QVAC-17304 feat[api]: add img2img support to SDK diffusion API

[maxim-smotrov]

🎯 What problem does this PR solve?

• The SDK diffusion() API only supported txt2img — users could not use the img2img capability of the underlying diffusion add-on.

📝 How does it solve it?

• Add init_image (Uint8Array on the client, base64 on the wire), strength, and img_cfg_scale to diffusionRequestSchema.
• Client diffusion() accepts init_image: Uint8Array and base64-encodes it before sending; cross-runtime encodeBase64 helper added in utils/encoding.ts (uses btoa so it works in Node, Bare, and React Native/Hermes).
• Server bare plugin decodes init_image to a Buffer and forwards init_image, strength, and img_cfg_scale to @qvac/diffusion-cpp.
• Bump @qvac/diffusion-cpp to ^0.2.0 (img2img-capable).

🧪 How was it tested?

• Unit tests added in test/unit/sdcpp-plugin.test.ts
• New runnable example examples/diffusion-img2img.ts demonstrating end-to-end usage.

🔌 API Changes

Here's an example usage of the new img2img API.

import { loadModel, diffusion, SD_V2_1_1B_Q8_0 } from "@qvac/sdk";
import fs from "fs";

const modelId = await loadModel({ modelSrc: SD_V2_1_1B_Q8_0, modelType: "diffusion" });

// SD / SDXL — SDEdit
const initImage = new Uint8Array(fs.readFileSync("input.png"));
const { outputs } = diffusion({
  modelId,
  prompt: "oil painting style, vibrant colors",
  init_image: initImage,
  strength: 0.5, // 0 = keep source, 1 = ignore source
});

// FLUX.2 — in-context conditioning
// NOTE: requires `prediction: "flux2_flow"` set on the model config at loadModel time.
// `strength` is ignored on this path.
const { outputs: fluxOutputs } = diffusion({
  modelId,
  prompt: "turn into watercolor",
  init_image: initImage,
});

const buffers = await outputs;
fs.writeFileSync("out.png", buffers[0]!);

[lauripiisang]

can't be merged - test changes will crash mobile

checks

[maxim-smotrov]

/review

[github-actions]

Tier-based Approval Status

**PR Tier:** TIER1

**Current Status:** ✅ APPROVED

**Requirements:**
- 1 Team Member approval ✅ (2/1)
- 1 Team Lead OR Management approval ✅ (1/1)



---
*This comment is automatically updated when reviews change.*