browser_shot

-- browser_shot: drive a disposable headless Chromium browser (Chrome, Edge or
-- Chromium) through the DevTools protocol and save PNG screenshots.
--
-- One call = one browser: a fresh temporary profile, the steps, the
-- screenshots, then Browser.close and girl's own preview stop (the browser's
-- tree by recorded identity, the profile removed) - also when a step fails.
-- Nothing outlives the call. The browser is started by girl's
-- platform.preview_chrome (a kill-on-close job, the profile under the session
-- scratch, never in the workspace); the DevTools protocol runs over a small RFC
-- 6455 WebSocket client on LuaXE's `socket` module, so the tool needs no Node,
-- no npm and no driver binary: only an installed Chromium-family browser.
--
-- Effects (all declared by the `browser.drive` permission and listed in the
-- approval before anything runs): starts a local browser process, loads the
-- given URL(s), performs the steps (clicks and `eval` can change what the page
-- talks to), writes PNG files to the session scratch (or `out_dir` inside the
-- workspace).
local fs = require("platform.fs")
local path = require("platform.path")
local json = require("platform.json")
local Preview = require("platform.preview_chrome")

local DEFAULT_WIDTH, DEFAULT_HEIGHT = 1440, 900
local STEP_TIMEOUT_MS = 15000
local LAUNCH_TIMEOUT_MS = 20000

-- ── small helpers ──────────────────────────────────────────────────────────
local function now_ms(socket) return socket.gettime() * 1000 end
local function js_string(value) return json.encode(tostring(value)) end

local function plain(value) -- json.null -> nil, recursively
    if value == json.null then return nil end
    if type(value) ~= "table" then return value end
    local out = {}
    for k, v in pairs(value) do out[k] = plain(v) end
    return out
end

local function parse_size(text)
    if text == nil then return nil end
    local w, h = tostring(text):match("^%s*(%d+)%s*[xX]%s*(%d+)%s*$")
    w, h = tonumber(w), tonumber(h)
    if not w or not h or w < 100 or h < 100 or w > 7680 or h > 7680 then
        error("size must look like 1440x900 (100..7680 per side), got " .. tostring(text), 0)
    end
    return w, h
end

local SCHEMES = { light = true, dark = true, ["no-preference"] = true }

local function check_url(url, label)
    if type(url) ~= "string" or url == "" then error((label or "url") .. " is required", 0) end
    local scheme = url:match("^(%a[%w+.-]*):")
    scheme = scheme and scheme:lower()
    if scheme ~= "http" and scheme ~= "https" and scheme ~= "file" and scheme ~= "about" then
        error((label or "url") .. " must be http, https, file or about: " .. url, 0)
    end
    return url
end

-- ── base64 (screenshots arrive base64-encoded) ─────────────────────────────
local B64 = {}
do
    local alphabet = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/"
    for i = 1, #alphabet do B64[alphabet:byte(i)] = i - 1 end
end
local function base64_decode(text)
    text = text:gsub("[^%w%+/]", "")
    local out, n = {}, 0
    local chunk = {}
    for i = 1, #text, 4 do
        local a, b, c, d = text:byte(i, i + 3)
        local v = B64[a] * 262144 + (b and B64[b] or 0) * 4096 + (c and B64[c] or 0) * 64 + (d and B64[d] or 0)
        local x = math.floor(v / 65536)
        local y = math.floor(v / 256) % 256
        local z = v % 256
        n = n + 1
        if d then chunk[n] = string.char(x, y, z)
        elseif c then chunk[n] = string.char(x, y)
        else chunk[n] = string.char(x) end
        if n == 4096 then out[#out + 1] = table.concat(chunk); chunk = {}; n = 0 end
    end
    out[#out + 1] = table.concat(chunk, "", 1, n)
    return table.concat(out)
end

local function png_size(bytes)
    if #bytes < 24 or bytes:sub(1, 8) ~= "\137PNG\r\n\26\n" then return nil end
    local function u32(i) local a, b, c, d = bytes:byte(i, i + 3) return ((a * 256 + b) * 256 + c) * 256 + d end
    return u32(17), u32(21)
end

