Skip to content

README.md

Latest commit

 

History

History
169 lines (126 loc) · 3.88 KB

README.md

File metadata and controls

169 lines (126 loc) · 3.88 KB
Raw

File Input Example

This example demonstrates how to process various types of file inputs (text, images, audio, and files) using OpenAI-compatible models.

Features

  • Text Input: Process plain text messages
  • Image Input: Analyze images (JPEG, PNG, GIF, WebP)
  • Audio Input: Process audio files (WAV format)
  • File Upload: Upload and analyze any file type
  • Streaming Support: Real-time streaming responses
  • Model Variants: Support for OpenAI and Hunyuan models

Quick Start

# Set your API key
export OPENAI_API_KEY="your-api-key-here"

# Basic text processing
go run main.go -text "Hello, how are you?"

# Image analysis
go run main.go -image path/to/image.jpg

# File analysis
go run main.go -file path/to/document.pdf -text "Analyze this file"

# Multiple inputs
go run main.go -text "What's in this image?" -image path/to/image.png

Command Line Options

Flag Description Default
-model Model to use gpt-4o
-variant Model variant (openai, hunyuan) openai
-text Text input to process -
-image Path to image file -
-audio Path to audio file -
-file Path to file for analysis -
-streaming Enable streaming mode true
-file-ids Use file IDs instead of base64 true

File Handling

The example supports two file handling modes:

File IDs Mode (Default)

Files are uploaded to the model provider and referenced by ID. This is the recommended approach for most use cases.

# Use file IDs (default)
go run main.go -file document.pdf -text "Analyze this PDF"

File Data Mode

File content is embedded directly in the message as base64 data.

# Use file data
go run main.go -file document.pdf -file-ids=false -text "Analyze this PDF"

Model Variants

OpenAI Variant

Standard OpenAI-compatible behavior with full file support.

go run main.go -variant openai -file data.json -text "Analyze this JSON"

Hunyuan Variant

Optimized for Hunyuan models with specific file handling.

go run main.go -variant hunyuan -file data.json -text "Analyze this JSON"

Examples

Text Analysis

go run main.go -text "Explain quantum computing in simple terms"

Image Analysis

go run main.go -image photo.jpg -text "What objects do you see in this image?"

Audio Processing

go run main.go -audio recording.wav -text "Transcribe this audio"

Document Analysis

go run main.go -file report.pdf -text "Summarize the key points in this document"

Custom Model

go run main.go -model gpt-4 -file code.py -text "Review this Python code"

Supported File Types

Images

  • JPEG (.jpg, .jpeg)
  • PNG (.png)
  • GIF (.gif)
  • WebP (.webp)

Audio

  • WAV (.wav)

Documents

  • Any file type (PDF, DOC, TXT, JSON, etc.)

Output Example

🚀 File Input Processing
Model: gpt-4o
Variant: openai
Streaming: true
File Mode: file_ids (recommended for Hunyuan/Gemini)
==================================================
✅ File processor ready!

📄 File input: document.pdf (mode: file_ids)
📤 File uploaded with ID: file-abc123def456
🤖 Assistant: I can see the contents of document.pdf. Here's what I found...

Configuration

API Key

Set your API key as an environment variable:

export OPENAI_API_KEY="your-api-key-here"

Streaming

Toggle streaming mode for real-time responses:

# Enable streaming (default)
go run main.go -streaming=true -text "Hello"

# Disable streaming
go run main.go -streaming=false -text "Hello"

Error Handling

The example includes comprehensive error handling for:

  • Invalid file paths
  • Unsupported file formats
  • API communication errors
  • Missing API keys

Dependencies

  • trpc-agent-go: Core framework
  • openai: Model provider
  • Standard library: context, flag, fmt, log, strings