ocr

type BillingReport ¶ added in v0.3.0

type BillingReport struct {
	KnownCost     Currency
	Indeterminate bool
}

func (r *BillingReport) Add(other BillingReport)

type Currency ¶ added in v0.3.0

type Currency int64

Currency is stored as billionths of a US dollar. That precision represents every baked-in per-token and per-page price exactly without concurrent float arithmetic.

func GeminiBatchCost(at time.Time, inputTokens, outputTokens int64) Currency

func GeminiCost(at time.Time, inputTokens, outputTokens int64) Currency

func GeminiCostWithPrices(prices GeminiTokenPrices, inputTokens, outputTokens int64) Currency

func MistralCost(pages int) Currency

func MistralPageCost() Currency

type Failure ¶ added in v0.3.0

type Failure struct {
	Kind FailureKind
	Err  error
}

func (e *Failure) Error() string

func (e *Failure) Unwrap() error

type FailureKind ¶ added in v0.3.0

type FailureKind string

const (
	FailureUnclassified FailureKind = "unclassified"
	FailureTemporary    FailureKind = "temporary"
	FailureDocument     FailureKind = "document"
	FailureGlobal       FailureKind = "global"
)

func ClassifyFailure(err error) FailureKind

type GeminiAmbiguousOperationError ¶ added in v0.3.0

type GeminiAmbiguousOperationError struct {
	Operation string
	Err       error
}

func (e *GeminiAmbiguousOperationError) Error() string

func (e *GeminiAmbiguousOperationError) Unwrap() error

type GeminiBatchAPIError ¶ added in v0.3.0

type GeminiBatchAPIError struct {
	StatusCode int
	Body       []byte
}

func (e *GeminiBatchAPIError) Error() string

type GeminiBatchTransport ¶ added in v0.3.0

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

GeminiBatchTransport keeps the discounted Batch API's Files and batch-job lifecycle explicit instead of hiding non-idempotent calls behind an SDK.

func NewGeminiBatchTransport(apiKey string) *GeminiBatchTransport

func (c *GeminiBatchTransport) CancelBatch(ctx context.Context, name string) error

func (c *GeminiBatchTransport) CreateBatch(
	ctx context.Context,
	model, displayName, inputFileName string,
) (GeminiRemoteBatch, error)

func (c *GeminiBatchTransport) DeleteBatch(ctx context.Context, name string) error

func (c *GeminiBatchTransport) DeleteFile(ctx context.Context, name string) error

func (c *GeminiBatchTransport) DownloadFile(ctx context.Context, name string) (io.ReadCloser, error)

func (c *GeminiBatchTransport) GetBatch(ctx context.Context, name string) (GeminiRemoteBatch, error)

func (c *GeminiBatchTransport) ListBatches(ctx context.Context) ([]GeminiRemoteBatch, error)

func (c *GeminiBatchTransport) ListFiles(ctx context.Context) ([]GeminiRemoteFile, error)

func (c *GeminiBatchTransport) UploadJSONL(
	ctx context.Context,
	displayName string,
	source io.ReadSeeker,
	size int64,
) (GeminiRemoteFile, error)

type GeminiClient ¶ added in v0.3.0

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

func NewGeminiClient(apiKey string, runAt time.Time) *GeminiClient

func NewGeminiClientFromEnv(runAt time.Time) (*GeminiClient, error)

func (c *GeminiClient) OCRFile(ctx context.Context, filePath string, fileType string) ([]PageResult, BillingReport, error)

func (c *GeminiClient) OCRRange(
	ctx context.Context,
	filePath, fileType string,
	start, end int,
) ([]PageResult, BillingReport, error)

func (c *GeminiClient) OCRRangeResult(
	ctx context.Context,
	filePath, fileType string,
	start, end int,
) (RangeResult, error)

func (c *GeminiClient) PrepareFileRequests(ctx context.Context, filePath, fileType string) ([]GeminiPreparedRequest, error)

func (c *GeminiClient) PrepareRangeRequest(
	ctx context.Context,
	filePath, fileType string,
	start, end int,
) (GeminiPreparedRequest, error)

func (c *GeminiClient) WalkFileRequests(
	ctx context.Context,
	filePath, fileType string,
	yield func(GeminiPreparedRequest) error,
	reject func(*GeminiRangeSizeError) error,
) error

type GeminiDecodedResult ¶ added in v0.3.0

type GeminiDecodedResult struct {
	Pages        []PageResult
	InputTokens  *int64
	OutputTokens *int64
	Billing      BillingReport
}

func DecodeGeminiBatchResult(
	body []byte,
	expectedPages int,
	fallbackModel string,
	prices GeminiTokenPrices,
) (GeminiDecodedResult, error)

type GeminiPlanningError ¶ added in v0.3.0

type GeminiPlanningError struct {
	PageStart int
	PageEnd   int
	Cause     error
}

func (e *GeminiPlanningError) Error() string

func (e *GeminiPlanningError) Unwrap() error

type GeminiPreparedRequest ¶ added in v0.3.0

type GeminiPreparedRequest struct {
	PageStart int
	PageEnd   int
	Body      []byte
}

GeminiClient uses the REST API directly so the OCR provider has no SDK dependency and its request and billing behavior remain explicit.

type GeminiRangeSizeError ¶ added in v0.3.0

type GeminiRangeSizeError struct {
	PageStart int
	PageEnd   int
	Cause     error
}

func (e *GeminiRangeSizeError) Error() string

func (e *GeminiRangeSizeError) Unwrap() error

type GeminiRemoteBatch ¶ added in v0.3.0

type GeminiRemoteBatch struct {
	Name           string
	DisplayName    string
	Model          string
	InputFileName  string
	State          string
	OutputFileName string
	ErrorMessage   string
	CreateTime     string
	UpdateTime     string
}

type GeminiRemoteFile ¶ added in v0.3.0

type GeminiRemoteFile struct {
	Name        string `json:"name"`
	DisplayName string `json:"displayName"`
	URI         string `json:"uri"`
	State       string `json:"state"`
	MIMEType    string `json:"mimeType"`
	SizeBytes   string `json:"sizeBytes"`
}

type GeminiTokenPrices ¶ added in v0.3.0

type GeminiTokenPrices struct {
	Input  Currency
	Output Currency
}

func GeminiBatchPrices(at time.Time) GeminiTokenPrices

func GeminiPrices(at time.Time) GeminiTokenPrices

type MistralClient ¶

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

func NewMistralClient(apiKey string) *MistralClient

func NewMistralClientFromEnv() (*MistralClient, error)

func (c *MistralClient) OCRFile(ctx context.Context, filePath string, fileType string) ([]PageResult, BillingReport, error)

func (c *MistralClient) OCRRangeResult(
	ctx context.Context,
	filePath, fileType string,
	start, end int,
) (RangeResult, error)

func (c *MistralClient) PricePerPage() float64

type PageResult ¶

type PageResult struct {
	PageIndex int
	Markdown  string
	Model     string
}

type Provider ¶

type Provider interface {
	OCRRangeResult(
		ctx context.Context,
		filePath string,
		fileType string,
		start int,
		end int,
	) (RangeResult, error)
}

Provider processes only a requested source-page range so successful pages from earlier models or runs never need to be sent over the network again.

type RangeResult ¶ added in v0.3.0

type RangeResult struct {
	Pages   []PageResult
	Billing BillingReport
}