-- ── where the files go ─────────────────────────────────────────────────────
-- The session scratch (girl's tool context), else the configured session dir,
-- else the OS temp directory.
local function scratch_of(context)
    local scratch = type(context) == "table" and context.scratch
    if type(scratch) == "string" and scratch ~= "" then return scratch end
    local config = type(context) == "table" and context.config
    local spec = type(config) == "table" and config.scratch
    if type(spec) == "table" and type(spec.session_dir) == "string" and spec.session_dir ~= "" then
        return spec.session_dir
    end
    return nil
end

local function default_out_dir(context)
    local scratch = scratch_of(context)
    if scratch then return path.join(scratch, "browser_shot") end
    local tmp = os.getenv("TEMP") or os.getenv("TMP") or os.getenv("TMPDIR") or "/tmp"
    return path.join(tmp, "girl-browser_shot")
end

-- ── a minimal WebSocket client (RFC 6455, text frames, client-masked) ──────
local function ws_connect(socket, port, ws_path)
    local conn = assert(socket.tcp())
    conn:settimeout(10)
    local ok, err = conn:connect("127.0.0.1", port)
    if not ok then error("cannot reach the browser's DevTools port " .. port .. ": " .. tostring(err), 0) end
    local key = {}
    for i = 1, 16 do key[i] = string.char(math.random(0, 255)) end
    local function b64(s) -- encoder for the 16-byte key only
        local chars = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/"
        local out = {}
        for i = 1, #s, 3 do
            local a, b, c = s:byte(i, i + 2)
            local v = a * 65536 + (b or 0) * 256 + (c or 0)
            local function at(shift) local idx = math.floor(v / shift) % 64 + 1 return chars:sub(idx, idx) end
            out[#out + 1] = at(262144) .. at(4096) .. (b and at(64) or "=") .. (c and at(1) or "=")
        end
        return table.concat(out)
    end
    conn:send("GET " .. ws_path .. " HTTP/1.1\r\nHost: 127.0.0.1:" .. port .. "\r\nUpgrade: websocket\r\n"
        .. "Connection: Upgrade\r\nSec-WebSocket-Key: " .. b64(table.concat(key)) .. "\r\nSec-WebSocket-Version: 13\r\n\r\n")
    local status = conn:receive("*l")
    if not status or not status:find(" 101 ", 1, true) then
        conn:close()
        error("the browser refused the DevTools WebSocket: " .. tostring(status), 0)
    end
    repeat local line = conn:receive("*l") until line == nil or line == ""
    local bit = require("bit")
    local ws = { conn = conn }
    function ws.send(text)
        local n = #text
        local head
        if n < 126 then head = string.char(0x81, 0x80 + n)
        elseif n < 65536 then head = string.char(0x81, 0x80 + 126, math.floor(n / 256), n % 256)
        else
            local bytes = {}
            local rest = n
            for i = 8, 1, -1 do bytes[i] = rest % 256; rest = math.floor(rest / 256) end
            head = string.char(0x81, 0x80 + 127, unpack(bytes))
        end
        local m = { math.random(0, 255), math.random(0, 255), math.random(0, 255), math.random(0, 255) }
        local out, chunk, c = { head, string.char(m[1], m[2], m[3], m[4]) }, {}, 0
        for i = 1, n do
            c = c + 1
            chunk[c] = string.char(bit.bxor(text:byte(i), m[(i - 1) % 4 + 1]))
            if c == 8192 then out[#out + 1] = table.concat(chunk); chunk = {}; c = 0 end
        end
        out[#out + 1] = table.concat(chunk, "", 1, c)
        local sent, err = conn:send(table.concat(out))
        if not sent then error("the DevTools connection dropped while sending: " .. tostring(err), 0) end
    end
    -- One whole message, or nil plus "timeout"/"closed".
    function ws.receive(timeout_s)
        conn:settimeout(timeout_s)
        local parts = {}
        while true do
            local h, err = conn:receive(2)
            if not h then return nil, err end
            conn:settimeout(10) -- the rest of a frame that has started arrives promptly
            local b1, b2 = h:byte(1, 2)
            local fin, opcode = b1 >= 128, b1 % 16
            local len = b2 % 128
            if len == 126 then
                local e = assert(conn:receive(2)); len = e:byte(1) * 256 + e:byte(2)
            elseif len == 127 then
                local e = assert(conn:receive(8)); len = 0
                for i = 1, 8 do len = len * 256 + e:byte(i) end
            end
            local payload = len > 0 and conn:receive(len) or ""
            if payload == nil then return nil, "closed" end
            if opcode == 8 then return nil, "closed"
            elseif opcode == 9 then -- ping -> pong (masked, same payload)
                local mask = string.char(0, 0, 0, 0)
                conn:send(string.char(0x8A, 0x80 + #payload) .. mask .. payload)
            elseif opcode == 1 or opcode == 0 or opcode == 2 then
                parts[#parts + 1] = payload
                if fin then return table.concat(parts) end
            end
        end
    end
    function ws.close() pcall(function() conn:close() end) end
    return ws
end

-- ── the DevTools session ──────────────────────────────────────────────────
local function http_get(socket, port, route)
    local conn = assert(socket.tcp())
    conn:settimeout(5)
    if not conn:connect("127.0.0.1", port) then conn:close() return nil end
    conn:send("GET " .. route .. " HTTP/1.1\r\nHost: 127.0.0.1:" .. port .. "\r\nConnection: close\r\n\r\n")
    local parts = {}
    while true do
        local chunk, _, partial = conn:receive(8192)
        parts[#parts + 1] = chunk or partial
        if not chunk then break end
    end
    conn:close()
    local text = table.concat(parts)
    return text:match("\r\n\r\n(.*)$")
end

local function new_cdp(socket, ws)
    local cdp = { ws = ws, next_id = 0, console = {}, events = {} }
    local function note(event)
        local params = event.params or {}
        if event.method == "Runtime.consoleAPICalled" and (params.type == "error" or params.type == "assert" or params.type == "warning") then
            local words = {}
            for _, arg in ipairs(params.args or {}) do
                local v = arg.value
                if v == nil or v == json.null then v = arg.description or arg.type end
                words[#words + 1] = type(v) == "table" and json.encode(plain(v)) or tostring(v)
            end
            cdp.console[#cdp.console + 1] = params.type .. ": " .. table.concat(words, " ")
        elseif event.method == "Runtime.exceptionThrown" then
            local d = params.exceptionDetails or {}
            local ex = d.exception or {}
            cdp.console[#cdp.console + 1] = "exception: " .. tostring(ex.description or d.text or "?")
                .. (d.url and d.url ~= json.null and (" @ " .. tostring(d.url) .. ":" .. tostring(d.lineNumber)) or "")
        elseif event.method == "Log.entryAdded" then
            local e = params.entry or {}
            if e.level == "error" then
                cdp.console[#cdp.console + 1] = "log: " .. tostring(e.text) .. (e.url and e.url ~= json.null and (" @ " .. tostring(e.url)) or "")
            end
        end
        cdp.events[#cdp.events + 1] = event.method
        if #cdp.events > 200 then table.remove(cdp.events, 1) end
    end
    -- Send one command and wait for ITS reply; events seen meanwhile are noted.
    function cdp.call(method, params, session, timeout_ms)
        cdp.next_id = cdp.next_id + 1
        local id = cdp.next_id
        local message = { id = id, method = method, params = params or json.object() }
        if session then message.sessionId = session end
        ws.send(json.encode(message))
        local deadline = now_ms(socket) + (timeout_ms or STEP_TIMEOUT_MS)
        while true do
            local left = (deadline - now_ms(socket)) / 1000
            if left <= 0 then error(method .. " timed out", 0) end
            local text, err = ws.receive(left)
            if not text then
                if err == "timeout" then error(method .. " timed out", 0) end
                error("the browser closed the DevTools connection during " .. method, 0)
            end
            local reply = json.decode(text)
            if reply.id == id then
                if reply.error and reply.error ~= json.null then
                    error(method .. ": " .. tostring(reply.error.message or json.encode(plain(reply.error))), 0)
                end
                return reply.result or {}
            elseif reply.method then
                note(reply)
                if cdp.waiting and reply.method == cdp.waiting then cdp.seen = true end
            end
        end
    end
    -- Drain events until `method` arrives (or time runs out). Returns true if seen.
    function cdp.wait_event(method, timeout_ms)
        if cdp.seen and cdp.waiting == method then cdp.seen = false; cdp.waiting = nil; return true end
        local deadline = now_ms(socket) + timeout_ms
        while true do
            local left = (deadline - now_ms(socket)) / 1000
            if left <= 0 then return false end
            local text = ws.receive(left)
            if not text then return false end
            local event = json.decode(text)
            if event.method then
                note(event)
                if event.method == method then return true end
            end
        end
    end
    return cdp
end

-- ── steps ──────────────────────────────────────────────────────────────────
local KEYS = {
    Enter = { code = "Enter", vk = 13, text = "\r" }, Tab = { code = "Tab", vk = 9 },
    Escape = { code = "Escape", vk = 27 }, Backspace = { code = "Backspace", vk = 8 },
    Delete = { code = "Delete", vk = 46 }, Space = { code = "Space", vk = 32, text = " ", key = " " },
    ArrowUp = { code = "ArrowUp", vk = 38 }, ArrowDown = { code = "ArrowDown", vk = 40 },
    ArrowLeft = { code = "ArrowLeft", vk = 37 }, ArrowRight = { code = "ArrowRight", vk = 39 },
    Home = { code = "Home", vk = 36 }, End = { code = "End", vk = 35 },
    PageUp = { code = "PageUp", vk = 33 }, PageDown = { code = "PageDown", vk = 34 },
}

local STEP_KINDS = { "goto", "wait", "wait_js", "click", "hover", "type", "key", "eval", "scroll", "size", "color_scheme", "shot" }

-- Normalize one step to { kind, ... } and validate it; never touches the browser.
local function parse_step(step, index)
    if type(step) ~= "table" then error("step " .. index .. " must be an object", 0) end
    local kind
    for _, name in ipairs(STEP_KINDS) do
        if step[name] ~= nil then
            if kind then error("step " .. index .. " names two actions (" .. kind .. ", " .. name .. ")", 0) end
            kind = name
        end
    end
    if not kind then error("step " .. index .. " has no action (one of: " .. table.concat(STEP_KINDS, ", ") .. ")", 0) end
    local v = step[kind]
    local out = { kind = kind, timeout_ms = tonumber(step.timeout_ms) or STEP_TIMEOUT_MS }
    if kind == "goto" then out.url = check_url(v, "step " .. index .. " goto")
    elseif kind == "wait" then
        if type(v) == "number" then out.ms = math.max(0, math.min(v, 60000))
        elseif type(v) == "string" and v ~= "" then out.selector = v
        else error("step " .. index .. " wait takes a selector or milliseconds", 0) end
    elseif kind == "wait_js" or kind == "eval" then
        if type(v) ~= "string" or v == "" then error("step " .. index .. " " .. kind .. " takes a JavaScript expression", 0) end
        out.js = v
    elseif kind == "click" or kind == "hover" or kind == "scroll" then
        if type(v) ~= "string" or v == "" then error("step " .. index .. " " .. kind .. " takes a CSS selector", 0) end
        out.selector = v
    elseif kind == "type" then
        if type(v) ~= "table" or type(v[1]) ~= "string" or v[2] == nil then
            error("step " .. index .. ' type takes ["selector", "text"]', 0)
        end
        out.selector, out.text = v[1], tostring(v[2])
        out.clear = step.clear ~= false
    elseif kind == "key" then
        if type(v) ~= "string" or v == "" then error("step " .. index .. " key takes a key name such as Enter", 0) end
        out.key = v
    elseif kind == "size" then out.width, out.height = parse_size(v)
    elseif kind == "color_scheme" then
        if not SCHEMES[v] then error("step " .. index .. " color_scheme is light, dark or no-preference", 0) end
        out.scheme = v
    elseif kind == "shot" then
        local name = type(v) == "string" and v or ""
        if name == "" or v == true then name = "shot-" .. index .. ".png" end
        name = path.basename(name)
        if not name:lower():match("%.png$") then name = name .. ".png" end
        if name:find("[^%w%._%-]") then error("step " .. index .. " shot name may use letters, digits, . _ -", 0) end
        out.name = name
        out.selector = type(step.selector) == "string" and step.selector ~= "" and step.selector or nil
        out.full_page = step.full_page == true
        out.padding = tonumber(step.padding) or 0
    end
    return out
end

local function describe(step)
    local k = step.kind
    if k == "goto" then return "goto " .. step.url end
    if k == "wait" then return step.ms and ("wait " .. step.ms .. " ms") or ("wait for " .. step.selector) end
    if k == "wait_js" then return "wait until JS: " .. step.js end
    if k == "eval" then return "eval JS: " .. step.js end
    if k == "type" then return "type " .. json.encode(step.text) .. " into " .. step.selector end
    if k == "key" then return "press " .. step.key end
    if k == "size" then return "viewport " .. step.width .. "x" .. step.height end
    if k == "color_scheme" then return "prefers-color-scheme: " .. step.scheme end
    if k == "shot" then
        return "screenshot " .. step.name .. (step.selector and (" of " .. step.selector) or "") .. (step.full_page and " (full page)" or "")
    end
    return k .. " " .. tostring(step.selector)
end

-- The whole request, resolved and validated, without touching the browser.
-- `pure` (the approval prompt) also reads no file: a scenario is named, not
-- expanded, there.
local function plan(args, context, pure)
    local scenario = {}
    if type(args.scenario) == "string" and args.scenario ~= "" then
        if args.steps ~= nil then error("pass steps OR a scenario file, not both", 0) end
        if pure then
            scenario = { steps = {}, url = args.url or "about:blank" }
        else
            local text = fs.read(context.workspace, args.scenario)
            local ok, decoded = pcall(json.decode, text)
            if not ok or type(decoded) ~= "table" then error("scenario " .. args.scenario .. " is not a JSON object", 0) end
            scenario = plain(decoded)
        end
    end
    local url = check_url(args.url or scenario.url, "url")
    local width, height = parse_size(args.size or scenario.size)
    width, height = width or DEFAULT_WIDTH, height or DEFAULT_HEIGHT
    local scheme = args.color_scheme or scenario.color_scheme
    if scheme ~= nil and not SCHEMES[scheme] then error("color_scheme is light, dark or no-preference", 0) end
    local raw = args.steps or scenario.steps or {}
    if type(raw) ~= "table" then error("steps must be an array", 0) end
    local steps = {}
    for i, step in ipairs(raw) do steps[i] = parse_step(step, i) end
    local has_shot = false
    for _, s in ipairs(steps) do if s.kind == "shot" then has_shot = true end end
    if not has_shot then steps[#steps + 1] = parse_step({ shot = "page.png" }, #steps + 1) end
    local out_dir
    if type(args.out_dir) == "string" and args.out_dir ~= "" then
        local resolved, why = path.resolve_inside(context.workspace, args.out_dir)
        if not resolved then error("out_dir must be inside the workspace (" .. tostring(why) .. ")", 0) end
        out_dir = resolved
    else
        out_dir = default_out_dir(context)
    end
    return {
        url = url, width = width, height = height, scheme = scheme, steps = steps,
        out_dir = out_dir, browser = args.browser, settle_ms = tonumber(args.settle_ms) or 150,
    }
end

-- JavaScript run in the page. Each returns a value by JSON.
local function js_rect(selector)
    return "(()=>{const el=document.querySelector(" .. js_string(selector) .. ");if(!el)return null;"
        .. "el.scrollIntoView({block:'center',inline:'center'});const r=el.getBoundingClientRect();"
        .. "return {x:r.left,y:r.top,w:r.width,h:r.height,sx:window.scrollX,sy:window.scrollY}})()"
end

local function run_steps(cdp, session, req, socket, shots, results, context)
    local function call(method, params, timeout) return cdp.call(method, params, session, timeout) end
    local function eval(js, timeout)
        local r = call("Runtime.evaluate", { expression = js, awaitPromise = true, returnByValue = true }, timeout)
        if r.exceptionDetails and r.exceptionDetails ~= json.null then
            local ex = r.exceptionDetails.exception or {}
            error("JavaScript threw: " .. tostring(ex.description or r.exceptionDetails.text), 0)
        end
        return plain(r.result and r.result.value)
    end
    local function settle()
        -- two animation frames, then the settle pause: a click's repaint lands
        eval("new Promise(r=>requestAnimationFrame(()=>requestAnimationFrame(()=>r(true))))", 5000)
        if req.settle_ms > 0 then socket.sleep(req.settle_ms / 1000) end
    end
    local function poll(js, timeout_ms, what)
        local deadline = now_ms(socket) + timeout_ms
        while true do
            local value = eval(js, timeout_ms)
            if value then return value end
            if now_ms(socket) > deadline then error("timed out after " .. timeout_ms .. " ms waiting for " .. what, 0) end
            socket.sleep(0.1)
        end
    end
    local function rect_of(selector, timeout_ms)
        return poll(js_rect(selector), timeout_ms, selector)
    end
    local function mouse(kind, x, y)
        call("Input.dispatchMouseEvent", { type = kind, x = x, y = y, button = "left", clickCount = 1 })
    end
    local function navigate(url, timeout_ms)
        cdp.waiting = "Page.loadEventFired"; cdp.seen = false
        local r = call("Page.navigate", { url = url }, timeout_ms)
        if r.errorText and r.errorText ~= json.null then error("could not load " .. url .. ": " .. tostring(r.errorText), 0) end
        if not cdp.wait_event("Page.loadEventFired", timeout_ms) then error("timed out loading " .. url, 0) end
        cdp.waiting = nil
    end
    local function viewport(w, h)
        call("Emulation.setDeviceMetricsOverride", { width = w, height = h, deviceScaleFactor = 1, mobile = false })
    end

    viewport(req.width, req.height)
    if req.scheme then
        call("Emulation.setEmulatedMedia", { features = { { name = "prefers-color-scheme", value = req.scheme } } })
    end
    navigate(req.url, STEP_TIMEOUT_MS * 2)
    settle()

    for i, step in ipairs(req.steps) do
        local k, t = step.kind, step.timeout_ms
        local ok, err = pcall(function()
            if k == "goto" then navigate(step.url, t)
            elseif k == "wait" then
                if step.ms then socket.sleep(step.ms / 1000)
                else rect_of(step.selector, t) end
            elseif k == "wait_js" then poll(step.js, t, "JS " .. step.js)
            elseif k == "eval" then results[#results + 1] = { step = i, value = eval(step.js, t) }
            elseif k == "click" or k == "hover" then
                local r = rect_of(step.selector, t)
                local x, y = r.x + r.w / 2, r.y + r.h / 2
                mouse("mouseMoved", x, y)
                if k == "click" then mouse("mousePressed", x, y); mouse("mouseReleased", x, y) end
            elseif k == "scroll" then rect_of(step.selector, t)
            elseif k == "type" then
                rect_of(step.selector, t)
                eval("(()=>{const el=document.querySelector(" .. js_string(step.selector) .. ");el.focus();"
                    .. (step.clear and "if('value' in el){el.value='';el.dispatchEvent(new Event('input',{bubbles:true}))}" or "")
                    .. "return true})()")
                call("Input.insertText", { text = step.text })
            elseif k == "key" then
                local spec = KEYS[step.key]
                if not spec and #step.key == 1 then spec = { key = step.key, code = "", vk = step.key:upper():byte(), text = step.key } end
                if not spec then error("unknown key " .. step.key, 0) end
                local base = { key = spec.key or step.key, code = spec.code, windowsVirtualKeyCode = spec.vk, nativeVirtualKeyCode = spec.vk }
                local down = { type = spec.text and "keyDown" or "rawKeyDown", text = spec.text }
                for kk, vv in pairs(base) do down[kk] = vv end
                call("Input.dispatchKeyEvent", down)
                local up = { type = "keyUp" }
                for kk, vv in pairs(base) do up[kk] = vv end
                call("Input.dispatchKeyEvent", up)
            elseif k == "size" then viewport(step.width, step.height)
            elseif k == "color_scheme" then
                call("Emulation.setEmulatedMedia", { features = { { name = "prefers-color-scheme", value = step.scheme } } })
            elseif k == "shot" then
                local params = { format = "png", captureBeyondViewport = step.full_page or step.selector ~= nil }
                if step.selector then
                    local r = rect_of(step.selector, t)
                    settle()
                    r = eval(js_rect(step.selector))
                    local pad = step.padding
                    params.clip = { x = math.max(0, r.x + r.sx - pad), y = math.max(0, r.y + r.sy - pad),
                        width = math.max(1, r.w + 2 * pad), height = math.max(1, r.h + 2 * pad), scale = 1 }
                elseif step.full_page then
                    local m = call("Page.getLayoutMetrics")
                    local size = m.cssContentSize or m.contentSize
                    params.clip = { x = 0, y = 0, width = size.width, height = size.height, scale = 1 }
                end
                local shot = call("Page.captureScreenshot", params, math.max(t, 30000))
                local bytes = base64_decode(shot.data or "")
                if not fs.ensure_dir(req.out_dir) then error("cannot create " .. req.out_dir, 0) end
                local file = path.join(req.out_dir, step.name)
                local handle = io.open(file, "wb")
                if not handle then error("cannot write " .. file, 0) end
                handle:write(bytes)
                handle:close()
                local w, h = png_size(bytes)
                shots[#shots + 1] = { name = step.name, path = file, width = w, height = h }
            end
            if k ~= "wait" and k ~= "shot" and k ~= "eval" then settle() end
        end)
        if not ok then return i, err end
    end
    return nil
end

