Files
avante.nvim/lua/avante/llm.lua
2025-06-23 03:13:37 +08:00

1070 lines
38 KiB
Lua

local api = vim.api
local fn = vim.fn
local uv = vim.uv
local curl = require("plenary.curl")
local Utils = require("avante.utils")
local Config = require("avante.config")
local Path = require("avante.path")
local Providers = require("avante.providers")
local LLMToolHelpers = require("avante.llm_tools.helpers")
local LLMTools = require("avante.llm_tools")
local HistoryMessage = require("avante.history_message")
---@class avante.LLM
local M = {}
M.CANCEL_PATTERN = "AvanteLLMEscape"
------------------------------Prompt and type------------------------------
local group = api.nvim_create_augroup("avante_llm", { clear = true })
---@param content AvanteLLMMessageContent
---@param cb fun(title: string | nil): nil
function M.summarize_chat_thread_title(content, cb)
local system_prompt =
[[Summarize the content as a title for the chat thread. The title should be a concise and informative summary of the conversation, capturing the main points and key takeaways. It should be no longer than 100 words and should be written in a clear and engaging style. The title should be suitable for use as the title of a chat thread on a messaging platform or other communication medium. /no_think]]
local response_content = ""
local provider = Providers.get_memory_summary_provider()
M.curl({
provider = provider,
prompt_opts = {
system_prompt = system_prompt,
messages = {
{ role = "user", content = content },
},
},
handler_opts = {
on_start = function(_) end,
on_chunk = function(chunk)
if not chunk then return end
response_content = response_content .. chunk
end,
on_stop = function(stop_opts)
if stop_opts.error ~= nil then
Utils.error(string.format("summarize chat thread title failed: %s", vim.inspect(stop_opts.error)))
return
end
if stop_opts.reason == "complete" then
response_content = Utils.trim_think_content(response_content)
response_content = Utils.trim(response_content, { prefix = "\n", suffix = "\n" })
response_content = Utils.trim(response_content, { prefix = '"', suffix = '"' })
local title = response_content
cb(title)
end
end,
},
})
end
---@param prev_memory string | nil
---@param history_messages avante.HistoryMessage[]
---@param cb fun(memory: avante.ChatMemory | nil): nil
function M.summarize_memory(prev_memory, history_messages, cb)
local system_prompt =
[[You are an expert coding assistant. Your goal is to generate a concise, structured summary of the conversation below that captures all essential information needed to continue development after context replacement. Include tasks performed, code areas modified or reviewed, key decisions or assumptions, test results or errors, and outstanding tasks or next steps.]]
if #history_messages == 0 then
cb(nil)
return
end
local latest_timestamp = nil
local latest_message_uuid = nil
for idx = #history_messages, 1, -1 do
local message = history_messages[idx]
if not message.is_dummy then
latest_timestamp = message.timestamp
latest_message_uuid = message.uuid
break
end
end
if not latest_timestamp or not latest_message_uuid then
cb(nil)
return
end
local conversation_items = vim
.iter(history_messages)
:map(function(msg) return msg.message.role .. ": " .. Utils.message_to_text(msg, history_messages) end)
:totable()
local conversation_text = table.concat(conversation_items, "\n")
local user_prompt = "Here is the conversation so far:\n"
.. conversation_text
.. "\n\nPlease summarize this conversation, covering:\n1. Tasks performed and outcomes\n2. Code files, modules, or functions modified or examined\n3. Important decisions or assumptions made\n4. Errors encountered and test or build results\n5. Remaining tasks, open questions, or next steps\nProvide the summary in a clear, concise format."
if prev_memory then user_prompt = user_prompt .. "\n\nThe previous summary is:\n\n" .. prev_memory end
local messages = {
{
role = "user",
content = user_prompt,
},
}
local response_content = ""
local provider = Providers.get_memory_summary_provider()
M.curl({
provider = provider,
prompt_opts = {
system_prompt = system_prompt,
messages = messages,
},
handler_opts = {
on_start = function(_) end,
on_chunk = function(chunk)
if not chunk then return end
response_content = response_content .. chunk
end,
on_stop = function(stop_opts)
if stop_opts.error ~= nil then
Utils.error(string.format("summarize memory failed: %s", vim.inspect(stop_opts.error)))
return
end
if stop_opts.reason == "complete" then
response_content = Utils.trim_think_content(response_content)
local memory = {
content = response_content,
last_summarized_timestamp = latest_timestamp,
last_message_uuid = latest_message_uuid,
}
cb(memory)
else
cb(nil)
end
end,
},
})
end
---@param user_input string
---@param cb fun(error: string | nil): nil
function M.generate_todos(user_input, cb)
local system_prompt =
[[You are an expert coding assistant. Please generate a todo list to complete the task based on the user input and pass the todo list to the add_todos tool.]]
