feat(coding-agent/tools): added image and binary file support to github file-read

- Switch github file-read tool to request json contents and handle base64 encoding. - Add support for parsing image attachments and detecting binary files from github contents.
2026-08-20 05:52:02 +02:00
parent bd5764aad4
commit 63b1146064
2 changed files with 159 additions and 9 deletions
@@ -6,9 +6,17 @@ import type {
 	AgentToolUpdateCallback,
 	ToolApprovalDecision,
 } from "@oh-my-pi/pi-agent-core";
-import { prompt, untilAborted } from "@oh-my-pi/pi-utils";
+import {
+	BINARY_SNIFF_BYTES,
+	formatBytes,
+	isProbablyBinaryHeader,
+	parseImageMetadata,
+	prompt,
+	untilAborted,
+} from "@oh-my-pi/pi-utils";
 import githubDescription from "../prompts/tools/github.md" with { type: "text" };
 import * as git from "../utils/git";
+import { loadImageAttachmentInput, webpExclusionForModel } from "../utils/image-loading";
 import type { ToolSession } from ".";
 import { buildTextResult, normalizeOptionalString, requireNonEmpty, resolveGitHubRepo } from "./gh-common";
 import { executePrCheckout, executePrCreate, executePrPush } from "./gh-pr-checkout";
@@ -23,6 +31,7 @@ import {
 import { executeRepoView } from "./gh-view";
 import type { OutputMeta } from "./output-meta";
 import { ToolError } from "./tool-errors";
+import { toolResult } from "./tool-result";

 export { parsePositiveDecimalInt, resolveDefaultRepoMemoized } from "./gh-common";
 export {
@@ -83,6 +92,34 @@ const githubSchema = type({

 type GithubInput = typeof githubSchema.infer;

+interface GitHubContentsFile {
+	type?: string;
+	encoding?: string;
+	size?: number;
+	content?: string;
+	html_url?: string | null;
+}
+
+type GitHubContentsResponse = GitHubContentsFile | GitHubContentsFile[];
+
+function isGitHubContentsFile(response: GitHubContentsResponse): response is GitHubContentsFile {
+	return !Array.isArray(response) && response.type === "file";
+}
+
+function buildBinaryFileReadResult(
+	filePath: string,
+	size: number,
+	sourceUrl: string,
+	repo: string,
+	branch: string | undefined,
+): AgentToolResult<GhToolDetails> {
+	return buildTextResult(
+		`[Cannot read binary file '${filePath}' (${formatBytes(size)}); not valid UTF-8 text. Open ${sourceUrl} to view it.]`,
+		sourceUrl,
+		{ repo, branch },
+	);
+}
+
 export interface GhToolDetails {
 	meta?: OutputMeta;
 	artifactId?: string;
@@ -232,15 +269,68 @@ async function executeFileRead(
 		"--method",
 		"GET",
 		"-H",
-		"Accept: application/vnd.github.raw+json",
+		"Accept: application/vnd.github+json",
+		"-H",
+		"Accept-Encoding: identity",
 	];
 	if (branch) {
 		args.push("-f", `ref=${branch}`);
 	}
-	const text = await git.github.text(session.cwd, args, signal, {
+	const response = await git.github.json<GitHubContentsResponse>(session.cwd, args, signal, {
 		repoProvided: true,
 		trimOutput: false,
 	});
-	const sourceUrl = `https://github.com/${repo}/blob/${encodeURIComponent(branch ?? "HEAD")}/${endpointPath}`;
-	return buildTextResult(text, sourceUrl, { repo, branch });
+	if (!isGitHubContentsFile(response)) {
+		throw new ToolError(`GitHub path '${filePath}' is not a file.`);
+	}
+
+	const fallbackSourceUrl = `https://github.com/${repo}/blob/${encodeURIComponent(branch ?? "HEAD")}/${endpointPath}`;
+	const sourceUrl = response.html_url || fallbackSourceUrl;
+	if (response.encoding !== "base64" || typeof response.content !== "string") {
+		const size =
+			typeof response.size === "number" && response.size >= 0 ? formatBytes(response.size) : "unknown size";
+		return buildTextResult(
+			`[GitHub did not return file bytes for '${filePath}' (${size}). Open ${sourceUrl} to view it.]`,
+			sourceUrl,
+			{ repo, branch },
+		);
+	}
+
+	const encoded = response.content.replaceAll(/\s/g, "");
+	const bytes = Buffer.from(encoded, "base64");
+	const imageMetadata = parseImageMetadata(bytes);
+	if (imageMetadata) {
+		const image = await loadImageAttachmentInput({
+			image: { type: "image", data: encoded, mimeType: imageMetadata.mimeType },
+			label: filePath,
+			uri: sourceUrl,
+			autoResize: session.settings.get("images.autoResize"),
+			excludeWebP: webpExclusionForModel(session.getActiveModel?.()),
+		});
+		if (image) {
+			const dimensions =
+				imageMetadata.width !== undefined && imageMetadata.height !== undefined
+					? `\nDimensions: ${imageMetadata.width}x${imageMetadata.height}`
+					: "";
+			return toolResult<GhToolDetails>({ repo, branch })
+				.content([
+					{
+						type: "text",
+						text: `Image file: ${filePath}\nMIME: ${image.mimeType}\nSize: ${formatBytes(bytes.byteLength)}${dimensions}`,
+					},
+					{ type: "image", data: image.data, mimeType: image.mimeType },
+				])
+				.sourceUrl(sourceUrl)
+				.done();
+		}
+	}
+
+	if (isProbablyBinaryHeader(bytes.subarray(0, BINARY_SNIFF_BYTES))) {
+		return buildBinaryFileReadResult(filePath, bytes.byteLength, sourceUrl, repo, branch);
+	}
+	try {
+		return buildTextResult(new TextDecoder("utf-8", { fatal: true }).decode(bytes), sourceUrl, { repo, branch });
+	} catch {
+		return buildBinaryFileReadResult(filePath, bytes.byteLength, sourceUrl, repo, branch);
+	}
 }
@@ -19,6 +19,9 @@ import * as git from "@oh-my-pi/pi-coding-agent/utils/git";
 import * as piUtils from "@oh-my-pi/pi-utils";
 import { $which, getAgentDir, hashPath, removeWithRetries, setAgentDir, WhichCachePolicy } from "@oh-my-pi/pi-utils";

