Merge branch 'development_action007_mahmoud' into development_actions007

This commit is contained in:
2025-03-11 20:41:52 +01:00
8 changed files with 843 additions and 381 deletions

60
examples/clients/jina.vsh Executable file
View File

@@ -0,0 +1,60 @@
#!/usr/bin/env -S v -n -w -gc none -cc tcc -d use_openssl -enable-globals run
import freeflowuniverse.herolib.clients.jina
mut jina_client := jina.get()!
// Create embeddings
embeddings := jina_client.create_embeddings(
input: ['Hello', 'World']
model: .jina_embeddings_v3
task: 'separation'
) or { panic('Error while creating embeddings: ${err}') }
println('Created embeddings: ${embeddings}')
// Rerank
rerank_result := jina_client.rerank(
model: .reranker_v2_base_multilingual
query: 'skincare products'
documents: ['Product A', 'Product B', 'Product C']
top_n: 2
) or { panic('Error while reranking: ${err}') }
println('Rerank result: ${rerank_result}')
// Train
train_result := jina_client.train(
model: .jina_clip_v1
input: [
jina.TrainingExample{
text: 'Sample text'
label: 'positive'
},
jina.TrainingExample{
image: 'https://letsenhance.io/static/73136da51c245e80edc6ccfe44888a99/1015f/MainBefore.jpg'
label: 'negative'
},
]
) or { panic('Error while training: ${err}') }
println('Train result: ${train_result}')
// Classify
classify_result := jina_client.classify(
model: .jina_clip_v1
input: [
jina.ClassificationInput{
text: 'A photo of a cat'
},
jina.ClassificationInput{
image: 'https://letsenhance.io/static/73136da51c245e80edc6ccfe44888a99/1015f/MainBefore.jpg'
},
]
labels: ['cat', 'dog']
) or { panic('Error while classifying: ${err}') }
println('Classification result: ${classify_result}')
classifiers := jina_client.list_classifiers() or { panic('Error fetching classifiers: ${err}') }
println('Classifiers: ${classifiers}')

