feat(routing): add canonical model pools and admin controls #352
There are no files selected for viewing
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
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
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
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
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
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,152 @@ | ||
| package config | ||
|
|
||
| import ( | ||
| "fmt" | ||
| "sort" | ||
| "strings" | ||
| "time" | ||
| ) | ||
|
|
||
| type RoutingStrategy string | ||
|
|
||
| const ( | ||
| RoutingStrategyPriorityFailover RoutingStrategy = "priority_failover" | ||
| RoutingStrategyWeightedRoundRobin RoutingStrategy = "weighted_round_robin" | ||
| ) | ||
|
|
||
| func normalizeRoutingStrategy(strategy RoutingStrategy) RoutingStrategy { | ||
| return RoutingStrategy(strings.ToLower(strings.TrimSpace(string(strategy)))) | ||
| } | ||
|
|
||
| func ResolveRoutingStrategy(strategy RoutingStrategy) RoutingStrategy { | ||
| strategy = normalizeRoutingStrategy(strategy) | ||
| if strategy == "" { | ||
| return RoutingStrategyPriorityFailover | ||
| } | ||
| return strategy | ||
| } | ||
|
|
||
| func (s RoutingStrategy) Valid() bool { | ||
| switch normalizeRoutingStrategy(s) { | ||
| case RoutingStrategyPriorityFailover, RoutingStrategyWeightedRoundRobin: | ||
| return true | ||
| default: | ||
| return false | ||
| } | ||
| } | ||
|
|
||
| // RoutingConfig holds canonical model pool routing configuration. | ||
| type RoutingConfig struct { | ||
| Defaults RoutingDefaultsConfig `yaml:"defaults"` | ||
| ModelPools map[string]ModelPoolConfig `yaml:"model_pools"` | ||
| } | ||
|
|
||
| // RoutingDefaultsConfig holds default routing behavior for canonical pools. | ||
| type RoutingDefaultsConfig struct { | ||
| Strategy RoutingStrategy `yaml:"strategy"` | ||
| SessionAffinity bool `yaml:"session_affinity"` | ||
| SessionAffinityTTL time.Duration `yaml:"session_affinity_ttl"` | ||
| Failover RoutingFailoverConfig `yaml:"failover"` | ||
| } | ||
|
|
||
| // RoutingFailoverConfig controls fallback between candidates within the same pool. | ||
| type RoutingFailoverConfig struct { | ||
| Enabled bool `yaml:"enabled"` | ||
| MaxAttempts int `yaml:"max_attempts"` | ||
| RetryOnStatuses []int `yaml:"retry_on_statuses"` | ||
| RetryOnModelErrors bool `yaml:"retry_on_model_errors"` | ||
| } | ||
|
|
||
| // ModelPoolConfig maps one public canonical model name to concrete provider candidates. | ||
| type ModelPoolConfig struct { | ||
| Candidates []ModelPoolCandidateConfig `yaml:"candidates"` | ||
| } | ||
|
|
||
| // ModelPoolCandidateConfig defines one concrete provider/model candidate. | ||
| type ModelPoolCandidateConfig struct { | ||
| Provider string `yaml:"provider"` | ||
| Model string `yaml:"model"` | ||
| Priority int `yaml:"priority"` | ||
| Weight int `yaml:"weight"` | ||
| } | ||
|
|
||
| func loadRoutingConfig(cfg *RoutingConfig) error { | ||
| if cfg == nil { | ||
| return nil | ||
| } | ||
|
|
||
| cfg.Defaults.Strategy = ResolveRoutingStrategy(cfg.Defaults.Strategy) | ||
| if !cfg.Defaults.Strategy.Valid() { | ||
| return fmt.Errorf("routing.defaults.strategy must be one of: priority_failover, weighted_round_robin") | ||
| } | ||
| if cfg.Defaults.SessionAffinityTTL <= 0 { | ||
| cfg.Defaults.SessionAffinityTTL = 30 * time.Minute | ||
| } | ||
|
Comment on lines
+75
to
+84
There was a problem hiding this comment.