-- ── the tool ──────────────────────────────────────────────────────────────
local function execute(args, context)
    local req = plan(args, context)
    local socket = import("socket")
    math.randomseed(os.time() + math.floor((socket.gettime() % 1) * 1e6))
    -- girl's preview launcher: the profile under the session scratch (never
    -- the workspace), the browser inside a kill-on-close job, its pid recorded
    -- with its identity. Port 0: the browser picks one and writes it into the
    -- profile's DevToolsActivePort.
    local browser, why = Preview.launch({
        scratch = scratch_of(context), workspace = context.workspace, config = context.config,
        port = 0, executable = req.browser, url = "about:blank",
        args = { "--disable-extensions", "--disable-background-networking", "--disable-sync",
            "--disable-component-update", "--mute-audio" },
    })
    if not browser then error(tostring(why), 0) end

    local result = { ok = false, url = req.url, browser = browser.executable, size = req.width .. "x" .. req.height,
        out_dir = req.out_dir, shots = {}, results = {}, console = {} }
    local ws, cdp
    local ok, err = pcall(function()
        local port
        local deadline = now_ms(socket) + LAUNCH_TIMEOUT_MS
        while not port do
            local handle = io.open(path.join(browser.profile, "DevToolsActivePort"), "rb")
            if handle then
                port = tonumber((handle:read("*a") or ""):match("^(%d+)"))
                handle:close()
            end
            if not port then
                if now_ms(socket) > deadline then error("the browser did not open a DevTools port within " .. LAUNCH_TIMEOUT_MS .. " ms", 0) end
                socket.sleep(0.1)
            end
        end
        local version = http_get(socket, port, "/json/version") or ""
        local ws_path = version:match('"webSocketDebuggerUrl"%s*:%s*"ws://[^/]+(/[^"]+)"')
        if not ws_path then error("the browser's DevTools answered no WebSocket URL", 0) end
        ws = ws_connect(socket, port, ws_path)
        cdp = new_cdp(socket, ws)
        local target = cdp.call("Target.createTarget", { url = "about:blank" })
        local attached = cdp.call("Target.attachToTarget", { targetId = target.targetId, flatten = true })
        local session = attached.sessionId
        for _, domain in ipairs({ "Page.enable", "Runtime.enable", "Log.enable" }) do cdp.call(domain, nil, session) end
        local failed_step, step_error = run_steps(cdp, session, req, socket, result.shots, result.results, context)
        if failed_step then
            result.failed_step = failed_step
            result.error = "step " .. failed_step .. " (" .. describe(req.steps[failed_step]) .. "): " .. tostring(step_error)
        else
            result.ok = true
        end
    end)
    if not ok then result.error = tostring(err) end
    if cdp then for _, line in ipairs(cdp.console) do result.console[#result.console + 1] = line end end