View File

@@ -0,0 +1,291 @@
module jina
import json
import freeflowuniverse.herolib.core.httpconnection
// ClassificationTrainAccess represents the accessibility of the classifier
pub enum ClassificationTrainAccess {
public // Classifier is publicly accessible
private // Classifier is private (default)
}
// TrainingExample represents a single training example (either text or image with a label)
pub struct TrainingExample {
pub mut:
text ?string // Optional text content
image ?string // Optional image URL
label string // Required label
}
// ClassificationTrainOutput represents the response from the training endpoint
pub struct ClassificationTrainOutput {
pub mut:
classifier_id string // Identifier of the trained classifier
num_samples int // Number of samples used in training
usage ClassificationTrainUsage // Token usage details
}
// ClassificationTrainUsage represents token usage for the training request
pub struct ClassificationTrainUsage {
pub mut:
total_tokens int // Total tokens consumed
}
// ClassificationTrain represents parameters for the training request
@[params]
pub struct ClassificationTrain {
pub mut:
model ?JinaModel // Optional model identifier (e.g., jina-clip-v1)
classifier_id ?string // Optional existing classifier ID
access ?ClassificationTrainAccess = .private // Accessibility, defaults to private
input []TrainingExample // Array of training examples
num_iters ?int = 10 // Number of training iterations, defaults to 10
}
// TrainRequest represents the JSON request body for the /v1/train endpoint
struct TrainRequest {
mut:
model ?string
classifier_id ?string
access ?string
input []TrainingExample
num_iters ?int
}
// Train a classifier by sending a POST request to /v1/train
pub fn (mut j Jina) train(params ClassificationTrain) !ClassificationTrainOutput {
// Validate that only one of model or classifier_id is provided
mut model_provided := false
mut classifier_id_provided := false
if _ := params.model {
model_provided = true
}
if _ := params.classifier_id {
classifier_id_provided = true
}
if model_provided && classifier_id_provided {
return error('Provide either model or classifier_id, not both')
}
if model := params.model {
if model == .jina_embeddings_v3 {
return error('jina-embeddings-v3 is not a valid model for classification')
}
}
// Validate each training example has exactly one of text or image
for example in params.input {
mut text_provided := false
mut image_provided := false
if _ := example.text {
text_provided = true
}
if _ := example.image {
image_provided = true
}
if text_provided && image_provided {
return error('Each training example must have either text or image, not both')
}
if !text_provided && !image_provided {
return error('Each training example must have either text or image')
}
}
// Construct the request body
mut request := TrainRequest{
input: params.input
}
if v := params.model {
request.model = v.to_string() // Convert JinaModel enum to string
}
if v := params.classifier_id {
request.classifier_id = v
}
if v := params.access {
request.access = match v {
.public { 'public' }
.private { 'private' }
}
}
if v := params.num_iters {
request.num_iters = v
}
// Create and send the HTTP request
req := httpconnection.Request{
method: .post
prefix: 'v1/train'
dataformat: .json
data: json.encode(request)
}
mut httpclient := j.httpclient()!
response := httpclient.post_json_str(req)!
result := json.decode(ClassificationTrainOutput, response)!
return result
}
// TextDoc represents a text document for classification
pub struct TextDoc {
pub mut:
text string // The text content
}
// ImageDoc represents an image document for classification
pub struct ImageDoc {
pub mut:
image string // The image URL or base64-encoded string
}
// ClassificationInput represents a single input for classification (text or image)
pub struct ClassificationInput {
pub mut:
text ?string // Optional text content
image ?string // Optional image content
}
// ClassificationOutput represents the response from the classify endpoint
pub struct ClassificationOutput {
pub mut:
data []ClassificationResult // List of classification results
usage ClassificationUsage // Token usage details
}
// ClassificationResult represents a single classification result
pub struct ClassificationResult {
pub mut:
index int // Index of the input
prediction string // Predicted label
score f64 // Confidence score
object string // Type of object (e.g., "classification")
predictions []LabelScore // List of label scores
}
// LabelScore represents a label and its corresponding score
pub struct LabelScore {
pub mut:
label string // Label name
score f64 // Confidence score
}
// ClassificationUsage represents token usage for the classification request
pub struct ClassificationUsage {
pub mut:
total_tokens int // Total tokens consumed
}
// ClassifyRequest represents the JSON request body for the /v1/classify endpoint
struct ClassifyRequest {
mut:
model ?string
classifier_id ?string
input []ClassificationInput
labels []string
}
// ClassifyParams represents parameters for the classification request
@[params]
pub struct ClassifyParams {
pub mut:
model ?JinaModel // Optional model identifier
classifier_id ?string // Optional classifier ID
input []ClassificationInput // Array of inputs (text or image)
labels []string // List of labels for classification
}
// Classify inputs by sending a POST request to /v1/classify
pub fn (mut j Jina) classify(params ClassifyParams) !ClassificationOutput {
// Validate that only one of model or classifier_id is provided
mut model_provided := false
mut classifier_id_provided := false
if _ := params.model {
model_provided = true
}
if _ := params.classifier_id {
classifier_id_provided = true
}
if model_provided && classifier_id_provided {
return error('Provide either model or classifier_id, not both')
}
if !model_provided && !classifier_id_provided {
return error('Either model or classifier_id must be provided')
}
// Validate each input has exactly one of text or image
for input in params.input {
mut text_provided := false
mut image_provided := false
if _ := input.text {
text_provided = true
}
if _ := input.image {
image_provided = true
}
if text_provided && image_provided {
return error('Each input must have either text or image, not both')
}
if !text_provided && !image_provided {
return error('Each input must have either text or image')
}
}
// Construct the request body
mut request := ClassifyRequest{
input: params.input
labels: params.labels
}
if v := params.model {
request.model = v.to_string() // Convert JinaModel enum to string
}
if v := params.classifier_id {
request.classifier_id = v
}
// Create and send the HTTP request
req := httpconnection.Request{
method: .post
prefix: 'v1/classify'
dataformat: .json
data: json.encode(request)
}
mut httpclient := j.httpclient()!
response := httpclient.post_json_str(req)!
result := json.decode(ClassificationOutput, response)!
return result
}
// Define the Classifier struct
pub struct Classifier {
pub mut:
classifier_id string
model_name string
labels []string
access string
updated_number int
used_number int
created_at string
updated_at string
used_at ?string
metadata map[string]string
}
// Implement the list_classifiers function
pub fn (mut j Jina) list_classifiers() ![]Classifier {
req := httpconnection.Request{
method: .get
prefix: 'v1/classifiers'
}
mut httpclient := j.httpclient()!
response := httpclient.get(req)!
println('response: ${response}')
classifiers := json.decode([]Classifier, response)!
return classifiers
}