local messages = {
{ role = "user", content = user_input },
}
local provider = Providers[Config.provider]
local tools = {
require("avante.llm_tools.add_todos"),
}
local history_messages = {}
cb = Utils.call_once(cb)
M.curl({
provider = provider,
prompt_opts = {
system_prompt = system_prompt,
messages = messages,
tools = tools,
},
handler_opts = {
on_start = function() end,
on_chunk = function() end,
on_messages_add = function(msgs)
msgs = vim.islist(msgs) and msgs or { msgs }
for _, msg in ipairs(msgs) do
if not msg.uuid then msg.uuid = Utils.uuid() end
local idx = nil
for i, m in ipairs(history_messages) do
if m.uuid == msg.uuid then
idx = i
break
end
end
if idx ~= nil then
history_messages[idx] = msg
else
table.insert(history_messages, msg)
end
end
end,
on_stop = function(stop_opts)
if stop_opts.error ~= nil then
Utils.error(string.format("generate todos failed: %s", vim.inspect(stop_opts.error)))
return
end
if stop_opts.reason == "tool_use" then
local uncalled_tool_uses = Utils.get_uncalled_tool_uses(history_messages)
for _, partial_tool_use in ipairs(uncalled_tool_uses) do
if partial_tool_use.state == "generated" and partial_tool_use.name == "add_todos" then
LLMTools.process_tool_use(tools, partial_tool_use, function() end, function() cb() end, {})
cb()
end
end
else
cb()
end
end,
},
})
end
---@param opts AvanteGeneratePromptsOptions
---@return AvantePromptOptions
function M.generate_prompts(opts)
local provider = opts.provider or Providers[Config.provider]
local mode = opts.mode or Config.mode
-- Check if the instructions contains an image path
local image_paths = {}
if opts.prompt_opts and opts.prompt_opts.image_paths then
image_paths = vim.list_extend(image_paths, opts.prompt_opts.image_paths)
end
local project_root = Utils.root.get()
Path.prompts.initialize(Path.prompts.get_templates_dir(project_root), project_root)
local system_info = Utils.get_system_info()
local selected_files = opts.selected_files or {}
if opts.selected_filepaths then
for _, filepath in ipairs(opts.selected_filepaths) do
local lines, error = Utils.read_file_from_buf_or_disk(filepath)
lines = lines or {}
local filetype = Utils.get_filetype(filepath)
if error ~= nil then
Utils.error("error reading file: " .. error)
else
local content = table.concat(lines, "\n")
table.insert(selected_files, { path = filepath, content = content, file_type = filetype })
end
end
end
local viewed_files = {}
if opts.history_messages then
for _, message in ipairs(opts.history_messages) do
local content = message.message.content
if type(content) ~= "table" then goto continue end
for _, item in ipairs(content) do
if type(item) ~= "table" then goto continue1 end
if item.type ~= "tool_use" then goto continue1 end
local tool_name = item.name
if tool_name ~= "view" then goto continue1 end
local path = item.input.path
if path then
local uniform_path = Utils.uniform_path(path)
viewed_files[uniform_path] = item.id
end
::continue1::
end
::continue::
end
end
selected_files = vim.iter(selected_files):filter(function(file) return viewed_files[file.path] == nil end):totable()
local template_opts = {
ask = opts.ask, -- TODO: add mode without ask instruction
code_lang = opts.code_lang,
selected_files = selected_files,
selected_code = opts.selected_code,
recently_viewed_files = opts.recently_viewed_files,
project_context = opts.project_context,
diagnostics = opts.diagnostics,
system_info = system_info,
model_name = provider.model or "unknown",
memory = opts.memory,
}
if opts.get_todos then
local todos = opts.get_todos()
if todos and #todos > 0 then template_opts.todos = vim.json.encode(todos) end
end
local system_prompt
if opts.prompt_opts and opts.prompt_opts.system_prompt then
system_prompt = opts.prompt_opts.system_prompt
else
system_prompt = Path.prompts.render_mode(mode, template_opts)
end
if Config.system_prompt ~= nil then
local custom_system_prompt = Config.system_prompt
if type(custom_system_prompt) == "function" then custom_system_prompt = custom_system_prompt() end