+const TINY_PNG_BASE64 =
+	"iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mP8z8DwHwAFBQIAX8jx0gAAAABJRU5ErkJggg==";
+
 // Isolate every `git` invocation in this file from the developer's host
 // configuration. The fixture spawns dozens of git subprocesses against tiny
 // throwaway repos; any leak from `~/.gitconfig` or system config (LFS filters,
@@ -433,8 +436,14 @@ describe("github tool", () => {
 		expect(text).toContain("Topics: cli, github");
 	});

-	it("reads repository files through the GitHub contents API", async () => {
-		const textSpy = vi.spyOn(git.github, "text").mockResolvedValue('{"version":"16.3.11"}\n');
+	it("reads repository text through GitHub's JSON contents API", async () => {
+		const jsonSpy = vi.spyOn(git.github, "json").mockResolvedValue({
+			type: "file",
+			encoding: "base64",
+			size: 20,
+			content: Buffer.from('{"version":"16.3.11"}\n').toString("base64"),
+			html_url: "https://github.com/can1357/oh-my-pi/blob/main/packages/coding-agent/package.json",
+		});
 		const tool = new GithubTool(createSession());
 		const result = await tool.execute("file-read", {
 			op: "file_read",
@@ -445,7 +454,7 @@ describe("github tool", () => {
 		const text = result.content[0]?.type === "text" ? result.content[0].text : "";

 		expect(text).toBe('{"version":"16.3.11"}\n');
-		expect(textSpy).toHaveBeenCalledWith(
+		expect(jsonSpy).toHaveBeenCalledWith(
 			"/tmp/test",
 			[
 				"api",
@@ -453,7 +462,9 @@ describe("github tool", () => {
 				"--method",
 				"GET",
 				"-H",
-				"Accept: application/vnd.github.raw+json",
+				"Accept: application/vnd.github+json",
+				"-H",
+				"Accept-Encoding: identity",
 				"-f",
 				"ref=main",
 			],
@@ -462,6 +473,55 @@ describe("github tool", () => {
 		);
 	});

+	it("returns GitHub images as model image content", async () => {
+		vi.spyOn(git.github, "json").mockResolvedValue({
+			type: "file",
+			encoding: "base64",
+			size: Buffer.byteLength(TINY_PNG_BASE64, "base64"),
+			content: TINY_PNG_BASE64,
+			html_url: "https://github.com/anomalyco/opencode/blob/main/packages/web/src/assets/lander/screenshot.png",
+		});
+		const tool = new GithubTool(
+			createSession("/tmp/test", Settings.isolated({ "github.enabled": true, "images.autoResize": false })),
+		);
+		const result = await tool.execute("file-read", {
+			op: "file_read",
+			repo: "anomalyco/opencode",
+			branch: "main",
+			path: "packages/web/src/assets/lander/screenshot.png",
+		});
+		const text = result.content[0]?.type === "text" ? result.content[0].text : "";
+
+		expect(text).toContain("Image file: packages/web/src/assets/lander/screenshot.png");
+		expect(result.content).toContainEqual({
+			type: "image",
+			data: TINY_PNG_BASE64,
+			mimeType: "image/png",
+		});
+	});
+
+	it("identifies files GitHub returns without inline bytes", async () => {
+		const sourceUrl = "https://github.com/anomalyco/opencode/blob/main/packages/web/src/assets/lander/screenshot.png";
+		vi.spyOn(git.github, "json").mockResolvedValue({
+			type: "file",
+			encoding: "none",
+			size: 2 * 1024 * 1024,
+			html_url: sourceUrl,
+		});
+		const tool = new GithubTool(createSession());
+		const result = await tool.execute("file-read", {
+			op: "file_read",
+			repo: "anomalyco/opencode",
+			branch: "main",
+			path: "packages/web/src/assets/lander/screenshot.png",
+		});
+		const text = result.content[0]?.type === "text" ? result.content[0].text : "";
+
+		expect(text).toContain("GitHub did not return file bytes");
+		expect(text).toContain("packages/web/src/assets/lander/screenshot.png");
+		expect(text).toContain(sourceUrl);
+	});
+
 	it("creates a pull request via gh and renders the resulting summary", async () => {
 		const textCalls: string[][] = [];
 		const textSpy = vi.spyOn(git.github, "text").mockImplementation(async (_cwd, args) => {