View File

@@ -1,185 +1,245 @@
module jina
import freeflowuniverse.herolib.core.httpconnection
import json
import os
import json
@[params]
pub struct CreateEmbeddingParams {
pub mut:
input []string @[required] // Input texts
model JinaModel @[required] // Model name
task string @[required] // Task type
type_ ?EmbeddingType // embedding type
truncate ?TruncateType // truncation type
late_chunking ?bool // Flag to determine if late chunking is applied
}
// Create embeddings for input texts
pub fn (mut j Jina) create_embeddings(input []string, model string, task string) !ModelEmbeddingOutput {
pub fn (mut j Jina) create_embeddings(params CreateEmbeddingParams) !ModelEmbeddingOutput {
task := task_type_from_string(params.task)!
mut embedding_input := TextEmbeddingInput{
model: model
input: input
task: task
input: params.input
model: params.model.to_string()
task: task
}
if v := params.type_ {
embedding_input.type_ = v
}
if v := params.truncate {
embedding_input.truncate = v
}
embedding_input.late_chunking = if _ := params.late_chunking { true } else { false }
req := httpconnection.Request{
method: .post
prefix: 'v1/embeddings'
method: .post
prefix: 'v1/embeddings'
dataformat: .json
data: embedding_input.to_json()
data: embedding_input.to_json()
}
response := j.http.get(req)!
mut httpclient := j.httpclient()!
response := httpclient.post_json_str(req)!
return parse_model_embedding_output(response)!
}
// Create embeddings with a TextDoc input
pub fn (mut j Jina) create_embeddings_with_docs(args TextEmbeddingInput) !ModelEmbeddingOutput {
req := httpconnection.Request{
method: .post
prefix: 'v1/embeddings'
dataformat: .json
data: json.encode(args)
}
response := j.http.get(req)!
return parse_model_embedding_output(response)!
@[params]
pub struct RerankParams {
pub mut:
model JinaRerankModel @[required]
query string @[required]
documents []string @[required]
top_n ?int // Optional: Number of top results to return
return_documents ?bool // Optional: Flag to determine if the documents should be returned
}
// Rerank documents based on a query
pub fn (mut j Jina) rerank(query string, documents []string, model string, top_n int) !RankingOutput {
mut rank_input := RankAPIInput{
model: model
query: query
documents: documents
top_n: top_n
pub fn (mut j Jina) rerank(params RerankParams) !RankingOutput {
mut rank_input := RerankInput{
model: params.model.to_string()
query: params.query
documents: params.documents
}
if v := params.top_n {
rank_input.top_n = v
}
if v := params.return_documents {
rank_input.return_documents = v
}
req := httpconnection.Request{
method: .post
prefix: 'v1/rerank'
method: .post
prefix: 'v1/rerank'
dataformat: .json
data: rank_input.to_json()
data: json.encode(rank_input)
}
response := j.http.get(req)!
return parse_ranking_output(response)!
mut httpclient := j.httpclient()!
response := httpclient.post_json_str(req)!
return json.decode(RankingOutput, response)!
}
// Simplified rerank function with default top_n
pub fn (mut j Jina) rerank_simple(query string, documents []string, model string) !RankingOutput {
return j.rerank(query, documents, model, 0)!
}
// // Create embeddings with a TextDoc input
// pub fn (mut j Jina) create_embeddings_with_docs(args TextEmbeddingInput) !ModelEmbeddingOutput {
// Classify input texts
pub fn (mut j Jina) classify(input []string, model string, labels []string) !ClassificationOutput {
mut classification_input := ClassificationAPIInput{
model: model
input: input
labels: labels
}
req := httpconnection.Request{
method: .post
prefix: 'v1/classify'
dataformat: .json
data: classification_input.to_json()
}
response := j.http.get(req)!
return parse_classification_output(response)!
}
// req := httpconnection.Request{
// method: .post
// prefix: 'v1/embeddings'
// dataformat: .json
// data: json.encode(args)
// }
// Train a classifier
pub fn (mut j Jina) train(examples []TrainingExample, model string, access string) !TrainingOutput {
mut training_input := TrainingAPIInput{
model: model
input: examples
access: access
}
req := httpconnection.Request{
method: .post
prefix: 'v1/train'
dataformat: .json
data: training_input.to_json()
}
response := j.http.get(req)!
return parse_training_output(response)!
}
// response := j.http.get(req)!
// return parse_model_embedding_output(response)!
// }
// List classifiers
pub fn (mut j Jina) list_classifiers() !string {
req := httpconnection.Request{
method: .get
prefix: 'v1/classifiers'
}
return j.http.get(req)!
}
// // Rerank documents based on a query
// pub fn (mut j Jina) rerank(query string, documents []string, model string, top_n int) !RankingOutput {
// mut rank_input := RankAPIInput{
// model: model
// query: query
// documents: documents
// top_n: top_n
// }
// Delete a classifier
pub fn (mut j Jina) delete_classifier(classifier_id string) !bool {
req := httpconnection.Request{
method: .delete
prefix: 'v1/classifiers/${classifier_id}'
}
j.http.get(req)!
return true
}
// req := httpconnection.Request{
// method: .post
// prefix: 'v1/rerank'
// dataformat: .json
// data: rank_input.to_json()
// }
// Create multi-vector embeddings
