AI Tools

LarkOcrTool

Extract text from images using the Feishu OCR API (POST /optical_char_recognition/v1/image/basic_recognize). Useful for processing screenshots or scanned documents inside an agent.

use synaptic::lark::{LarkConfig, LarkOcrTool};
use synaptic::core::Tool;
use serde_json::json;

let config = LarkConfig::new("cli_xxx", "secret_xxx");
let tool = LarkOcrTool::new(config);

// Pass a base64-encoded image
let result = tool.call(json!({
    "image_base64": "<base64-encoded-image>"
})).await?;

println!("Recognized text: {}", result["text"]);

LarkTranslateTool

Translate text between languages using the Feishu Translation API (POST /translation/v1/text/translate). Supports all language pairs offered by the Feishu platform.

use synaptic::lark::{LarkConfig, LarkTranslateTool};
use synaptic::core::Tool;
use serde_json::json;

let config = LarkConfig::new("cli_xxx", "secret_xxx");
let tool = LarkTranslateTool::new(config);

let result = tool.call(json!({
    "source_language": "zh",
    "target_language": "en",
    "text": "你好，世界！"
})).await?;

println!("Translation: {}", result["translated_text"]);

LarkAsrTool

Transcribe audio files to text using the Feishu Speech-to-Text API (POST /speech_to_text/v1/speech/file_recognize). Supply a file_key of a previously uploaded audio file.

use synaptic::lark::{LarkAsrTool, LarkConfig};
use synaptic::core::Tool;
use serde_json::json;

let config = LarkConfig::new("cli_xxx", "secret_xxx");
let tool = LarkAsrTool::new(config);

let result = tool.call(json!({
    "file_key": "file_xxx"
})).await?;

println!("Transcript: {}", result["recognition_text"]);

LarkDocProcessTool

Extract structured entities from documents using the Feishu Document AI API (POST /document_ai/v1/entity/recognize). Returns structured key-value pairs from forms, invoices, and other document types.

use synaptic::lark::{LarkConfig, LarkDocProcessTool};
use synaptic::core::Tool;
use serde_json::json;

let config = LarkConfig::new("cli_xxx", "secret_xxx");
let tool = LarkDocProcessTool::new(config);

let result = tool.call(json!({
    "task_type": "invoice",
    "file_key": "file_xxx"
})).await?;

println!("Entities: {}", result["entities"]);