.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -5293,9 +5293,15 @@ components:
RateLimitingResourceLimit:
type: object
required:
- methods
- resource
- limit
properties:
methods:
type: array
items:
type: string
enum: [ GET, POST, PUT, DELETE, PATCH, OPTIONS, HEAD, "*" ]
resource:
type: string
description: Explicit resource path to apply the limit to.
Expand Down
Loading