pub fn (mut j Jina) create_multi_vector(input []string, model string) !ColbertModelEmbeddingsOutput {
mut data := map[string]json.Any{}
data['model'] = model
data['input'] = input
req := httpconnection.Request{
method: .post
prefix: 'v1/multi-embeddings'
dataformat: .json
data: json.encode(data)
}
response := j.http.get(req)!
return parse_colbert_model_embeddings_output(response)!
}
// response := j.http.get(req)!
// return parse_ranking_output(response)!
// }
// Start a bulk embedding job
pub fn (mut j Jina) start_bulk_embedding(file_path string, model string, email string) !BulkEmbeddingJobResponse {
// This endpoint requires multipart/form-data which is not directly supported by the current HTTPConnection
// We need to implement a custom solution for this
return error('Bulk embedding is not implemented yet')
}
// // Simplified rerank function with default top_n
// pub fn (mut j Jina) rerank_simple(query string, documents []string, model string) !RankingOutput {
// return j.rerank(query, documents, model, 0)!
// }
// Check the status of a bulk embedding job
pub fn (mut j Jina) check_bulk_embedding_status(job_id string) !BulkEmbeddingJobResponse {
req := httpconnection.Request{
method: .get
prefix: 'v1/bulk-embeddings/${job_id}'
}
response := j.http.get(req)!
return parse_bulk_embedding_job_response(response)!
}
// // Classify input texts
// pub fn (mut j Jina) classify(input []string, model string, labels []string) !ClassificationOutput {
// mut classification_input := ClassificationAPIInput{
// model: model
// input: input
// labels: labels
// }
// Download the result of a bulk embedding job
pub fn (mut j Jina) download_bulk_embedding_result(job_id string) !DownloadResultResponse {
req := httpconnection.Request{
method: .post
prefix: 'v1/bulk-embeddings/${job_id}/download-result'
}
response := j.http.get(req)!
return parse_download_result_response(response)!
}
// req := httpconnection.Request{
// method: .post
// prefix: 'v1/classify'
// dataformat: .json
// data: classification_input.to_json()
// }
// Check if the API key is valid by making a simple request
pub fn (mut j Jina) check_auth() !bool {
req := httpconnection.Request{
method: .get
prefix: '/'
}
j.http.get(req) or {
return error('Failed to connect to Jina API: ${err}')
}
// If we get a response, the API key is valid
return true
}
// response := j.http.get(req)!
// return parse_classification_output(response)!
// }
// // Train a classifier
// pub fn (mut j Jina) train(examples []TrainingExample, model string, access string) !TrainingOutput {
// mut training_input := TrainingAPIInput{
// model: model
// input: examples
// access: access
// }
// req := httpconnection.Request{
// method: .post
// prefix: 'v1/train'
// dataformat: .json
// data: training_input.to_json()
// }
// response := j.http.get(req)!
// return parse_training_output(response)!
// }
// // List classifiers
// pub fn (mut j Jina) list_classifiers() !string {
// req := httpconnection.Request{
// method: .get
// prefix: 'v1/classifiers'
// }
// return j.http.get(req)!
// }
// // Delete a classifier
// pub fn (mut j Jina) delete_classifier(classifier_id string) !bool {
// req := httpconnection.Request{
// method: .delete
// prefix: 'v1/classifiers/${classifier_id}'
// }
// j.http.get(req)!
// return true
// }
// // Create multi-vector embeddings
// pub fn (mut j Jina) create_multi_vector(input []string, model string) !ColbertModelEmbeddingsOutput {
// mut data := map[string]json.Any{}
// data['model'] = model
// data['input'] = input
// req := httpconnection.Request{
// method: .post
// prefix: 'v1/multi-embeddings'
// dataformat: .json
// data: json.encode(data)
// }
// response := j.http.get(req)!
// return parse_colbert_model_embeddings_output(response)!
// }
// // Start a bulk embedding job
// pub fn (mut j Jina) start_bulk_embedding(file_path string, model string, email string) !BulkEmbeddingJobResponse {
// // This endpoint requires multipart/form-data which is not directly supported by the current HTTPConnection
// // We need to implement a custom solution for this
// return error('Bulk embedding is not implemented yet')
// }
// // Check the status of a bulk embedding job
// pub fn (mut j Jina) check_bulk_embedding_status(job_id string) !BulkEmbeddingJobResponse {
// req := httpconnection.Request{
// method: .get
// prefix: 'v1/bulk-embeddings/${job_id}'
// }
// response := j.http.get(req)!
// return parse_bulk_embedding_job_response(response)!
// }
// // Download the result of a bulk embedding job
// pub fn (mut j Jina) download_bulk_embedding_result(job_id string) !DownloadResultResponse {
// req := httpconnection.Request{
// method: .post
// prefix: 'v1/bulk-embeddings/${job_id}/download-result'
// }
// response := j.http.get(req)!
// return parse_download_result_response(response)!
// }
// // Check if the API key is valid by making a simple request
// pub fn (mut j Jina) check_auth() !bool {
// req := httpconnection.Request{
// method: .get
// prefix: '/'
// }
// j.http.get(req) or {
// return error('Failed to connect to Jina API: ${err}')
// }
// // If we get a response, the API key is valid
// return true
// }

