Values must respect the configuration's conditional logic:
# Valid: haiku model allows up to 2048 tokens
config = llm_config(values={
"model_name": "haiku",
"max_tokens": 2000
})
# Invalid: sonnet model only allows up to 1024 tokens
config = llm_config(values={
"model_name": "sonnet",
"max_tokens": 2000 # Will raise error
})
Numeric Bounds Validation
Numeric parameters undergo bounds validation, if specified:
# These will raise validation errors:
config = llm_config(values={
"config_dct.temperature": 1.5, # Exceeds max=1
"max_tokens": -10 # Below min=0
})
Variable Selection Methods
To ensure we pass only the required variables, we have two filtering approaches:
