cli

Documentation

Index

• func Execute() int
• func NewRootCmd() *cobra.Command
• type ExitCodeError
  • func (e *ExitCodeError) Error() string
• type LLMCallCounts
  • func (c LLMCallCounts) Total() int64
• type ModelCapabilities
  • func ResolveCapabilities(provider, model string) (ModelCapabilities, bool)
• type Project
  • func ListAnalyzedProjects(cacheDir string) ([]Project, error)

Functions

func Execute

func Execute() int

func NewRootCmd

func NewRootCmd() *cobra.Command

Types

type ExitCodeError

type ExitCodeError struct {
	Code int
}

ExitCodeError signals to Execute that the CLI should exit with the given non-zero code without printing any additional error text. The subcommand that returns this error is responsible for having already written any user-facing output.

func (*ExitCodeError) Error

func (e *ExitCodeError) Error() string

type LLMCallCounts

type LLMCallCounts struct {
	Small, Typical, Large int64
}

LLMCallCounts is a snapshot of how many LLM calls have flowed through each tier of an llmTiering. Reported at the end of an analyze run when --verbose is set.

func (LLMCallCounts) Total

func (c LLMCallCounts) Total() int64

Total returns the sum across all tiers.

type ModelCapabilities

type ModelCapabilities struct {
	Provider string
	Model    string
	ToolUse  bool
	Vision   bool
	// MaxCompletionTokens is the per-model output cap. Zero means "use the
	// BifrostClient default". Set explicitly only for models whose API rejects
	// the default 32k request (e.g., Groq's llama-4-scout caps at 8192).
	MaxCompletionTokens int
	// MaxInputTokens is the per-model input cap including system prompt,
	// tool definitions, and accumulated chat history. Zero disables the
	// budget gate (used for self-hosted ollama/lmstudio "*" rows where the
	// user picks the model). The decorator gates sends at 0.9 × this value;
	// see .plans/2026-05-07-token-budget-design.md.
	MaxInputTokens int
}

ModelCapabilities describes which optional LLM features a (provider, model) pair supports. Looked up via ResolveCapabilities at tier construction time and travels with the LLMClient so analysis branches without naming providers.

func ResolveCapabilities

func ResolveCapabilities(provider, model string) (ModelCapabilities, bool)

ResolveCapabilities returns the capability flags for (provider, model). The bool is true when the provider is recognized; for known providers with an unknown model, it returns a zero-value ModelCapabilities and true so the caller can run with no optional features rather than failing the run.

type Project

type Project struct {
	Name    string
	SiteDir string
}

Project is one analyzed repo whose Hugo site lives under cacheDir/<Name>/site.

func ListAnalyzedProjects

func ListAnalyzedProjects(cacheDir string) ([]Project, error)

ListAnalyzedProjects returns every immediate subdirectory of cacheDir that contains a `site` subdirectory. A non-existent cacheDir is treated as "no analyzed projects" (not an error) so the caller can produce one helpful message.