benchmark

type BenchmarkRun struct {
	Mode    string      `json:"mode"`
	RunAt   time.Time   `json:"run_at"`
	Results []RunResult `json:"results"`
}

type ComparisonReport struct {
	WithCCG        *BenchmarkRun `json:"with_ccg"`
	WithoutCCG     *BenchmarkRun `json:"without_ccg,omitempty"`
	Matches        []MatchResult `json:"matches"`
	MatchesWithout []MatchResult `json:"matches_without,omitempty"`
}

type ContentBlock struct {
	Type  string          `json:"type"`
	Text  string          `json:"text,omitempty"`
	Name  string          `json:"name,omitempty"` // tool_use
	ID    string          `json:"id,omitempty"`
	Input json.RawMessage `json:"input,omitempty"`
}

type Corpus struct {
	Version string  `yaml:"version" json:"version,omitempty"`
	Queries []Query `yaml:"queries" json:"queries"`
}

type Executor interface {
	Execute(ctx context.Context, args []string, prompt, dir string) ([]byte, error)
}

type MatchResult struct {
	QueryID                 string  `json:"query_id"`
	FileHitRatio            float64 `json:"file_hit_ratio"`
	SymbolHitRatio          float64 `json:"symbol_hit_ratio"`
	StrictSymbolHitRatio    float64 `json:"strict_symbol_hit_ratio"`
	TentativeSymbolHitRatio float64 `json:"tentative_symbol_hit_ratio"`
	LLMStrictBias           float64 `json:"llm_strict_bias"`
	StrictContaminationRate float64 `json:"strict_contamination_rate"`
	ToolAwareStrictRatio    float64 `json:"tool_aware_strict_ratio"`
	ToolAwareTentativeRatio float64 `json:"tool_aware_tentative_ratio"`
	TotalToolCalls          int     `json:"total_tool_calls"`
	CcgToolCalls            int     `json:"ccg_tool_calls"`
	TotalInputTokens        int     `json:"total_input_tokens"`
}

type MessagePayload struct {
	Role    string         `json:"role"`
	Content []ContentBlock `json:"content"`
	Usage   *UsageInfo     `json:"usage,omitempty"`
}

type NaiveTokensOptions struct {
	Excludes     []string
	MaxFileBytes int64
}

type NodeExpander interface {
	GetEdgesFrom(ctx context.Context, nodeID uint) ([]model.Edge, error)
	GetNodesByIDs(ctx context.Context, ids []uint) ([]model.Node, error)
	GetAnnotation(ctx context.Context, nodeID uint) (*model.Annotation, error)
}

type Query struct {
	ID                       string   `yaml:"id"                       json:"id"`
	Description              string   `yaml:"description"              json:"description"`
	ExpectedFiles            []string `yaml:"expected_files"           json:"expected_files,omitempty"`
	ExpectedSymbols          []string `yaml:"expected_symbols"         json:"expected_symbols,omitempty"`
	ExpectedStrictSymbols    []string `yaml:"expected_strict_symbols"  json:"expected_strict_symbols,omitempty"`
	ExpectedTentativeSymbols []string `yaml:"expected_tentative_symbols" json:"expected_tentative_symbols,omitempty"`
	Difficulty               string   `yaml:"difficulty"               json:"difficulty,omitempty"`
}

type QuerySegment struct {
	QueryID  string
	Messages []SessionMessage
}

type RunResult struct {
	QueryID      string     `json:"query_id"`
	ToolCalls    []ToolCall `json:"tool_calls,omitempty"`
	FilesRead    []string   `json:"files_read,omitempty"`
	Answer       string     `json:"answer,omitempty"`
	InputTokens  int        `json:"input_tokens"`
	OutputTokens int        `json:"output_tokens"`
	ElapsedMs    int64      `json:"elapsed_ms"`
	Error        string     `json:"error,omitempty"`
}

type RunTokenBenchOptions struct {
	Naive NaiveTokensOptions
}

type Runner struct {
	// contains filtered or unexported fields
}

type RunnerConfig struct {
	Mode         string // "with-ccg" | "without-ccg"
	CWD          string // benchmark workspace directory
	MaxToolCalls int
	TimeoutSec   int
}

type SearchBackend interface {
	Query(ctx context.Context, db *gorm.DB, query string, limit int) ([]model.Node, error)
}

type SessionMessage struct {
	Type      string          `json:"type"`
	Message   *MessagePayload `json:"message,omitempty"`
	Content   json.RawMessage `json:"content,omitempty"`
	ToolUseID string          `json:"tool_use_id,omitempty"`
	IsError   bool            `json:"is_error,omitempty"`
}

type TokenBenchResult struct {
	QueryID         string  `json:"query_id"`
	NaiveTokens     int     `json:"naive_tokens"`
	GraphTokens     int     `json:"graph_tokens"`
	Ratio           float64 `json:"ratio"`
	SearchElapsedMs int64   `json:"search_elapsed_ms"`
	ResultCount     int     `json:"result_count"`
	// Recall: 정답 파일/심볼이 결과에 포함되었는지 측정
	FilesHit     int     `json:"files_hit"`
	FilesTotal   int     `json:"files_total"`
	SymbolsHit   int     `json:"symbols_hit"`
	SymbolsTotal int     `json:"symbols_total"`
	Recall       float64 `json:"recall"`
}

type ToolCall struct {
	Tool      string `json:"tool"`
	ToolUseID string `json:"tool_use_id,omitempty"`
	Input     string `json:"input,omitempty"`
	Output    string `json:"output,omitempty"`
}

type UsageInfo struct {
	InputTokens  int `json:"input_tokens"`
	OutputTokens int `json:"output_tokens"`
}