if custom_system_prompt ~= nil and custom_system_prompt ~= "" and custom_system_prompt ~= "null" then
system_prompt = system_prompt .. "\n\n" .. custom_system_prompt
end
end
---@type AvanteLLMMessage[]
local context_messages = {}
if opts.prompt_opts and opts.prompt_opts.messages then
context_messages = vim.list_extend(context_messages, opts.prompt_opts.messages)
end
if opts.project_context ~= nil and opts.project_context ~= "" and opts.project_context ~= "null" then
local project_context = Path.prompts.render_file("_project.avanterules", template_opts)
if project_context ~= "" then
table.insert(context_messages, { role = "user", content = project_context, visible = false, is_context = true })
end
end
if opts.diagnostics ~= nil and opts.diagnostics ~= "" and opts.diagnostics ~= "null" then
local diagnostics = Path.prompts.render_file("_diagnostics.avanterules", template_opts)
if diagnostics ~= "" then
table.insert(context_messages, { role = "user", content = diagnostics, visible = false, is_context = true })
end
end
if #selected_files > 0 or opts.selected_code ~= nil then
local code_context = Path.prompts.render_file("_context.avanterules", template_opts)
if code_context ~= "" then
table.insert(context_messages, { role = "user", content = code_context, visible = false, is_context = true })
end
end
if opts.memory ~= nil and opts.memory ~= "" and opts.memory ~= "null" then
local memory = Path.prompts.render_file("_memory.avanterules", template_opts)
if memory ~= "" then
table.insert(context_messages, { role = "user", content = memory, visible = false, is_context = true })
end
end
local pending_compaction_history_messages = {}
if opts.prompt_opts and opts.prompt_opts.pending_compaction_history_messages then
pending_compaction_history_messages =
vim.list_extend(pending_compaction_history_messages, opts.prompt_opts.pending_compaction_history_messages)
end
local context_window = provider.context_window
if context_window and context_window > 0 then
Utils.debug("Context window", context_window)
if opts.get_tokens_usage then
local tokens_usage = opts.get_tokens_usage()
if tokens_usage then
local target_tokens = context_window * 0.9
local tokens_count = tokens_usage.prompt_tokens + tokens_usage.completion_tokens
Utils.debug("Tokens count", tokens_count)
if tokens_count > target_tokens then pending_compaction_history_messages = opts.history_messages end
end
end
end
---@type AvanteLLMMessage[]
local messages = vim.deepcopy(context_messages)
for _, msg in ipairs(opts.history_messages or {}) do
local message = msg.message
if msg.is_user_submission then
message = vim.deepcopy(message)
local content = message.content
if type(content) == "string" then
message.content = "<task>" .. content .. "</task>"
elseif type(content) == "table" then
for idx, item in ipairs(content) do
if type(item) == "string" then
item = "<task>" .. item .. "</task>"
content[idx] = item
elseif type(item) == "table" and item.type == "text" then
item.content = "<task>" .. item.content .. "</task>"
content[idx] = item
end
end
end
end
table.insert(messages, message)
end
messages = vim
.iter(messages)
:filter(function(msg) return type(msg.content) ~= "string" or msg.content ~= "" end)
:totable()
if opts.instructions ~= nil and opts.instructions ~= "" then
messages = vim.list_extend(messages, { { role = "user", content = opts.instructions } })
end
opts.session_ctx = opts.session_ctx or {}
opts.session_ctx.system_prompt = system_prompt
opts.session_ctx.messages = messages
local tools = {}
if opts.tools then tools = vim.list_extend(tools, opts.tools) end
if opts.prompt_opts and opts.prompt_opts.tools then tools = vim.list_extend(tools, opts.prompt_opts.tools) end
---@type AvantePromptOptions
return {
system_prompt = system_prompt,
messages = messages,
image_paths = image_paths,
tools = tools,
pending_compaction_history_messages = pending_compaction_history_messages,
}
end
---@param opts AvanteGeneratePromptsOptions
