Syn

syn

CLI for [Synthetic.new](https://synthetic.new) AI API.

Install

go install github.com/dotcommander/syn@latest

Setup

export SYN_API_KEY="your-api-key"

Or create `~/.config/syn/config.yaml`:

api:
  key: "your-api-key"
  model: "hf:deepseek-ai/DeepSeek-V3.2"

Usage

Chat

# One-shot
syn "Explain goroutines"

# With file context
syn -f main.go "Review this code"

# Pipe input
pbpaste | syn "Summarize this"
cat error.log | syn "What went wrong?"

# Select model
syn -m kimi "Complex reasoning task"
syn -m coder "Refactor this function"

# JSON output
syn --json "List 3 facts about Go"

Interactive REPL

syn chat

Commands: `/help`, `/clear`, `/model`, `/context`, `/exit`

Web Search

syn search "golang error handling best practices"
syn search --json "react server components"

Vision

syn vision -f screenshot.png "What's in this image?"
syn vision -f https://example.com/diagram.png "Explain this diagram"

Embeddings

syn embed "Hello world"
syn embed "Text 1" "Text 2" "Text 3"
syn embed --json "For vector storage"

Models

syn model list

Evaluation

# Run insight-extraction eval across all listed models
syn eval --dataset testdata/eval/walter_lewin

# Save machine-readable report
syn eval --format json --out analysis-results/eval-report.json

# Evaluate only selected models
syn eval --models "hf:deepseek-ai/DeepSeek-V3.2,hf:moonshotai/Kimi-K2-Thinking"

# Keep persistent score history + leaderboard
syn eval --history analysis-results/eval-history.jsonl --leaderboard-out analysis-results/eval-leaderboard.md

Model Aliases

Alias	Model
kimi	Kimi-K2.5
qwen	Qwen3-235B-Thinking
coder	Qwen3-Coder-480B
glm	GLM-4.7
gpt	GPT-OSS-120B
deepseek	DeepSeek-V3.2 (default)
r1	DeepSeek