View File

@@ -0,0 +1,87 @@
module jina
import time
fn setup_client() !&Jina {
mut client := get()!
return client
}
fn test_create_embeddings() {
time.sleep(1 * time.second)
mut client := setup_client()!
embeddings := client.create_embeddings(
input: ['Hello', 'World']
model: .jina_embeddings_v3
task: 'separation'
) or { panic('Error while creating embeddings: ${err}') }
assert embeddings.data.len > 0
assert embeddings.object == 'list' // Check the object type
assert embeddings.model == 'jina-embeddings-v3'
}
fn test_rerank() {
time.sleep(1 * time.second)
mut client := setup_client()!
rerank_result := client.rerank(
model: .reranker_v2_base_multilingual
query: 'skincare products'
documents: ['Product A', 'Product B', 'Product C']
top_n: 2
) or { panic('Error while reranking: ${err}') }
assert rerank_result.results.len == 2
assert rerank_result.model == 'jina-reranker-v2-base-multilingual'
}
fn test_train() {
time.sleep(1 * time.second)
mut client := setup_client()!
train_result := client.train(
model: .jina_clip_v1
input: [
TrainingExample{
text: 'A photo of a cat'
label: 'cat'
},
TrainingExample{
text: 'A photo of a dog'
label: 'dog'
},
]
) or { panic('Error while training: ${err}') }
assert train_result.classifier_id.len > 0
assert train_result.num_samples == 2
}
fn test_classify() {
time.sleep(1 * time.second)
mut client := setup_client()!
classify_result := client.classify(
model: .jina_clip_v1
input: [
ClassificationInput{
text: 'A photo of a cat'
},
ClassificationInput{
image: 'https://letsenhance.io/static/73136da51c245e80edc6ccfe44888a99/1015f/MainBefore.jpg'
},
]
labels: ['cat', 'dog']
) or { panic('Error while classifying: ${err}') }
assert classify_result.data.len == 2
assert classify_result.data[0].prediction in ['cat', 'dog']
assert classify_result.data[1].prediction in ['cat', 'dog']
assert classify_result.data[0].object == 'classification'
assert classify_result.data[1].object == 'classification'
}
fn test_get_classifiers() {
time.sleep(1 * time.second)
mut client := setup_client()!
classifiers := client.list_classifiers() or { panic('Error fetching classifiers: ${err}') }
assert classifiers.len != 0
}

View File

@@ -1,6 +1,5 @@
module jina
import freeflowuniverse.herolib.data.paramsparser
import freeflowuniverse.herolib.data.encoderhero
import freeflowuniverse.herolib.core.httpconnection
import net.http
@@ -17,16 +16,29 @@ const env_key = 'JINAKEY'
@[heap]
pub struct Jina {
pub mut:
name string = 'default'
secret string
base_url string = api_base_url
http httpconnection.HTTPConnection @[str: skip]
name string = 'default'
secret string
base_url string = api_base_url
// http httpconnection.HTTPConnection @[str: skip]
}
fn (mut self Jina) httpclient() !&httpconnection.HTTPConnection {
mut http_conn := httpconnection.new(
name: 'Jina_vclient'
url: self.base_url
)!
// Add authentication header if API key is provided
if self.secret.len > 0 {
http_conn.default_header.add(.authorization, 'Bearer ${self.secret}')
}
return http_conn
}
// your checking & initialization code if needed
fn obj_init(mycfg_ Jina) !Jina {
mut mycfg := mycfg_
// Get API key from environment variable if not set
if mycfg.secret == '' {
if env_key in os.environ() {
@@ -35,16 +47,7 @@ fn obj_init(mycfg_ Jina) !Jina {
return error('Jina API key not provided and ${env_key} environment variable not set')
}
}
// Initialize HTTP connection
mut header := http.new_header()
header.add_custom('Authorization', 'Bearer ${mycfg.secret}')
mycfg.http = httpconnection.HTTPConnection{
base_url: mycfg.base_url
default_header: header
}
return mycfg
}

