Add OpenGradient plugin

typescript/.changeset/tall-pants-turn.md

@@ -0,0 +1,5 @@
+---
+"@goat-sdk/plugin-opengradient": patch
+---
+
+Release package

typescript/packages/plugins/opengradient/README.md

@@ -0,0 +1,44 @@
+<div align="center">
+<a href="https://github.com/goat-sdk/goat">
+<img src="https://github.com/user-attachments/assets/5fc7f121-259c-492c-8bca-f15fe7eb830c" alt="GOAT" width="100px" height="auto" style="object-fit: contain;">
+</a>
+</div>
+
+# OpenGradient GOAT Plugin
+
+This plugin integrates the OpenGradient service with the GOAT SDK, providing on-chain ML model inference and LLM interactions.
+
+## Installation
+```bash
+npm install @goat-sdk/plugin-opengradient
+yarn add @goat-sdk/plugin-opengradient
+pnpm add @goat-sdk/plugin-opengradient
+```
+
+## Usage
+```typescript
+import { opengradient } from "@goat-sdk/plugin-opengradient";
+
+const tools = await getOnChainTools({
+    wallet: viem(wallet),
+    plugins: [
+        opengradient(),
+    ],
+});
+```
+
+## Tools
+* `opengradient_model_inference` - Run inference on machine learning models using OpenGradient
+* `opengradient_llm_completion` - Generate text completions using LLMs through OpenGradient
+* `opengradient_llm_chat` - Interact with LLMs using a chat interface through OpenGradient
+
+
+<footer>
+<br/>
+<br/>
+<div>
+<a href="https://github.com/goat-sdk/goat">
+  <img src="https://github.com/user-attachments/assets/59fa5ddc-9d47-4d41-a51a-64f6798f94bd" alt="GOAT" width="100%" height="auto" style="object-fit: contain; max-width: 800px;">
+</a>
+</div>
+</footer>

typescript/packages/plugins/opengradient/package.json

@@ -0,0 +1,33 @@
+{
+    "name": "@goat-sdk/plugin-opengradient",
+    "version": "0.1.0",
+    "files": ["dist/**/*", "README.md", "package.json"],
+    "scripts": {
+        "build": "tsup",
+        "clean": "rm -rf dist",
+        "test": "vitest run --passWithNoTests"
+    },
+    "main": "./dist/index.js",
+    "module": "./dist/index.mjs",
+    "types": "./dist/index.d.ts",
+    "sideEffects": false,
+    "homepage": "https://ohmygoat.dev",
+    "repository": {
+        "type": "git",
+        "url": "git+https://github.com/goat-sdk/goat.git"
+    },
+    "license": "MIT",
+    "bugs": {
+        "url": "https://github.com/goat-sdk/goat/issues"
+    },
+    "keywords": ["ai", "agents", "web3"],
+    "dependencies": {
+        "@goat-sdk/core": "workspace:*",
+        "opengradient-sdk": "1.0.0",
+        "web3": "4.16.0",
+        "zod": "catalog:"
+    },
+    "peerDependencies": {
+        "@goat-sdk/core": "workspace:*"
+    }
+}

typescript/packages/plugins/opengradient/src/index.ts

@@ -0,0 +1,3 @@
+export * from "./opengradient.plugin";
+export * from "./parameters";
+export * from "./types";

typescript/packages/plugins/opengradient/src/opengradient.plugin.ts

@@ -0,0 +1,19 @@
+import { type Chain, PluginBase } from "@goat-sdk/core";
+import { OpengradientService } from "./opengradient.service";
+import { OpenGradientConfig } from "./types";
+
+export type OpengradientPluginCtorParams = {
+    config: OpenGradientConfig;
+};
+
+export class OpengradientPlugin extends PluginBase {
+    constructor({ config }: OpengradientPluginCtorParams) {
+        super("opengradient", [new OpengradientService(config)]);
+    }
+
+    supportsChain = (_chain: Chain) => true;
+}
+
+export function opengradient(params: OpengradientPluginCtorParams) {
+    return new OpengradientPlugin(params);
+}

