zai - a Zig AI Library!

Installation

1. Declare zai as a project dependency with zig fetch:

   # latest version
   zig fetch --save git+https://github.com/AdjectiveAllison/zai.git#main
   
   # specific commit
   zig fetch --save git+https://github.com/AdjectiveAllison/zai.git#COMMIT

2. Expose zai as a module in your project's build.zig:

   pub fn build(b: *std.Build) void {
       const target = b.standardTargetOptions(.{});
       const optimize = b.standardOptimizeOption(.{});
   
       const opts = .{ .target = target, .optimize = optimize };   // 👈
       const zai = b.dependency("zai", opts).module("zai"); // 👈
   
       const exe = b.addExecutable(.{
           .name = "my-project",
           .root_source_file = b.path("src/main.zig"),
           .target = target,
           .optimize = optimize,
       });
   
       exe.root_module.addImport("zai", zai); // 👈
   
       // ...
   }

3. Import Zig AI into your code:

   const zai = @import("zai");

Features

• Support for chat completions and embeddings
• Streaming capabilities for real-time responses
• Easy integration with Zig projects
• Configurable model selection
• Automatic provider-specific API handling (in the works)

Usage

Chat Completion

Here's a basic example of using zai for chat completion:

const std = @import("std");
const zai = @import("zai");

pub fn main() !void {
    var gpa = std.heap.GeneralPurposeAllocator(.{}){};
    defer _ = gpa.deinit();
    const allocator = gpa.allocator();

    var ai: zai.AI = undefined;
    try ai.init(allocator, zai.Provider.OctoAI);
    defer ai.deinit();

    var messages = [_]zai.Message{
        .{
            .role = "system",
            .content = "You are a helpful AI assistant.",
        },
        .{
            .role = "user",
            .content = "Write one sentence about how cool it would be to use zig to call language models.",
        },
    };

    const payload = zai.CompletionPayload{
        .model = "meta-llama-3.1-8b-instruct",
        .messages = &messages,
        .temperature = 0.1,
        .stream = true,
    };

    var chat_completion: zai.ChatCompletion = undefined;
    chat_completion.init(allocator);
    defer chat_completion.deinit();

    try chat_completion.streamAndPrint(&ai, payload);
}

Embeddings

Here's how to use zai for generating embeddings:

const std = @import("std");
const zai = @import("zai");

pub fn main() !void {
    var gpa = std.heap.GeneralPurposeAllocator(.{}){};
    defer _ = gpa.deinit();
    const allocator = gpa.allocator();

    var ai: zai.AI = undefined;
    try ai.init(allocator, zai.Provider.OctoAI);
    defer ai.deinit();

    const input_text = "Zig is a general-purpose programming language and toolchain for maintaining robust, optimal, and reusable software.";

    const payload = zai.Embeddings.EmbeddingsPayload{
        .input = input_text,
        .model = "thenlper/gte-large",
    };

    var embeddings: zai.Embeddings = undefined;
    embeddings.init(allocator);
    defer embeddings.deinit();

    try embeddings.request(&ai, payload);

    std.debug.print("Embedding vector length: {d}\n", .{embeddings.embedding.len});
}

Examples

You can find more detailed examples in the examples/ directory:

1. Chat Completion
2. Embeddings

To run an example, use:

zig build chat-completion
# or
zig build embeddings

Configuration

zai uses environment variables for API keys (right now). Set the appropriate environment variable for your chosen provider:

• OpenAI: OPENAI_API_KEY
• OctoAI: OCTOAI_TOKEN
• TogetherAI: TOGETHER_API_KEY
• OpenRouter: OPENROUTER_API_KEY

License

zai is released under the MIT License. See the LICENSE file for details.