View File

@@ -2,53 +2,53 @@ module jina
import json
// JinaModelEnumerator represents the available models for Jina API
pub enum JinaModelEnumerator {
clip_v1 // jina-clip-v1, 223M, 768
clip_v2 // jina-clip-v2, 865M, 1024
embeddings_v2_base_en // jina-embeddings-v2-base-en, 137M, 768
embeddings_v2_base_es // jina-embeddings-v2-base-es, 161M, 768
embeddings_v2_base_de // jina-embeddings-v2-base-de, 161M, 768
embeddings_v2_base_zh // jina-embeddings-v2-base-zh, 161M, 768
embeddings_v2_base_code // jina-embeddings-v2-base-code, 137M, 768
embeddings_v3 // jina-embeddings-v3, 570M, 1024
// JinaModel represents the available Jina models
pub enum JinaModel {
jina_clip_v1
jina_clip_v2
jina_embeddings_v2_base_en
jina_embeddings_v2_base_es
jina_embeddings_v2_base_de
jina_embeddings_v2_base_zh
jina_embeddings_v2_base_code
jina_embeddings_v3
}
// to_string converts JinaModelEnumerator enum to its string representation
pub fn (m JinaModelEnumerator) to_string() string {
// to_string converts a JinaModel enum to its string representation as expected by the Jina API
pub fn (m JinaModel) to_string() string {
return match m {
.clip_v1 { 'jina-clip-v1' }
.clip_v2 { 'jina-clip-v2' }
.embeddings_v2_base_en { 'jina-embeddings-v2-base-en' }
.embeddings_v2_base_es { 'jina-embeddings-v2-base-es' }
.embeddings_v2_base_de { 'jina-embeddings-v2-base-de' }
.embeddings_v2_base_zh { 'jina-embeddings-v2-base-zh' }
.embeddings_v2_base_code { 'jina-embeddings-v2-base-code' }
.embeddings_v3 { 'jina-embeddings-v3' }
.jina_clip_v1 { 'jina-clip-v1' }
.jina_clip_v2 { 'jina-clip-v2' }
.jina_embeddings_v2_base_en { 'jina-embeddings-v2-base-en' }
.jina_embeddings_v2_base_es { 'jina-embeddings-v2-base-es' }
.jina_embeddings_v2_base_de { 'jina-embeddings-v2-base-de' }
.jina_embeddings_v2_base_zh { 'jina-embeddings-v2-base-zh' }
.jina_embeddings_v2_base_code { 'jina-embeddings-v2-base-code' }
.jina_embeddings_v3 { 'jina-embeddings-v3' }
}
}
// from_string converts string to JinaModelEnumerator enum
pub fn jina_model_from_string(s string) ?JinaModelEnumerator {
// from_string converts a string to a JinaModel enum, returning an error if the string is invalid
pub fn jina_model_from_string(s string) !JinaModel {
return match s {
'jina-clip-v1' { JinaModelEnumerator.clip_v1 }
'jina-clip-v2' { JinaModelEnumerator.clip_v2 }
'jina-embeddings-v2-base-en' { JinaModelEnumerator.embeddings_v2_base_en }
'jina-embeddings-v2-base-es' { JinaModelEnumerator.embeddings_v2_base_es }
'jina-embeddings-v2-base-de' { JinaModelEnumerator.embeddings_v2_base_de }
'jina-embeddings-v2-base-zh' { JinaModelEnumerator.embeddings_v2_base_zh }
'jina-embeddings-v2-base-code' { JinaModelEnumerator.embeddings_v2_base_code }
'jina-embeddings-v3' { JinaModelEnumerator.embeddings_v3 }
else { error('Invalid model string: $s') }
'jina-clip-v1' { JinaModel.jina_clip_v1 }
'jina-clip-v2' { JinaModel.jina_clip_v2 }
'jina-embeddings-v2-base-en' { JinaModel.jina_embeddings_v2_base_en }
'jina-embeddings-v2-base-es' { JinaModel.jina_embeddings_v2_base_es }
'jina-embeddings-v2-base-de' { JinaModel.jina_embeddings_v2_base_de }
'jina-embeddings-v2-base-zh' { JinaModel.jina_embeddings_v2_base_zh }
'jina-embeddings-v2-base-code' { JinaModel.jina_embeddings_v2_base_code }
'jina-embeddings-v3' { JinaModel.jina_embeddings_v3 }
else { error('Invalid Jina model string: ${s}') }
}
}
// EmbeddingType represents the available embedding types
pub enum EmbeddingType {
float // "float"
base64 // "base64"
binary // "binary"
ubinary // "ubinary"
float // "float"
base64 // "base64"
binary // "binary"
ubinary // "ubinary"
}
// to_string converts EmbeddingType enum to its string representation
@@ -68,17 +68,17 @@ pub fn embedding_type_from_string(s string) !EmbeddingType {
'base64' { EmbeddingType.base64 }
'binary' { EmbeddingType.binary }
'ubinary' { EmbeddingType.ubinary }
else { error('Invalid embedding type string: $s') }
else { error('Invalid embedding type string: ${s}') }
}
}
// TaskType represents the available task types for embeddings
pub enum TaskType {
retrieval_query // "retrieval.query"
retrieval_passage // "retrieval.passage"
text_matching // "text-matching"
classification // "classification"
separation // "separation"
retrieval_query // "retrieval.query"
retrieval_passage // "retrieval.passage"
text_matching // "text-matching"
classification // "classification"
separation // "separation"
}
// to_string converts TaskType enum to its string representation
@@ -100,13 +100,13 @@ pub fn task_type_from_string(s string) !TaskType {
'text-matching' { TaskType.text_matching }
'classification' { TaskType.classification }
'separation' { TaskType.separation }
else { error('Invalid task type string: $s') }
else { error('Invalid task type string: ${s}') }
}
}