---@return integer
function M.calculate_tokens(opts)
local prompt_opts = M.generate_prompts(opts)
local tokens = Utils.tokens.calculate_tokens(prompt_opts.system_prompt)
for _, message in ipairs(prompt_opts.messages) do
tokens = tokens + Utils.tokens.calculate_tokens(message.content)
end
return tokens
end
local parse_headers = function(headers_file)
local headers = {}
local file = io.open(headers_file, "r")
if file then
for line in file:lines() do
line = line:gsub("\r$", "")
local key, value = line:match("^%s*(.-)%s*:%s*(.*)$")
if key and value then headers[key] = value end
end
file:close()
end
return headers
end
---@param opts avante.CurlOpts
function M.curl(opts)
local provider = opts.provider
local prompt_opts = opts.prompt_opts
local handler_opts = opts.handler_opts
local orig_on_stop = handler_opts.on_stop
local stopped = false
---@param stop_opts AvanteLLMStopCallbackOptions
handler_opts.on_stop = function(stop_opts)
if stop_opts and not stop_opts.streaming_tool_use then
if stopped then return end
stopped = true
end
if orig_on_stop then return orig_on_stop(stop_opts) end
end
---@type AvanteCurlOutput
local spec = provider:parse_curl_args(prompt_opts)
---@type string
local current_event_state = nil
local turn_ctx = {}
turn_ctx.turn_id = Utils.uuid()
local response_body = ""
---@param line string
local function parse_stream_data(line)
local event = line:match("^event:%s*(.+)$")
if event then
current_event_state = event
return
end
local data_match = line:match("^data:%s*(.+)$")
if data_match then
response_body = ""
provider:parse_response(turn_ctx, data_match, current_event_state, handler_opts)
else
response_body = response_body .. line
local ok, jsn = pcall(vim.json.decode, response_body)
if ok then
if jsn.error then
handler_opts.on_stop({ reason = "error", error = jsn.error })
else
provider:parse_response(turn_ctx, response_body, current_event_state, handler_opts)
end
response_body = ""
end
end
end
local function parse_response_without_stream(data)
provider:parse_response_without_stream(data, current_event_state, handler_opts)
end
local completed = false
local active_job
local temp_file = fn.tempname()
local curl_body_file = temp_file .. "-request-body.json"
local resp_body_file = temp_file .. "-response-body.txt"
local headers_file = temp_file .. "-response-headers.txt"
local json_content = vim.json.encode(spec.body)
fn.writefile(vim.split(json_content, "\n"), curl_body_file)
Utils.debug("curl request body file:", curl_body_file)
Utils.debug("curl response body file:", resp_body_file)
Utils.debug("curl headers file:", headers_file)
local function cleanup()
if Config.debug then return end
vim.schedule(function()
fn.delete(curl_body_file)
pcall(fn.delete, resp_body_file)
fn.delete(headers_file)
end)
end
local headers_reported = false
active_job = curl.post(spec.url, {
headers = spec.headers,
proxy = spec.proxy,
insecure = spec.insecure,
body = curl_body_file,
raw = spec.rawArgs,
dump = { "-D", headers_file },
stream = function(err, data, _)
if not headers_reported and opts.on_response_headers then
headers_reported = true
opts.on_response_headers(parse_headers(headers_file))
end
if err then
completed = true
handler_opts.on_stop({ reason = "error", error = err })
return
end
if not data then return end
if Config.debug then
if type(data) == "string" then
local file = io.open(resp_body_file, "a")
if file then
file:write(data .. "\n")
file:close()
end
end
end
vim.schedule(function()
if provider.parse_stream_data ~= nil then
provider:parse_stream_data(turn_ctx, data, handler_opts)
else
parse_stream_data(data)
end
end)
end,
on_error = function(err)
if err.exit == 23 then
local xdg_runtime_dir = os.getenv("XDG_RUNTIME_DIR")
if not xdg_runtime_dir or fn.isdirectory(xdg_runtime_dir) == 0 then
Utils.error(
"$XDG_RUNTIME_DIR="
.. xdg_runtime_dir
.. " is set but does not exist. curl could not write output. Please make sure it exists, or unset.",
{ title = "Avante" }
)
elseif not uv.fs_access(xdg_runtime_dir, "w") then
Utils.error(
"$XDG_RUNTIME_DIR="
.. xdg_runtime_dir
.. " exists but is not writable. curl could not write output. Please make sure it is writable, or unset.",
{ title = "Avante" }
)
end
end
active_job = nil
if not completed then
completed = true
cleanup()
handler_opts.on_stop({ reason = "error", error = err })
end
end,
callback = function(result)
active_job = nil
cleanup()
local headers_map = vim.iter(result.headers):fold({}, function(acc, value)
local pieces = vim.split(value, ":")
local key = pieces[1]
local remain = vim.list_slice(pieces, 2)
if not remain then return acc end
local val = Utils.trim_spaces(table.concat(remain, ":"))
acc[key] = val
return acc
end)
if result.status >= 400 then
if provider.on_error then
provider.on_error(result)
else
Utils.error("API request failed with status " .. result.status, { once = true, title = "Avante" })
end
local retry_after = 10
if headers_map["retry-after"] then retry_after = tonumber(headers_map["retry-after"]) or 10 end
if result.status == 429 then
handler_opts.on_stop({ reason = "rate_limit", retry_after = retry_after })
return
end
vim.schedule(function()
if not completed then
completed = true
handler_opts.on_stop({
reason = "error",
error = "API request failed with status " .. result.status .. ". Body: " .. vim.inspect(result.body),
})
end
end)
end
-- If stream is not enabled, then handle the response here
if provider:is_disable_stream() and result.status == 200 then
vim.schedule(function()
completed = true
parse_response_without_stream(result.body)
end)
end
if result.status == 200 and spec.url:match("https://openrouter.ai") then
local content_type = headers_map["content-type"]
if content_type and content_type:match("text/html") then
handler_opts.on_stop({
reason = "error",
error = "Your openrouter endpoint setting is incorrect, please set it to https://openrouter.ai/api/v1",
})
end
end
end,
})
api.nvim_create_autocmd("User", {
group = group,
pattern = M.CANCEL_PATTERN,
once = true,
callback = function()
-- Error: cannot resume dead coroutine
if active_job then
-- Mark as completed first to prevent error handler from running
completed = true
-- Attempt to shutdown the active job, but ignore any errors
xpcall(function() active_job:shutdown() end, function(err)
Utils.debug("Ignored error during job shutdown: " .. vim.inspect(err))
return err
end)
Utils.debug("LLM request cancelled")
active_job = nil
-- Clean up and notify of cancellation
cleanup()
vim.schedule(function() handler_opts.on_stop({ reason = "cancelled" }) end)
end
end,
})
return active_job
end
---@param opts AvanteLLMStreamOptions
function M._stream(opts)
-- Reset the cancellation flag at the start of a new request
if LLMToolHelpers then LLMToolHelpers.is_cancelled = false end
local provider = opts.provider or Providers[Config.provider]
opts.session_ctx = opts.session_ctx or {}
if not opts.session_ctx.on_messages_add then opts.session_ctx.on_messages_add = opts.on_messages_add end
if not opts.session_ctx.on_state_change then opts.session_ctx.on_state_change = opts.on_state_change end
if not opts.session_ctx.on_start then opts.session_ctx.on_start = opts.on_start end
if not opts.session_ctx.on_chunk then opts.session_ctx.on_chunk = opts.on_chunk end
if not opts.session_ctx.on_stop then opts.session_ctx.on_stop = opts.on_stop end
if not opts.session_ctx.on_tool_log then opts.session_ctx.on_tool_log = opts.on_tool_log end
if not opts.session_ctx.get_history_messages then
opts.session_ctx.get_history_messages = opts.get_history_messages
end
---@cast provider AvanteProviderFunctor
local prompt_opts = M.generate_prompts(opts)
if
prompt_opts.pending_compaction_history_messages
and #prompt_opts.pending_compaction_history_messages > 0
and opts.on_memory_summarize
then
opts.on_memory_summarize(prompt_opts.pending_compaction_history_messages)
return
end
local resp_headers = {}
---@type AvanteHandlerOptions
local handler_opts = {
on_messages_add = opts.on_messages_add,