typescript/packages/plugins/opengradient/src/opengradient.service.ts

@@ -0,0 +1,83 @@
+import { Tool } from "@goat-sdk/core";
+import { WalletClientBase } from "@goat-sdk/core";
+import { Client } from "opengradient-sdk";
+import { LLMChatParameters, LLMCompletionParameters, ModelInferenceParameters } from "./parameters";
+import { OpenGradientConfig } from "./types";
+
+export class OpengradientService {
+    private client: Client | null = null;
+
+    constructor(private config: OpenGradientConfig) {}
+
+    private getClient(clientConfig: OpenGradientConfig): Client {
+        if (!this.client) {
+            this.client = new Client({
+                privateKey: clientConfig.privateKey,
+            });
+        }
+        return this.client;
+    }
+
+    @Tool({
+        name: "opengradient_model_inference",
+        description: "Run inference on a machine learning model using OpenGradient",
+    })
+    async runModelInference(walletClient: WalletClientBase, parameters: ModelInferenceParameters) {
+        const client = this.getClient(this.config);
+        const [txHash, modelOutput] = await client.infer(
+            parameters.modelCid,
+            parameters.inferenceMode,
+            parameters.modelInput,
+        );
+
+        return {
+            transactionHash: txHash,
+            output: modelOutput,
+        };
+    }
+
+    @Tool({
+        name: "opengradient_llm_completion",
+        description: "Generate text completions using an LLM through OpenGradient",
+    })
+    async runLLMCompletion(walletClient: WalletClientBase, parameters: LLMCompletionParameters) {
+        const client = this.getClient(this.config);
+        const [txHash, completion] = await client.llmCompletion(
+            parameters.modelCid,
+            parameters.inferenceMode,
+            parameters.prompt,
+            parameters.maxTokens,
+            parameters.stopSequence,
+            parameters.temperature,
+        );
+
+        return {
+            transactionHash: txHash,
+            completion,
+        };
+    }
+
+    @Tool({
+        name: "opengradient_llm_chat",
+        description: "Interact with an LLM using a chat interface through OpenGradient",
+    })
+    async runLLMChat(walletClient: WalletClientBase, parameters: LLMChatParameters) {
+        const client = this.getClient(this.config);
+        const [txHash, finishReason, message] = await client.llmChat(
+            parameters.modelCid,
+            parameters.inferenceMode,
+            parameters.messages,
+            parameters.maxTokens,
+            parameters.stopSequence,
+            parameters.temperature,
+            parameters.tools,
+            parameters.toolChoice,
+        );
+
+        return {
+            transactionHash: txHash,
+            finishReason,
+            message,
+        };
+    }
+}

typescript/packages/plugins/opengradient/src/parameters.ts

@@ -0,0 +1,50 @@
+import { createToolParameters } from "@goat-sdk/core";
+import { z } from "zod";
+
+export class ModelInferenceParameters extends createToolParameters(
+    z.object({
+        modelCid: z.string().describe("CID of the model to run inference with"),
+        inferenceMode: z.number().describe("Inference mode to use (0 for VANILLA, 1 for TEE)"),
+        modelInput: z.record(z.any()).describe("Input to the model in the form of a JSON object"),
+    }),
+) {}
+
+export class LLMCompletionParameters extends createToolParameters(
+    z.object({
+        modelCid: z.string().describe("CID of the LLM model to use"),
+        inferenceMode: z.number().describe("Inference mode to use (0 for VANILLA, 1 for TEE)"),
+        prompt: z.string().describe("Text prompt for the LLM"),
+        maxTokens: z.number().default(100).describe("Maximum number of tokens to generate"),
+        stopSequence: z.array(z.string()).default([]).describe("Sequences that will stop generation if encountered"),
+        temperature: z.number().default(0).describe("Temperature for sampling (0.0 to 1.0)"),
+    }),
+) {}
+
+export class LLMChatParameters extends createToolParameters(
+    z.object({
+        modelCid: z.string().describe("CID of the LLM model to use"),
+        inferenceMode: z.number().describe("Inference mode to use (0 for VANILLA, 1 for TEE)"),
+        messages: z
+            .array(
+                z.object({
+                    role: z.enum(["system", "user", "assistant", "tool"]).describe("Role of the message sender"),
+                    content: z.string().describe("Content of the message"),
+                    toolCalls: z.array(z.any()).optional().describe("Tool calls made in this message"),
+                    toolCallId: z.string().optional().describe("ID of the tool call this message is responding to"),
+                    name: z.string().optional().describe("Name of the entity sending the message"),
+                }),
+            )
+            .describe("Array of conversation messages"),
+        maxTokens: z.number().default(100).describe("Maximum number of tokens to generate"),
+        stopSequence: z.array(z.string()).default([]).describe("Sequences that will stop generation if encountered"),
+        temperature: z.number().default(0).describe("Temperature for sampling (0.0 to 1.0)"),
+        tools: z.array(z.any()).default([]).describe("Tools available to the model"),
+        toolChoice: z.string().optional().describe("Tool choice strategy ('auto', 'none', or specific tool name)"),
+    }),
+) {}
+
+export class ClientConfigParameters extends createToolParameters(
+    z.object({
+        privateKey: z.string().describe("Private key for authentication with OpenGradient"),
+    }),
+) {}

typescript/packages/plugins/opengradient/src/types/index.ts

@@ -0,0 +1,3 @@
+export type OpenGradientConfig = {
+    privateKey: string;
+};

typescript/packages/plugins/opengradient/tsconfig.json

@@ -0,0 +1,6 @@
+{
+    "$schema": "https://json.schemastore.org/tsconfig",
+    "extends": "../../../tsconfig.base.json",
+    "include": ["src/**/*"],
+    "exclude": ["node_modules", "dist"]
+}

typescript/packages/plugins/opengradient/tsup.config.ts

@@ -0,0 +1,6 @@
+import { defineConfig } from "tsup";
+import { treeShakableConfig } from "../../../tsup.config.base";
+
+export default defineConfig({
+    ...treeShakableConfig,
+});

typescript/packages/plugins/opengradient/turbo.json

@@ -0,0 +1,11 @@
+{
+    "$schema": "https://turbo.build/schema.json",
+    "extends": ["//"],
+    "tasks": {
+        "build": {
+            "inputs": ["src/**", "tsup.config.ts", "!./**/*.test.{ts,tsx}", "tsconfig.json"],
+            "dependsOn": ["^build"],
+            "outputs": ["dist/**"]
+        }
+    }
+}