Configuration Reference¶

Route-Switch uses YAML or JSON configuration files. All services (CLI, gateway, optimizer) share the same schema.

Complete Example¶

model_providers:
  openai:
    api_key: "your-key"
    base_url: "https://api.openai.com/v1"
    models: ["gpt-4o", "gpt-4", "gpt-3.5-turbo"]
    options:
      temperature: 0.2
    rate_limit: 1200
  anthropic:
    api_key: "your-anthropic-key"
    models: ["claude-3-opus", "claude-3-sonnet"]
  mock:
    api_key: mock

mipro_v2:
  num_candidates: 5
  max_bootstrapped_demos: 3
  max_labeled_demos: 2
  num_trials: 20
  minibatch_size: 5
  minibatch_full_eval_steps: 3
  num_instruction_candidates: 4
  evaluation_strategy: "Similarity"

evaluation:
  default_strategy: "Similarity"
  threshold: 0.7
  max_retries: 3

api:
  timeout_seconds: 30
  max_retries: 3

dataset:
  base_path: "data/prompts"
  max_records: 1000

gateway:
  addr: ":8080"
  strategy: "performance_based"
  fallback_threshold: 0.8
  optimization:
    enabled: true
    interval_seconds: 1800
  combinations:
    - id: "default"
      name: "default"
      prompt: "Default onboarding prompt: {topic}"
      model: "gpt-4"
      provider: "openai"
      is_primary: true
      weight: 10
      metadata:
        optimized: false

analytics:
  driver: "duckdb"
  path: "data/analytics/metrics.duckdb"

Sections¶

model_providers¶

Define provider credentials and settings.

model_providers:
  openai:
    api_key: "your-key"
    base_url: "https://api.openai.com/v1"
    models: ["gpt-4o", "gpt-4", "gpt-3.5-turbo"]
    options:
      temperature: 0.2
    rate_limit: 1200

Field	Type	Description
`api_key`	string	Provider API key
`base_url`	string	API base URL
`models`	string[]	Available models
`options`	object	Provider-specific options (temperature, etc.)
`rate_limit`	int	Requests per minute limit

Provider Selection:

--provider gollm (default) - Uses all configured providers
--provider <name> - Restricts to specific provider
--provider mock - Local testing

mipro_v2¶

Controls the MIPROv2 optimization loop.

mipro_v2:
  num_candidates: 5
  max_bootstrapped_demos: 3
  max_labeled_demos: 2
  num_trials: 20
  minibatch_size: 5
  minibatch_full_eval_steps: 3
  num_instruction_candidates: 4
  evaluation_strategy: "Similarity"

Field	Type	Default	Description
`num_candidates`	int	5	Final prompt candidates to evaluate
`max_bootstrapped_demos`	int	3	Max examples for bootstrapping
`max_labeled_demos`	int	2	Max labeled training examples
`num_trials`	int	20	Bayesian optimization trials
`minibatch_size`	int	5	Evaluation minibatch size
`minibatch_full_eval_steps`	int	3	Full evaluation frequency
`num_instruction_candidates`	int	4	Instruction proposals to generate
`evaluation_strategy`	string	"Similarity"	Evaluation strategy

evaluation¶

Default evaluation settings.

evaluation:
  default_strategy: "Similarity"
  threshold: 0.7
  max_retries: 3

Field	Type	Default	Description
`default_strategy`	string	"Similarity"	`Similarity`, `Keyword`, or `ExactMatch`
`threshold`	float	0.7	Success threshold (0-1)
`max_retries`	int	3	Retry count for failed evaluations

api¶

HTTP client settings for provider calls.

api:
  timeout_seconds: 30
  max_retries: 3

Field	Type	Default	Description
`timeout_seconds`	int	30	Request timeout
`max_retries`	int	3	Retry count

dataset¶

Per-prompt dataset storage settings.

dataset:
  base_path: "data/prompts"
  max_records: 1000

Field	Type	Default	Description
`base_path`	string	"data/prompts"	Storage directory
`max_records`	int	1000	Max records per prompt

gateway¶

Gateway server configuration.

gateway:
  addr: ":8080"
  strategy: "performance_based"
  fallback_threshold: 0.8
  optimization:
    enabled: true
    interval_seconds: 1800
  combinations:
    - id: "default"
      name: "default"
      prompt: "Default prompt: {topic}"
      model: "gpt-4"
      provider: "openai"
      is_primary: true
      weight: 10

Field	Type	Default	Description
`addr`	string	":8080"	Listen address
`strategy`	string	"round_robin"	Load balancing strategy
`fallback_threshold`	float	-	Success rate floor (0-1)
`optimization.enabled`	bool	false	Enable background optimization
`optimization.interval_seconds`	int	3600	Optimization interval
`combinations`	array	-	Static prompt/model combinations

Strategies:

round_robin - Even rotation
weighted_round_robin - Respect weights
performance_based - Route by performance

Combination Fields:

Field	Type	Description
`id`	string	Unique identifier
`name`	string	Display name
`prompt`	string	Prompt template
`model`	string	Model name
`provider`	string	Provider name
`is_primary`	bool	Primary combination flag
`weight`	int	Load balancing weight
`metadata`	object	Additional metadata

analytics¶

Analytics storage configuration.

analytics:
  driver: "duckdb"
  path: "data/analytics/metrics.duckdb"

Field	Type	Default	Description
`driver`	string	"duckdb"	Storage driver
`path`	string	-	Database file path

Loading Configuration¶

CLI¶

./route-switch --config config.yaml

Go¶

cfgManager := config.NewSimpleConfigManager()
if err := cfgManager.Load("config.yaml"); err != nil {
    log.Fatalf("load config: %v", err)
}
config := cfgManager.GetConfig()

Validation¶

The configuration manager validates:

Counts (num_candidates, num_trials, max_records) > 0
Thresholds between 0 and 1
API timeout > 0
Dataset base path is non-empty

Additional validation occurs at service startup.

Environment Variables¶

Provider keys can be read from environment:

Variable	Description
`OPENAI_API_KEY`	OpenAI API key
`ANTHROPIC_API_KEY`	Anthropic API key
`GOOGLE_API_KEY`	Google AI API key

Environment variables override config file values.