import type { AgentTool, AgentToolResult } from "@mariozechner/pi-ai"; import { bashTool, codingTools, readTool } from "@mariozechner/pi-coding-agent"; import { type TSchema, Type } from "@sinclair/typebox"; import { detectMime } from "../media/mime.js"; import { startWebLoginWithQr, waitForWebLogin } from "../web/login-qr.js"; import { createClawdisTools } from "./clawdis-tools.js"; import { sanitizeToolResultImages } from "./tool-images.js"; // TODO(steipete): Remove this wrapper once pi-mono ships file-magic MIME detection // for `read` image payloads in `@mariozechner/pi-coding-agent` (then switch back to `codingTools` directly). type ToolContentBlock = AgentToolResult["content"][number]; type ImageContentBlock = Extract; type TextContentBlock = Extract; async function sniffMimeFromBase64( base64: string, ): Promise { const trimmed = base64.trim(); if (!trimmed) return undefined; const take = Math.min(256, trimmed.length); const sliceLen = take - (take % 4); if (sliceLen < 8) return undefined; try { const head = Buffer.from(trimmed.slice(0, sliceLen), "base64"); return await detectMime({ buffer: head }); } catch { return undefined; } } function rewriteReadImageHeader(text: string, mimeType: string): string { // pi-coding-agent uses: "Read image file [image/png]" if (text.startsWith("Read image file [") && text.endsWith("]")) { return `Read image file [${mimeType}]`; } return text; } async function normalizeReadImageResult( result: AgentToolResult, filePath: string, ): Promise> { const content = Array.isArray(result.content) ? result.content : []; const image = content.find( (b): b is ImageContentBlock => !!b && typeof b === "object" && (b as { type?: unknown }).type === "image" && typeof (b as { data?: unknown }).data === "string" && typeof (b as { mimeType?: unknown }).mimeType === "string", ); if (!image) return result; if (!image.data.trim()) { throw new Error(`read: image payload is empty (${filePath})`); } const sniffed = await sniffMimeFromBase64(image.data); if (!sniffed) return result; if (!sniffed.startsWith("image/")) { throw new Error( `read: file looks like ${sniffed} but was treated as ${image.mimeType} (${filePath})`, ); } if (sniffed === image.mimeType) return result; const nextContent = content.map((block) => { if ( block && typeof block === "object" && (block as { type?: unknown }).type === "image" ) { const b = block as ImageContentBlock & { mimeType: string }; return { ...b, mimeType: sniffed } satisfies ImageContentBlock; } if ( block && typeof block === "object" && (block as { type?: unknown }).type === "text" && typeof (block as { text?: unknown }).text === "string" ) { const b = block as TextContentBlock & { text: string }; return { ...b, text: rewriteReadImageHeader(b.text, sniffed), } satisfies TextContentBlock; } return block; }); return { ...result, content: nextContent }; } type AnyAgentTool = AgentTool; function normalizeToolParameters(tool: AnyAgentTool): AnyAgentTool { const schema = tool.parameters && typeof tool.parameters === "object" ? (tool.parameters as Record) : undefined; if (!schema) return tool; if ("type" in schema && "properties" in schema) return tool; if (!Array.isArray(schema.anyOf)) return tool; return { ...tool, parameters: { ...schema, type: "object", properties: schema.properties ?? {}, additionalProperties: "additionalProperties" in schema ? schema.additionalProperties : true, } as TSchema, }; } function createWhatsAppLoginTool(): AnyAgentTool { return { label: "WhatsApp Login", name: "whatsapp_login", description: "Generate a WhatsApp QR code for linking, or wait for the scan to complete.", parameters: Type.Object({ action: Type.Union([Type.Literal("start"), Type.Literal("wait")]), timeoutMs: Type.Optional(Type.Number()), force: Type.Optional(Type.Boolean()), }), execute: async (_toolCallId, args) => { const action = (args as { action?: string })?.action ?? "start"; if (action === "wait") { const result = await waitForWebLogin({ timeoutMs: typeof (args as { timeoutMs?: unknown }).timeoutMs === "number" ? (args as { timeoutMs?: number }).timeoutMs : undefined, }); return { content: [{ type: "text", text: result.message }], details: { connected: result.connected }, }; } const result = await startWebLoginWithQr({ timeoutMs: typeof (args as { timeoutMs?: unknown }).timeoutMs === "number" ? (args as { timeoutMs?: number }).timeoutMs : undefined, force: typeof (args as { force?: unknown }).force === "boolean" ? (args as { force?: boolean }).force : false, }); if (!result.qrDataUrl) { return { content: [ { type: "text", text: result.message, }, ], details: { qr: false }, }; } const text = [ result.message, "", "Open WhatsApp → Linked Devices and scan:", "", `![whatsapp-qr](${result.qrDataUrl})`, ].join("\n"); return { content: [{ type: "text", text }], details: { qr: true }, }; }, }; } function createClawdisReadTool(base: AnyAgentTool): AnyAgentTool { return { ...base, execute: async (toolCallId, params, signal) => { const result = (await base.execute( toolCallId, params, signal, )) as AgentToolResult; const record = params && typeof params === "object" ? (params as Record) : undefined; const filePath = typeof record?.path === "string" ? String(record.path) : ""; const normalized = await normalizeReadImageResult(result, filePath); return sanitizeToolResultImages(normalized, `read:${filePath}`); }, }; } function createClawdisBashTool(base: AnyAgentTool): AnyAgentTool { return { ...base, execute: async (toolCallId, params, signal) => { const result = (await base.execute( toolCallId, params, signal, )) as AgentToolResult; return sanitizeToolResultImages(result, "bash"); }, }; } export function createClawdisCodingTools(): AnyAgentTool[] { const base = (codingTools as unknown as AnyAgentTool[]).map((tool) => tool.name === readTool.name ? createClawdisReadTool(tool) : tool.name === bashTool.name ? createClawdisBashTool(tool) : (tool as AnyAgentTool), ); return [...base, createWhatsAppLoginTool(), ...createClawdisTools()].map( normalizeToolParameters, ); }