    -- Cleanup, whatever happened above: ask the browser to close, then girl's
    -- preview stop - the tree by its recorded identity (never by name), the
    -- profile removed, and the process table read back.
    if cdp then pcall(cdp.call, "Browser.close", nil, nil, 5000) end
    if ws then ws.close() end
    local stopped, stop_error = Preview.stop(browser)
    if not stopped then result.cleanup = "the browser cleanup did not finish: " .. tostring(stop_error) end
    if #result.console == 0 then result.console = nil end
    if #result.results == 0 then result.results = nil end
    return result
end

return function(registry)
    registry:register {
        name = "browser_shot",
        description = "Open a URL in a disposable headless Chromium browser (Chrome, Edge or Chromium must be installed), "
            .. "optionally run interaction steps (click, type, key, wait, eval, resize, color scheme), and save PNG "
            .. "screenshots of the viewport, the full page or one element. Returns the screenshot paths (view them "
            .. "with read_file), eval results and the page's console errors. Use it to check how a web UI really "
            .. "renders. Each call starts and closes its own browser.",
        input_schema = {
            type = "object",
            properties = {
                url = { type = "string", description = "Page to open first (http, https, file or about)." },
                steps = {
                    type = "array",
                    description = "Actions in order; each object names ONE action. "
                        .. '{"goto":url} {"wait":"css"|ms} {"wait_js":"expr"} {"click":"css"} {"hover":"css"} '
                        .. '{"scroll":"css"} {"type":["css","text"],"clear":true} {"key":"Enter|Escape|Tab|ArrowDown|a"} '
                        .. '{"eval":"js expr (awaited; value returned)"} {"size":"390x844"} {"color_scheme":"dark"} '
                        .. '{"shot":"name.png","selector":"css (element clip)","full_page":true,"padding":8}. '
                        .. "Any step may carry timeout_ms (default 15000). Without a shot step one viewport shot is taken at the end.",
                    items = { type = "object" },
                },
                scenario = { type = "string", description = "Workspace path of a JSON file {url?, size?, color_scheme?, steps} (instead of steps)." },
                size = { type = "string", description = "Viewport WIDTHxHEIGHT, default 1440x900." },
                color_scheme = { type = "string", enum = { "light", "dark", "no-preference" }, description = "Emulated prefers-color-scheme." },
                out_dir = { type = "string", description = "Workspace-relative folder for the PNGs; default: the session scratch (browser_shot/)." },
                browser = { type = "string", description = "Browser executable path or name; default: auto-detect Chrome, Edge, Chromium." },
                settle_ms = { type = "integer", description = "Pause after each action before the next one (default 150)." },
            },
            additionalProperties = false,
        },
        permission = "browser.drive",
        resource = function(args)
            local url = type(args) == "table" and args.url
            if type(url) == "string" and url ~= "" then return url end
            return type(args) == "table" and args.scenario and ("scenario " .. tostring(args.scenario)) or "browser"
        end,
        approval_details = function(args, context)
            args, context = args or {}, context or {}
            local ok, req = pcall(plan, args, context, true)
            if not ok then return { lines = { "invalid request: " .. tostring(req) } } end
            local lines = {
                "headless browser → " .. (args.scenario and not args.url and "(url from the scenario)" or req.url)
                    .. "  (" .. req.width .. "x" .. req.height .. (req.scheme and (", " .. req.scheme) or "") .. ")",
            }
            if args.scenario then
                lines[#lines + 1] = "steps: from the scenario file " .. tostring(args.scenario)
            else
                for i, step in ipairs(req.steps) do lines[#lines + 1] = string.format("%2d. %s", i, describe(step)) end
            end
            lines[#lines + 1] = "screenshots → " .. req.out_dir
            return { lines = lines }
        end,
        run = execute,
    }
end

• 1.0.0 3 days ago · 34 KB · download

  md5 8263277943982f08b5ca830f6d1c26cb