The loader restores durations, attempts, and retry statuses when they are omitted, but it does not restore boolean defaults. If a config file includes a partial Context Used: CLAUDE.md (source) |
||
| if cfg.Defaults.Failover.MaxAttempts <= 0 { | ||
| cfg.Defaults.Failover.MaxAttempts = 3 | ||
| } | ||
| if len(cfg.Defaults.Failover.RetryOnStatuses) == 0 { | ||
| cfg.Defaults.Failover.RetryOnStatuses = []int{429, 500, 502, 503, 504} | ||
| } | ||
|
Comment on lines
+88
to
+90
Contributor
There was a problem hiding this comment. Validate Invalid HTTP codes (e.g., Proposed fix if len(cfg.Defaults.Failover.RetryOnStatuses) == 0 {
cfg.Defaults.Failover.RetryOnStatuses = []int{429, 500, 502, 503, 504}
}
+ for idx, status := range cfg.Defaults.Failover.RetryOnStatuses {
+ if status < 100 || status > 599 {
+ return fmt.Errorf("routing.defaults.failover.retry_on_statuses[%d] must be a valid HTTP status code (100-599)", idx)
+ }
+ }🤖 Prompt for AI Agents |
||
|
|
||
| if len(cfg.ModelPools) == 0 { | ||
| cfg.ModelPools = nil | ||
| return nil | ||
| } | ||
|
|
||
| normalized := make(map[string]ModelPoolConfig, len(cfg.ModelPools)) | ||
| keys := make([]string, 0, len(cfg.ModelPools)) | ||
| for key := range cfg.ModelPools { | ||
| keys = append(keys, key) | ||
| } | ||
| sort.Strings(keys) | ||
|
|
||
| for _, key := range keys { | ||
| trimmedKey := strings.TrimSpace(key) | ||
| if trimmedKey == "" { | ||
| return fmt.Errorf("routing.model_pools: model key cannot be empty") | ||
| } | ||
| if _, exists := normalized[trimmedKey]; exists { | ||
| return fmt.Errorf("routing.model_pools: duplicate model key after trimming: %q", trimmedKey) | ||
| } | ||
| pool := cfg.ModelPools[key] | ||
| if len(pool.Candidates) == 0 { | ||
| return fmt.Errorf("routing.model_pools[%q]: at least one candidate is required", trimmedKey) | ||
| } | ||
|
|
||
| seenCandidates := make(map[string]struct{}, len(pool.Candidates)) | ||
| normalizedCandidates := make([]ModelPoolCandidateConfig, 0, len(pool.Candidates)) | ||
| for idx, candidate := range pool.Candidates { | ||
| candidate.Provider = strings.TrimSpace(candidate.Provider) | ||
| candidate.Model = strings.TrimSpace(candidate.Model) | ||
| if candidate.Provider == "" { | ||
| return fmt.Errorf("routing.model_pools[%q].candidates[%d].provider is required", trimmedKey, idx) | ||
| } | ||
| if candidate.Model == "" { | ||
| return fmt.Errorf("routing.model_pools[%q].candidates[%d].model is required", trimmedKey, idx) | ||
| } | ||
| candidateKey := candidate.Provider + "/" + candidate.Model | ||
| if _, exists := seenCandidates[candidateKey]; exists { | ||
| return fmt.Errorf("routing.model_pools[%q]: duplicate candidate %q", trimmedKey, candidateKey) | ||
| } | ||
| seenCandidates[candidateKey] = struct{}{} | ||
|
|
||
| switch cfg.Defaults.Strategy { | ||
| case RoutingStrategyPriorityFailover: | ||
| if candidate.Priority <= 0 { | ||
| return fmt.Errorf("routing.model_pools[%q].candidates[%d].priority must be > 0 for priority_failover", trimmedKey, idx) | ||
| } | ||
| case RoutingStrategyWeightedRoundRobin: | ||
| if candidate.Weight <= 0 { | ||
| return fmt.Errorf("routing.model_pools[%q].candidates[%d].weight must be > 0 for weighted_round_robin", trimmedKey, idx) | ||
| } | ||
| } | ||
|
|
||
| normalizedCandidates = append(normalizedCandidates, candidate) | ||
| } | ||
| normalized[trimmedKey] = ModelPoolConfig{Candidates: normalizedCandidates} | ||
| } | ||
|
|
||
| cfg.ModelPools = normalized | ||
| return nil | ||
| } | ||
Oops, something went wrong.
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.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Assert loaded failover fields in the routing YAML test.
Line 730 through Line 735 sets failover inputs, but the test only validates strategy, affinity, TTL, and candidate count. A regression in failover parsing could pass unnoticed.
✅ Suggested test assertions
if cfg.Routing.Defaults.SessionAffinityTTL != 45*time.Minute { t.Fatalf("SessionAffinityTTL = %s, want 45m", cfg.Routing.Defaults.SessionAffinityTTL) } + failover := cfg.Routing.Defaults.Failover + if !failover.Enabled { + t.Fatal("expected Failover.Enabled=true from YAML") + } + if failover.MaxAttempts != 5 { + t.Fatalf("Failover.MaxAttempts = %d, want 5", failover.MaxAttempts) + } + if !reflect.DeepEqual(failover.RetryOnStatuses, []int{429, 503}) { + t.Fatalf("Failover.RetryOnStatuses = %v, want [429 503]", failover.RetryOnStatuses) + } + if failover.RetryOnModelErrors { + t.Fatal("expected Failover.RetryOnModelErrors=false from YAML") + } pool := cfg.Routing.ModelPools["claude-sonnet-4-6"]As per coding guidelines: “Add or update tests for behavior changes to cover request translation, response normalization, error handling, default configuration, and provider-specific parameter mapping.”
🤖 Prompt for AI Agents