parser

Documentation

Overview

Package parser provides functionality for parsing prescription images using different AI backends. It supports multiple AI services (OpenAI and Gemini) for processing prescription forms and extracting structured data from them, with vector embedding generation for similarity search.

Index

• Constants
• Variables
• func GenerateSchema[T any]() map[string]interface{}
• type GeminiParser
  • func NewGeminiParser(cfg config.Config, ds datastore.Datastore, logger *zap.Logger) (*GeminiParser, error)
  • func (p *GeminiParser) GetEmbedding(ctx context.Context, prescription models.Prescription) ([]float32, error)
  • func (p *GeminiParser) ParseImage(ctx context.Context, fileName string, file io.Reader) (string, error)
  • func (p *GeminiParser) ScoreResult(ctx context.Context, expectedJSON, outputJSON string) (models.ParserResultScore, error)
  • func (p *GeminiParser) UploadImage(ctx context.Context, fileName string, file io.Reader) (string, error)
• type OpenAIParser
  • func NewOpenAIParser(cfg config.Config, ds datastore.Datastore, logger *zap.Logger) (*OpenAIParser, error)
  • func (p *OpenAIParser) GetEmbedding(ctx context.Context, prescription models.Prescription) ([]float32, error)
  • func (p *OpenAIParser) ParseImage(ctx context.Context, fileName string, file io.Reader) (string, error)
  • func (p *OpenAIParser) ScoreResult(ctx context.Context, expectedJSON, outputJSON string) (models.ParserResultScore, error)
  • func (p *OpenAIParser) UploadImage(ctx context.Context, fileName string, file io.Reader) (string, error)
• type Parser
  • func NewParser(cfg config.Config, ds datastore.Datastore, logger *zap.Logger) (Parser, error)

Constants

const (
	// JobTypeParsePrescription represents a job for parsing a prescription image.
	JobTypeParsePrescription = "parse_prescription"
)

Job type constants used to identify different types of asynchronous jobs in the job tracking system.

Variables

var PrescriptionResponseSchema = GenerateSchema[models.Prescription]()

PrescriptionResponseSchema is a JSON schema that defines the structure for prescription data extraction when using the OpenAI API. It's generated from the models.Prescription struct using reflection.

var ResultScoreSchema = GenerateSchema[models.ParserResultScore]()

ResultScoreSchema is a JSON schema that defines the structure for result score data extraction when using the OpenAI API. It's generated from the models.ParserResultScore struct using reflection.

Functions

func GenerateSchema

func GenerateSchema[T any]() map[string]interface{}

GenerateSchema creates a JSON schema from a Go type using reflection. It configures the schema generator to comply with the subset of JSON schema that is supported by the OpenAI API.

The generic type T is the struct type from which to generate the schema. Returns a map representation of the JSON schema that can be provided to the API.

Types

type GeminiParser

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

GeminiParser implements the Parser interface using Google Gemini services. It leverages Gemini's multimodal capabilities to process prescription images and extract structured data from them.

func NewGeminiParser

func NewGeminiParser(cfg config.Config, ds datastore.Datastore, logger *zap.Logger) (*GeminiParser, error)

NewGeminiParser creates a new Gemini-based parser. It initializes a client for the Gemini API with the provided API key and returns a parser instance ready for processing prescription images.

func (*GeminiParser) GetEmbedding

func (p *GeminiParser) GetEmbedding(ctx context.Context, prescription models.Prescription) ([]float32, error)

GetEmbedding generates embeddings for a prescription using Gemini embeddings API. It converts the prescription to JSON and sends it to the Gemini API to generate a vector representation for similarity search.

func (*GeminiParser) ParseImage

func (p *GeminiParser) ParseImage(ctx context.Context, fileName string, file io.Reader) (string, error)

ParseImage handles parsing a prescription image using Gemini multimodal API. It creates an asynchronous job to process the image and returns the job ID. The actual processing is done in a separate goroutine.

func (*GeminiParser) ScoreResult

func (p *GeminiParser) ScoreResult(ctx context.Context, expectedJSON, outputJSON string) (models.ParserResultScore, error)

ScoreResult scores the result of a parser against a validated expected JSON.

func (*GeminiParser) UploadImage

func (p *GeminiParser) UploadImage(ctx context.Context, fileName string, file io.Reader) (string, error)

UploadImage uploads an image using the Gemini Files API. It validates the file type, uploads the file to Gemini, and returns a URI that can be used to reference the image in subsequent API calls.

type OpenAIParser

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

OpenAIParser implements the Parser interface using OpenAI services. It uses OpenAI's vision and embedding capabilities to process prescription images and extract structured data from them.

func NewOpenAIParser

func NewOpenAIParser(cfg config.Config, ds datastore.Datastore, logger *zap.Logger) (*OpenAIParser, error)

NewOpenAIParser creates a new OpenAI-based parser. It initializes a connection to the OpenAI API with the provided API key and returns a parser instance ready for processing prescription images.

func (*OpenAIParser) GetEmbedding

func (p *OpenAIParser) GetEmbedding(ctx context.Context, prescription models.Prescription) ([]float32, error)

GetEmbedding generates embeddings for a prescription using OpenAI. It converts the prescription to JSON and sends it to the OpenAI API to generate a vector representation for similarity search.

func (*OpenAIParser) ParseImage

func (p *OpenAIParser) ParseImage(ctx context.Context, fileName string, file io.Reader) (string, error)

ParseImage handles parsing a prescription image using OpenAI vision API. It creates an asynchronous job to process the image and returns the job ID. The actual processing is done in a separate goroutine.

func (*OpenAIParser) ScoreResult

func (p *OpenAIParser) ScoreResult(ctx context.Context, expectedJSON, outputJSON string) (models.ParserResultScore, error)

ScoreResult scores the result of a parser against a validated expected JSON.

func (*OpenAIParser) UploadImage

func (p *OpenAIParser) UploadImage(ctx context.Context, fileName string, file io.Reader) (string, error)

UploadImage uploads an image to the OpenAI API. It validates the file type, uploads the file to OpenAI, and returns an ID that can be used to reference the image in subsequent API calls.

type Parser

type Parser interface {
	// ParseImage processes a prescription image asynchronously and returns a job ID for tracking parsing progress.
	// It takes a filename and file reader, initiates an asynchronous job, and returns the job ID.
	ParseImage(ctx context.Context, fileName string, file io.Reader) (string, error)

	// GetEmbedding generates an embedding vector for a prescription.
	// This vector representation can be used for similarity searches and document clustering.
	GetEmbedding(ctx context.Context, prescription models.Prescription) ([]float32, error)

	// UploadImage uploads an image to persistent storage and returns its ID.
	// The image can then be referenced in subsequent API calls.
	UploadImage(ctx context.Context, fileName string, file io.Reader) (string, error)

	// ScoreResult scores the result of a parser against a validated expected JSON.
	ScoreResult(ctx context.Context, expectedJSON, outputJSON string) (models.ParserResultScore, error)
}

Parser defines the interface for prescription parsing services

func NewParser

func NewParser(cfg config.Config, ds datastore.Datastore, logger *zap.Logger) (Parser, error)

NewParser creates a new instance of a Parser implementation based on config. It returns the appropriate parser implementation (OpenAI or Gemini) based on the configuration. Returns an error if the parser backend specified in config is not supported.