refactor(render): make lazy-render always active, fix cursor and gg

- Remove opts.lazy: lazy-render is always active as a perf optimization
- Decouple hidden_count from lazy-render truncation (only max_messages counts)
- Fix cursor jump after load_more: preserve position via anchor message
- Add gg keymap: load all messages before jumping to top, so full history
  is searchable

Author: jensenojs

lua/opencode/ui/output_window.lua

@@ -593,6 +593,15 @@ end
 function M.setup_keymaps(windows)
   local keymap = require('opencode.keymap')
   keymap.setup_window_keymaps(config.keymap.output_window, windows.output_buf)
+
+  -- When lazy-render is active, gg only reaches the top of rendered content.
+  -- Load all messages first so gg reaches the true start of history.
+  -- See tests/unit/renderer_lazy_spec.lua: "gg loads all messages and makes them searchable"
+  vim.keymap.set('n', 'gg', function()
+    local renderer = require('opencode.ui.renderer')
+    renderer.load_all_messages()
+    return 'gg'
+  end, { buffer = windows.output_buf, expr = true, remap = true })
 end
 
 ---@param windows OpencodeWindowState
@@ -633,13 +642,42 @@ function M.setup_autocmds(windows, group)
     end,
   })
 
-  -- Lazy-render: load more messages on scroll-to-top (debounced)
-  local debounced_load_more = require('opencode.util').debounce(function()
-    local renderer = require('opencode.ui.renderer')
-    if renderer.load_more_messages() then
-      pcall(vim.api.nvim_win_set_cursor, windows.output_win, { 3, 0 })
-    end
-  end, 150)
+    -- Lazy-render: load more messages on scroll-to-top (debounced)
+    local debounced_load_more = require('opencode.util').debounce(function()
+      local renderer = require('opencode.ui.renderer')
+      local render_state = require('opencode.ui.renderer.ctx').render_state
+      -- Save the message at the top of the viewport before loading more
+      local ok, cursor = pcall(vim.api.nvim_win_get_cursor, windows.output_win)
+      local anchor_msg_id = nil
+      local anchor_was_at_line = nil
+      if ok and cursor then
+        local top_line = cursor[1]
+        for _, msg in ipairs(state.messages or {}) do
+          local msg_id = msg.info and msg.info.id or ''
+          if not msg_id:match('^__opencode_') then
+            local rendered = render_state:get_message(msg_id)
+            if rendered and rendered.line_start and rendered.line_start >= top_line then
+              anchor_msg_id = msg_id
+              anchor_was_at_line = rendered.line_start
+              break
+            end
+          end
+        end
+      end
+
+      if renderer.load_more_messages() then
+        -- Restore cursor to the anchor message's new position
+        if anchor_msg_id then
+          local rendered = render_state:get_message(anchor_msg_id)
+          if rendered and rendered.line_start then
+            pcall(vim.api.nvim_win_set_cursor, windows.output_win, { rendered.line_start, 0 })
+            return
+          end
+        end
+        -- Fallback: move to top of buffer
+        pcall(vim.api.nvim_win_set_cursor, windows.output_win, { 1, 0 })
+      end
+    end, 150)
   vim.api.nvim_create_autocmd('WinScrolled', {
     group = group,
     buffer = windows.output_buf,

lua/opencode/ui/renderer.lua

@@ -349,20 +349,16 @@ function M._render_full_session_data(session_data, opts)
   local visible_messages, hidden_count = get_visible_session_messages(state.messages)
   local revert_index = get_revert_index(state.messages)
 
-  -- Lazy-render: only render viewport-sized message count.
-  -- Active when ctx.lazy_render_count is set or opts.lazy=true.
-  if lazy_limit == nil and opts.lazy then
-    local initial = get_initial_render_count()
-    if #visible_messages > initial then
-      lazy_limit = initial
+    if lazy_limit == nil then
+      local initial = get_initial_render_count()
+      if #visible_messages > initial then
+        lazy_limit = initial
+      end
+    end
+    ctx.lazy_render_count = lazy_limit
+    if lazy_limit and #visible_messages > lazy_limit then
+      visible_messages = vim.list_slice(visible_messages, #visible_messages - lazy_limit + 1)
     end
-  end
-  -- Persist back to ctx so it survives the next M.reset()
-  ctx.lazy_render_count = lazy_limit
-  if lazy_limit and #visible_messages > lazy_limit then
-    hidden_count = hidden_count + (#visible_messages - lazy_limit)
-    visible_messages = vim.list_slice(visible_messages, #visible_messages - lazy_limit + 1)
-  end
 
   local t_format_start = vim.uv.hrtime()
   flush.begin_bulk_mode()
@@ -435,10 +431,9 @@ function M.render_from_cache(session_data)
   if not output_window.mounted() or not state.api_client then
     return
   end
-  M._render_full_session_data(session_data, {
-    restore_model_from_messages = true,
-    lazy = ctx.lazy_render_count ~= nil,
-  })
+    M._render_full_session_data(session_data, {
+      restore_model_from_messages = true,
+    })
   local active_session = state.active_session
   if active_session and active_session.id then
     require('opencode.ui.question_window').restore_pending_question(active_session.id)
@@ -468,17 +463,38 @@ function M.load_more_messages()
   return true
 end
 
+---Load all remaining messages and re-render.
+---Used when user explicitly navigates to the top (gg) to ensure
+---the full history is available for navigation and search.
+---@return boolean Whether any messages were loaded
+function M.load_all_messages()
+  if not state.messages then
+    return false
+  end
+  local total = #get_visible_session_messages(state.messages)
+  if total == 0 then
+    return false
+  end
+  local current = ctx.lazy_render_count or 0
+  if current >= total then
+    return false
+  end
+
+  ctx.lazy_render_count = total
+  M.render_from_cache(state.messages)
+  return true
+end
+
 ---Fetch the active session from the server and render it
 ---@return Promise<OpencodeMessage[]>
 function M.render_full_session()
   if not output_window.mounted() or not state.api_client then
     return Promise.new():resolve(nil)
   end
   return fetch_session():and_then(function(session_data)
-    M._render_full_session_data(session_data, {
-      restore_model_from_messages = true,
-      lazy = true,
-    })
+      M._render_full_session_data(session_data, {
+        restore_model_from_messages = true,
+      })
     local active_session = state.active_session
     if active_session and active_session.id then
       require('opencode.ui.question_window').restore_pending_question(active_session.id)