import { afterEach, describe, expect, test } from "bun:test" import { mkdirSync, mkdtempSync, writeFileSync } from "node:fs" import { tmpdir } from "node:os" import { join } from "node:path" import { createApp } from "../src/app.ts" import type { AskReply } from "../src/bus/index.ts" import { paths } from "../src/config/paths.ts" import { attachmentsFor } from "../src/project/attach.ts" import { delta, fakeProvider, toolCall, type Fake } from "./fake-provider.ts" // A 1×1 PNG. const PNG = Buffer.from("iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mP8z8BQDwAEhQGAhKmMIQAAAABJRU5ErkJggg==", "base64") let fake: Fake | undefined afterEach(() => fake?.stop()) function project() { const root = mkdtempSync(join(tmpdir(), "ph-img-")) writeFileSync(join(root, "shot.png"), PNG) return root } describe("images", () => { test("@image: attached as an image with vision; without, the model is told why not", () => { const root = project() const ctx = { root, cwd: root, readFiles: new Set(), fileStamps: new Map() } const [withVision] = attachmentsFor("look at @shot.png", ctx, true) expect(withVision!.image).toEqual({ type: "image", mime: "image/png", data: PNG.toString("base64") }) const [without] = attachmentsFor("look at @shot.png", ctx, false) expect(without!.image).toBeUndefined() expect(without!.text).toContain("not attached: this model has no vision") }) function setup(vision: boolean, script: Parameters[0]) { fake = fakeProvider(script) mkdirSync(paths.config, { recursive: true }) writeFileSync(join(paths.config, "connections.yaml"), `connections:\n f:\n dialect: openai-chat\n base_url: ${fake.url}\n models:\n m: { vision: ${vision} }\n`, { mode: 0o600 }) writeFileSync(join(paths.config, "config.yaml"), "model: f/m\n") return createApp({ cwd: project(), mode: "edit", store: false, asker: { ask: async (): Promise => ({ kind: "once" }) } }) } test("view_image: offered only with vision; the image follows the tool results as a user turn", async () => { const blind = setup(false, [{ chunks: [delta({ content: "ok" })] }]) await blind.engine.prompt("hi") expect(fake!.requests[0].tools.map((t: any) => t.function.name)).not.toContain("view_image") fake!.stop() const app = setup(true, [{ chunks: [toolCall(0, "c1", "view_image", '{"path":"shot.png"}')] }, { chunks: [delta({ content: "A single pixel." })] }]) await app.engine.prompt("what is in shot.png?") expect(fake!.requests[0].tools.map((t: any) => t.function.name)).toContain("view_image") const msgs = fake!.requests[1].messages expect(msgs.at(-2)).toMatchObject({ role: "tool", tool_call_id: "c1" }) expect(msgs.at(-1).role).toBe("user") expect(msgs.at(-1).content[1]).toEqual({ type: "image_url", image_url: { url: `data:image/png;base64,${PNG.toString("base64")}` } }) }) test("a prompt with @image reaches an openai-chat vision model as image_url", async () => { const app = setup(true, [{ chunks: [delta({ content: "ok" })] }]) const { attachmentsFor: att } = await import("../src/project/attach.ts") const atts = att("see @shot.png", app.engine.o.toolCtx, true) await app.engine.prompt("see @shot.png", atts.flatMap((a) => (a.image ? [a.text, a.image] : [a.text]))) const content = fake!.requests[0].messages.at(-1).content expect(content.map((p: any) => p.type)).toEqual(["text", "text", "image_url"]) }) })