on_state_change = opts.on_state_change,
update_tokens_usage = opts.update_tokens_usage,
on_start = opts.on_start,
on_chunk = opts.on_chunk,
on_stop = function(stop_opts)
if stop_opts.usage and opts.update_tokens_usage then opts.update_tokens_usage(stop_opts.usage) end
---@param partial_tool_use_list AvantePartialLLMToolUse[]
---@param tool_use_index integer
---@param tool_results AvanteLLMToolResult[]
local function handle_next_tool_use(partial_tool_use_list, tool_use_index, tool_results, streaming_tool_use)
if tool_use_index > #partial_tool_use_list then
---@type avante.HistoryMessage[]
local messages = {}
for _, tool_result in ipairs(tool_results) do
messages[#messages + 1] = HistoryMessage:new({
role = "user",
content = {
{
type = "tool_result",
tool_use_id = tool_result.tool_use_id,
content = tool_result.content,
is_error = tool_result.is_error,
is_user_declined = tool_result.is_user_declined,
},
},
})
end
if opts.on_messages_add then opts.on_messages_add(messages) end
local the_last_tool_use = partial_tool_use_list[#partial_tool_use_list]
if the_last_tool_use and the_last_tool_use.name == "attempt_completion" then
opts.on_stop({ reason = "complete" })
return
end
local new_opts = vim.tbl_deep_extend("force", opts, {
history_messages = opts.get_history_messages(),
})
if provider.get_rate_limit_sleep_time then
local sleep_time = provider:get_rate_limit_sleep_time(resp_headers)
if sleep_time and sleep_time > 0 then
Utils.info("Rate limit reached. Sleeping for " .. sleep_time .. " seconds ...")
vim.defer_fn(function() M._stream(new_opts) end, sleep_time * 1000)
return
end
end
M._stream(new_opts)
return
end
local partial_tool_use = partial_tool_use_list[tool_use_index]
---@param result string | nil
---@param error string | nil
local function handle_tool_result(result, error)
-- Special handling for cancellation signal from tools
if error == LLMToolHelpers.CANCEL_TOKEN then
Utils.debug("Tool execution was cancelled by user")
if opts.on_chunk then opts.on_chunk("\n*[Request cancelled by user during tool execution.]*\n") end
if opts.on_messages_add then
local message = HistoryMessage:new({
role = "assistant",
content = "\n*[Request cancelled by user during tool execution.]*\n",
}, {
just_for_display = true,
})
opts.on_messages_add({ message })
end
return opts.on_stop({ reason = "cancelled" })
end
local is_user_declined = error and error:match("^User declined")
local tool_result = {
tool_use_id = partial_tool_use.id,
content = error ~= nil and error or result,
is_error = error ~= nil, -- Keep this as error to prevent processing as success
is_user_declined = is_user_declined ~= nil,
}
table.insert(tool_results, tool_result)
return handle_next_tool_use(partial_tool_use_list, tool_use_index + 1, tool_results)
end
local is_edit_tool_use = Utils.is_edit_func_call_tool_use(partial_tool_use)
local support_streaming = false
local llm_tool = vim.iter(prompt_opts.tools):find(function(tool) return tool.name == partial_tool_use.name end)
if llm_tool then support_streaming = llm_tool.support_streaming == true end
if partial_tool_use.state == "generating" and not is_edit_tool_use and not support_streaming then return end
if type(partial_tool_use.input) == "table" then partial_tool_use.input.tool_use_id = partial_tool_use.id end
if partial_tool_use.state == "generating" then
if type(partial_tool_use.input) == "table" then
partial_tool_use.input.streaming = true
LLMTools.process_tool_use(
prompt_opts.tools,
partial_tool_use,
function() end,
function() end,
opts.session_ctx
)
end
return
else
if streaming_tool_use then return end
end
-- Either on_complete handles the tool result asynchronously or we receive the result and error synchronously when either is not nil
local result, error = LLMTools.process_tool_use(
prompt_opts.tools,
partial_tool_use,
opts.on_tool_log,
handle_tool_result,
opts.session_ctx
)
if result ~= nil or error ~= nil then return handle_tool_result(result, error) end
end