// TruncateType represents the available truncation options
pub enum TruncateType {
none // "NONE"
none_ // "NONE"
start // "START"
end // "END"
}
@@ -114,7 +114,7 @@ pub enum TruncateType {
// to_string converts TruncateType enum to its string representation
pub fn (t TruncateType) to_string() string {
return match t {
.none { 'NONE' }
.none_ { 'NONE' }
.start { 'START' }
.end { 'END' }
}
@@ -123,90 +123,89 @@ pub fn (t TruncateType) to_string() string {
// from_string converts string to TruncateType enum
pub fn truncate_type_from_string(s string) !TruncateType {
return match s {
'NONE' { TruncateType.none }
'NONE' { TruncateType.none_ }
'START' { TruncateType.start }
'END' { TruncateType.end }
else { error('Invalid truncate type string: $s') }
else { error('Invalid truncate type string: ${s}') }
}
}
// TextEmbeddingInputRaw represents the raw input for text embedding requests as sent to the server
struct TextEmbeddingInputRaw {
mut:
model string = 'jina-embeddings-v2-base-en'
input []string @[required]
task string // Optional: task type as string
type_ string @[json: 'type'] // Optional: embedding type as string
truncate string // Optional: "NONE", "START", "END"
late_chunking bool // Optional: Flag to determine if late chunking is applied
model string = 'jina-embeddings-v2-base-en'
input []string @[required]
task string // Optional: task type as string
type_ string @[json: 'type'] // Optional: embedding type as string
truncate string // Optional: "NONE", "START", "END"
late_chunking bool // Optional: Flag to determine if late chunking is applied
}
// TextEmbeddingInput represents the input for text embedding requests with enum types
pub struct TextEmbeddingInput {
pub mut:
model JinaModelEnumerator = JinaModelEnumerator.embeddings_v2_base_en
input []string @[required]
task TaskType // task type
type_ EmbeddingType // embedding type
truncate TruncateType // truncation type
late_chunking bool //Flag to determine if late chunking is applied
model string = 'jina-embeddings-v2-base-en'
input []string @[required]
task TaskType // task type
type_ ?EmbeddingType // embedding type
truncate ?TruncateType // truncation type
late_chunking ?bool // Flag to determine if late chunking is applied
}
// dumps converts TextEmbeddingInput to JSON string
pub fn (t TextEmbeddingInput) dumps() !string {
mut raw := TextEmbeddingInputRaw{
model: t.model.to_string()
input: t.input
late_chunking: t.late_chunking
model: t.model
input: t.input
late_chunking: if v := t.late_chunking { true } else { false }
}
raw.task = t.task.to_string()
raw.type_ = t.type_.to_string()
raw.truncate = t.truncate.to_string()
if v := t.type_ {
raw.type_ = v.to_string()
}
if v := t.truncate {
raw.truncate = v.to_string()
}
return json.encode(raw)
}
// from_raw converts TextEmbeddingInputRaw to TextEmbeddingInput
pub fn loads_text_embedding_input(text string ) !TextEmbeddingInput {
// TODO: go from text to InputObject over json
mut input := TextEmbeddingInput{
model: jina_model_from_string(raw.model)?
input: raw.input
late_chunking: raw.late_chunking
}
if raw.task != '' {
input.task = task_type_from_string(raw.task)!
}
if raw.type_ != '' {
input.type_ = embedding_type_from_string(raw.type_)!
}
if raw.truncate != '' {
input.truncate = truncate_type_from_string(raw.truncate)!
}
return input
}
// pub fn loads_text_embedding_input(text string) !TextEmbeddingInput {
// // TODO: go from text to InputObject over json
// // mut input := TextEmbeddingInput{
// // model: jina_model_from_string(raw.model)?
// // input: raw.input
// // late_chunking: raw.late_chunking
// // }
// // if raw.task != '' {
// // input.task = task_type_from_string(raw.task)!
// // }
// // if raw.type_ != '' {
// // input.type_ = embedding_type_from_string(raw.type_)!
// // }
// // if raw.truncate != '' {
// // input.truncate = truncate_type_from_string(raw.truncate)!
// // }
// return TextEmbeddingInput{}
// }
// loads converts a JSON string to TextEmbeddingInput
pub fn loads(text string) !TextEmbeddingInput {
// First decode the JSON string to the raw struct
raw := json.decode(TextEmbeddingInputRaw, text) or {
return error('Failed to decode JSON: $err')
}
// Then convert the raw struct to the typed struct
return text_embedding_input_from_raw(raw)
}
// pub fn loads(text string) !TextEmbeddingInput {
// // First decode the JSON string to the raw struct
// raw := json.decode(TextEmbeddingInputRaw, text) or {
// return error('Failed to decode JSON: ${err}')
// }
// TextDoc represents a document with ID and text for embedding
pub struct TextDoc {
pub mut:
id string
text string
}
// // Then convert the raw struct to the typed struct
// return text_embedding_input_from_raw(raw)
// }
// ModelEmbeddingOutput represents the response from embedding requests
pub struct ModelEmbeddingOutput {

