OptimizationCandidate interface

Package:

@azure/ai-projects

Aggregated evaluation result for a single candidate agent configuration across all tasks.

Properties

avg_score	Average composite score across all tasks.
avg_tokens	Average token usage across all tasks.
candidate_id	Server-assigned candidate identifier. Use with GET /candidates/{id} sub-endpoints.
config	The agent configuration that produced this candidate.
eval_id	Foundry evaluation identifier used to score this candidate.
eval_run_id	Foundry evaluation run identifier for this candidate's scoring run.
is_pareto_optimal	Whether this candidate is on the Pareto frontier (score vs cost).
mutations	What was mutated from the baseline (e.g., {system_prompt: 'new prompt'}).
name	Display name of the candidate (e.g., 'baseline', 'instruction-v2').
pass_rate	Fraction of tasks that met the pass threshold.
promotion	Promotion metadata. Null if the candidate has not been promoted.
task_scores	Individual task-level scores.

Property Details

avg_score

Average composite score across all tasks.

avg_score: number

Property Value

number

avg_tokens

Average token usage across all tasks.

avg_tokens: number

Property Value

number

candidate_id

Server-assigned candidate identifier. Use with GET /candidates/{id} sub-endpoints.

candidate_id?: string

Property Value

string

config

The agent configuration that produced this candidate.

config: OptimizationAgentDefinition

Property Value

OptimizationAgentDefinition

eval_id

Foundry evaluation identifier used to score this candidate.

eval_id?: string

Property Value

string

eval_run_id

Foundry evaluation run identifier for this candidate's scoring run.

eval_run_id?: string

Property Value

string

is_pareto_optimal

Whether this candidate is on the Pareto frontier (score vs cost).

is_pareto_optimal: boolean

Property Value

boolean

mutations

What was mutated from the baseline (e.g., {system_prompt: 'new prompt'}).

mutations: Record<string, any>

Property Value

Record<string, any>

name

Display name of the candidate (e.g., 'baseline', 'instruction-v2').

name: string

Property Value

string

pass_rate

Fraction of tasks that met the pass threshold.

pass_rate: number

Property Value

number

promotion

Promotion metadata. Null if the candidate has not been promoted.

promotion?: PromotionInfo

Property Value

PromotionInfo

task_scores

Individual task-level scores.

task_scores: OptimizationTaskResult[]

Property Value

OptimizationTaskResult[]