if stop_opts.reason == "cancelled" then
if opts.on_chunk then opts.on_chunk("\n*[Request cancelled by user.]*\n") end
if opts.on_messages_add then
local message = HistoryMessage:new({
role = "assistant",
content = "\n*[Request cancelled by user.]*\n",
}, {
just_for_display = true,
})
opts.on_messages_add({ message })
end
return opts.on_stop({ reason = "cancelled" })
end
local history_messages = opts.get_history_messages and opts.get_history_messages({ all = true }) or {}
local uncalled_tool_uses = Utils.get_uncalled_tool_uses(history_messages)
if stop_opts.reason == "complete" and Config.mode == "agentic" then
local completed_attempt_completion_tool_use = nil
for idx = #history_messages, 1, -1 do
local message = history_messages[idx]
if message.is_user_submission then break end
if not Utils.is_tool_use_message(message) then goto continue end
if message.message.content[1].name ~= "attempt_completion" then break end
completed_attempt_completion_tool_use = message
if message then break end
::continue::
end
local unfinished_todos = {}
if opts.get_todos then
local todos = opts.get_todos()
unfinished_todos = vim.tbl_filter(
function(todo) return todo.status ~= "done" or todo.status ~= "cancelled" end,
todos
)
end
local user_reminder_count = opts.session_ctx.user_reminder_count or 0
if
not completed_attempt_completion_tool_use
and opts.on_messages_add
and (user_reminder_count < 3 or #unfinished_todos > 0)
then
opts.session_ctx.user_reminder_count = user_reminder_count + 1
Utils.debug("user reminder count", user_reminder_count)
local message
if #unfinished_todos > 0 then
message = HistoryMessage:new({
role = "user",
content = "<user-reminder>You should use tool calls to answer the question, for example, use update_todo_status if the task step is done or cancelled.</user-reminder>",
}, {
visible = false,
})
else
message = HistoryMessage:new({
role = "user",
content = "<user-reminder>You should use tool calls to answer the question, for example, use attempt_completion if the job is done.</user-reminder>",
}, {
visible = false,
})
end
opts.on_messages_add({ message })
local new_opts = vim.tbl_deep_extend("force", opts, {
history_messages = opts.get_history_messages(),
})
if provider.get_rate_limit_sleep_time then
local sleep_time = provider:get_rate_limit_sleep_time(resp_headers)
if sleep_time and sleep_time > 0 then
Utils.info("Rate limit reached. Sleeping for " .. sleep_time .. " seconds ...")
vim.defer_fn(function() M._stream(new_opts) end, sleep_time * 1000)
return
end
end
M._stream(new_opts)
return
end
end
if stop_opts.reason == "tool_use" then
opts.session_ctx.user_reminder_count = 0
return handle_next_tool_use(uncalled_tool_uses, 1, {}, stop_opts.streaming_tool_use)
end
if stop_opts.reason == "rate_limit" then
local msg_content = "*[Rate limit reached. Retrying in " .. stop_opts.retry_after .. " seconds ...]*"
if opts.on_chunk then opts.on_chunk("\n" .. msg_content .. "\n") end
local message
if opts.on_messages_add then
message = HistoryMessage:new({
role = "assistant",
content = "\n\n" .. msg_content,
}, {
just_for_display = true,
})
opts.on_messages_add({ message })
end
local timer = vim.loop.new_timer()
if timer then
local retry_after = stop_opts.retry_after
local function countdown()
timer:start(
1000,
0,
vim.schedule_wrap(function()
if retry_after > 0 then retry_after = retry_after - 1 end
local msg_content_ = "*[Rate limit reached. Retrying in " .. retry_after .. " seconds ...]*"
if opts.on_chunk then opts.on_chunk([[\033[1A\033[K]] .. "\n" .. msg_content_ .. "\n") end
if opts.on_messages_add and message then
message.message.content = "\n\n" .. msg_content_
opts.on_messages_add({ message })
end
countdown()
end)
)
end
countdown()
end
Utils.info("Rate limit reached. Retrying in " .. stop_opts.retry_after .. " seconds", { title = "Avante" })
vim.defer_fn(function()
if timer then timer:stop() end
M._stream(opts)
end, stop_opts.retry_after * 1000)
return
end
return opts.on_stop(stop_opts)
end,
}