View File

@@ -1,91 +1,6 @@
module jina
// RankAPIInput represents the input for reranking requests
// model:
// jina-reranker-v2-base-multilingual, 278M
// jina-reranker-v1-base-en, 137M
// jina-reranker-v1-tiny-en, 33M
// jina-reranker-v1-turbo-en, 38M
// jina-colbert-v1-en, 137M
pub struct RankAPIInputRAW {
pub mut:
model string @[required]
query string @[required]
documents []string @[required]
top_n int // Optional: Number of top results to return
}
// RankingOutput represents the response from reranking requests
pub struct RankingOutput {
pub mut:
model string
results []RankResult
usage Usage
object string
}
// RankResult represents a single reranking result
pub struct RankResult {
pub mut:
document string
index int
relevance_score f64
}
// ClassificationAPIInput represents the input for classification requests
pub struct ClassificationAPIInput {
pub mut:
model string @[required]
input []string @[required]
labels []string @[required]
}
// ClassificationOutput represents the response from classification requests
pub struct ClassificationOutput {
pub mut:
model string
data []ClassificationData
usage Usage
object string
}
// ClassificationData represents a single classification result
pub struct ClassificationData {
pub mut:
classifications []Classification
index int
}
// Classification represents a single label classification with score
pub struct Classification {
pub mut:
label string
score f64
}
// TrainingExample represents a single training example for classifier training
pub struct TrainingExample {
pub mut:
text string
label string
}
// TrainingAPIInput represents the input for training a classifier
pub struct TrainingAPIInput {
pub mut:
model string @[required]
input []TrainingExample @[required]
access string // Optional: "public" or "private"
}
// TrainingOutput represents the response from training a classifier
pub struct TrainingOutput {
pub mut:
classifier_id string
model string
status string
object string
}
import json
// BulkEmbeddingJobResponse represents the response from bulk embedding operations
pub struct BulkEmbeddingJobResponse {
@@ -136,9 +51,9 @@ pub mut:
// ValidationError represents a single validation error
pub struct ValidationError {
pub mut:
loc []string
msg string
type_ string @[json: 'type'] // 'type' is a keyword, so we need to specify the JSON name
loc []string
msg string
type_ string @[json: 'type'] // 'type' is a keyword, so we need to specify the JSON name
}
// Serialize and deserialize functions for the main request/response types
@@ -158,36 +73,16 @@ pub fn parse_model_embedding_output(json_str string) !ModelEmbeddingOutput {
return json.decode(ModelEmbeddingOutput, json_str)
}
// Serialize RankAPIInput to JSON
pub fn (input RankAPIInput) to_json() string {
return json.encode(input)
}
// // Serialize RankAPIInput to JSON
// pub fn (input RankAPIInput) to_json() string {
// return json.encode(input)
// }
// Parse JSON to RankingOutput
pub fn parse_ranking_output(json_str string) !RankingOutput {
return json.decode(RankingOutput, json_str)
}
// Serialize ClassificationAPIInput to JSON
pub fn (input ClassificationAPIInput) to_json() string {
return json.encode(input)
}
// Parse JSON to ClassificationOutput
pub fn parse_classification_output(json_str string) !ClassificationOutput {
return json.decode(ClassificationOutput, json_str)
}
// Serialize TrainingAPIInput to JSON
pub fn (input TrainingAPIInput) to_json() string {
return json.encode(input)
}
// Parse JSON to TrainingOutput
pub fn parse_training_output(json_str string) !TrainingOutput {
return json.decode(TrainingOutput, json_str)
}
// Parse JSON to BulkEmbeddingJobResponse
pub fn parse_bulk_embedding_job_response(json_str string) !BulkEmbeddingJobResponse {
return json.decode(BulkEmbeddingJobResponse, json_str)

View File

@@ -0,0 +1,67 @@
module jina
import json
pub enum JinaRerankModel {
reranker_v2_base_multilingual // 278M
reranker_v1_base_en // 137M
reranker_v1_tiny_en // 33M
reranker_v1_turbo_en // 38M
colbert_v1_en // 137M
}
// RankAPIInput represents the input for reranking requests
pub struct RerankInput {
pub mut:
model string @[required] // Model name
query string @[required] // Query text
documents []string @[required] // Document texts
top_n ?int // Optional: Number of top results to return
return_documents ?bool // Optional: Flag to determine if the documents should be returned
}
// RankingOutput represents the response from reranking requests
pub struct RankingOutput {
pub mut:
model string
results []RankResult
usage Usage
object string
}
// RankResult represents a single reranking result
pub struct RankResult {
pub mut:
document RankDocument
index int
relevance_score f64
}
// RankDocument represents a single document for reranking
pub struct RankDocument {
pub mut:
text string
}
// to_string converts a JinaRerankModel enum to its string representation as expected by the Jina API
pub fn (m JinaRerankModel) to_string() string {
return match m {
.reranker_v2_base_multilingual { 'jina-reranker-v2-base-multilingual' }
.reranker_v1_base_en { 'jina-reranker-v1-base-en' }
.reranker_v1_tiny_en { 'jina-reranker-v1-tiny-en' }
.reranker_v1_turbo_en { 'jina-reranker-v1-turbo-en' }
.colbert_v1_en { 'jina-colbert-v1-en' }
}
}
// from_string converts a string to a JinaRerankModel enum, returning an error if the string is invalid
pub fn jina_rerank_model_from_string(s string) !JinaRerankModel {
return match s {
'jina-reranker-v2-base-multilingual' { JinaRerankModel.reranker_v2_base_multilingual }
'jina-reranker-v1-base-en' { JinaRerankModel.reranker_v1_base_en }
'jina-reranker-v1-tiny-en' { JinaRerankModel.reranker_v1_tiny_en }
'jina-reranker-v1-turbo-en' { JinaRerankModel.reranker_v1_turbo_en }
'jina-colbert-v1-en' { JinaRerankModel.colbert_v1_en }
else { error('Invalid JinaRerankModel string: ${s}') }
}
}