Documentation
¶
Index ¶
Constants ¶
View Source
const ( DefaultOpenAIModel = "gpt-5.2" DefaultAnthropicModel = "claude-sonnet-4-6" DefaultAzureModel = "gpt-5.2" DefaultMistralModel = "mistral-large-latest" DefaultBedrockModel = "global.anthropic.claude-sonnet-4-6-v1:0" )
Default models for each provider. Update these when bumping model versions.
Variables ¶
This section is empty.
Functions ¶
func LLMRubricT ¶
func NumEvalsOrSkip ¶
func RecordScore ¶
func RecordScore(e *EvalT, result *EvalResult)
RecordScore records the score of an eval in a JSONL file.
Types ¶
type Eval ¶
type Eval struct {
LLM llm.LanguageModel
GraderLLM llm.LanguageModel
Prompts *llm.Prompts
// contains filtered or unexported fields
}
func NewEvalWithProvider ¶ added in v1.7.0
NewEvalWithProvider creates an Eval instance with a specific provider
type EvalLogLine ¶
type EvalResult ¶
type RubricResult ¶
type ThreadExport ¶
type ThreadExport struct {
Posts map[string]*model.Post `json:"posts"`
Channel *model.Channel `json:"channel"`
Team *model.Team `json:"team"`
Users map[string]*model.User `json:"users"`
FileInfos map[string]*model.FileInfo `json:"file_infos"`
Files map[string][]byte `json:"files"`
// Helper fields not in the JSON
RootPost *model.Post `json:"-"`
PostList *model.PostList `json:"-"`
}
ThreadExport represents the format of exported thread data
func LoadThreadFromJSON ¶
func LoadThreadFromJSON(t *EvalT, path string) *ThreadExport
LoadThreadFromJSON loads post data from a JSON file containing exported Mattermost thread data and returns it as ThreadData containing Posts, RootPost, and PostList for testing
func (*ThreadExport) LatestPost ¶
func (t *ThreadExport) LatestPost() *model.Post
func (*ThreadExport) RequestingUser ¶
func (t *ThreadExport) RequestingUser() *model.User
func (*ThreadExport) String ¶
func (t *ThreadExport) String() string
Click to show internal directories.
Click to hide internal directories.