return M.curl({
provider = provider,
prompt_opts = prompt_opts,
handler_opts = handler_opts,
on_response_headers = function(headers) resp_headers = headers end,
})
end
local function _merge_response(first_response, second_response, opts)
local prompt = "\n" .. Config.dual_boost.prompt
prompt = prompt
:gsub("{{[%s]*provider1_output[%s]*}}", function() return first_response end)
:gsub("{{[%s]*provider2_output[%s]*}}", function() return second_response end)
prompt = prompt .. "\n"
if opts.instructions == nil then opts.instructions = "" end
-- append this reference prompt to the prompt_opts messages at last
opts.instructions = opts.instructions .. prompt
M._stream(opts)
end
local function _collector_process_responses(collector, opts)
if not collector[1] or not collector[2] then
Utils.error("One or both responses failed to complete")
return
end
_merge_response(collector[1], collector[2], opts)
end
local function _collector_add_response(collector, index, response, opts)
collector[index] = response
collector.count = collector.count + 1
if collector.count == 2 then
collector.timer:stop()
_collector_process_responses(collector, opts)
end
end
function M._dual_boost_stream(opts, Provider1, Provider2)
Utils.debug("Starting Dual Boost Stream")
local collector = {
count = 0,
responses = {},
timer = uv.new_timer(),
timeout_ms = Config.dual_boost.timeout,
}
-- Setup timeout
collector.timer:start(
collector.timeout_ms,
0,
vim.schedule_wrap(function()
if collector.count < 2 then
Utils.warn("Dual boost stream timeout reached")
collector.timer:stop()
-- Process whatever responses we have
_collector_process_responses(collector, opts)
end
end)
)
-- Create options for both streams
local function create_stream_opts(index)
local response = ""
return vim.tbl_extend("force", opts, {
on_chunk = function(chunk)
if chunk then response = response .. chunk end
end,
on_stop = function(stop_opts)
if stop_opts.error then
Utils.error(string.format("Stream %d failed: %s", index, stop_opts.error))
return
end
Utils.debug(string.format("Response %d completed", index))
_collector_add_response(collector, index, response, opts)
end,
})
end
-- Start both streams
local success, err = xpcall(function()
local opts1 = create_stream_opts(1)
opts1.provider = Provider1
M._stream(opts1)
local opts2 = create_stream_opts(2)
opts2.provider = Provider2
M._stream(opts2)
end, function(err) return err end)
if not success then Utils.error("Failed to start dual_boost streams: " .. tostring(err)) end
end
---@param opts AvanteLLMStreamOptions
function M.stream(opts)
local is_completed = false
if opts.on_tool_log ~= nil then
local original_on_tool_log = opts.on_tool_log
opts.on_tool_log = vim.schedule_wrap(function(...)
if not original_on_tool_log then return end
return original_on_tool_log(...)
end)
end
if opts.on_chunk ~= nil then
local original_on_chunk = opts.on_chunk
opts.on_chunk = vim.schedule_wrap(function(chunk)
if is_completed then return end
if original_on_chunk then return original_on_chunk(chunk) end
end)
end
if opts.on_stop ~= nil then
local original_on_stop = opts.on_stop
opts.on_stop = vim.schedule_wrap(function(stop_opts)
if is_completed then return end
if stop_opts.reason == "complete" or stop_opts.reason == "error" or stop_opts.reason == "cancelled" then
is_completed = true
end
return original_on_stop(stop_opts)
end)
end
local valid_dual_boost_modes = {
legacy = true,
}
opts.mode = opts.mode or Config.mode
if Config.dual_boost.enabled and valid_dual_boost_modes[opts.mode] then
M._dual_boost_stream(
opts,
Providers[Config.dual_boost.first_provider],
Providers[Config.dual_boost.second_provider]
)
else
M._stream(opts)
end
end
function M.cancel_inflight_request()
if LLMToolHelpers.is_cancelled ~= nil then LLMToolHelpers.is_cancelled = true end
if LLMToolHelpers.confirm_popup ~= nil then
LLMToolHelpers.confirm_popup:cancel()
LLMToolHelpers.confirm_popup = nil
end
api.nvim_exec_autocmds("User", { pattern = M.CANCEL_PATTERN })
end
return M