mirror of
https://ghfast.top/https://github.com/discourse/discourse-ai.git
synced 2026-07-16 11:36:23 +08:00
823e8ef490
Implement streaming tool call implementation for Anthropic and Open AI.
When calling:
llm.generate(..., partial_tool_calls: true) do ...
Partials may contain ToolCall instances with partial: true, These tool calls are partially populated with json partially parsed.
So for example when performing a search you may get:
ToolCall(..., {search: "hello" })
ToolCall(..., {search: "hello world" })
The library used to parse json is:
https://github.com/dgraham/json-stream
We use a fork cause we need access to the internal buffer.
This prepares internals to perform partial tool calls, but does not implement it yet.
78 lines
2 KiB
Ruby
78 lines
2 KiB
Ruby
# frozen_string_literal: true
|
|
|
|
module DiscourseAi
|
|
module Completions
|
|
module Endpoints
|
|
class CannedResponse
|
|
CANNED_RESPONSE_ERROR = Class.new(StandardError)
|
|
|
|
def initialize(responses)
|
|
@responses = responses
|
|
@completions = 0
|
|
@dialect = nil
|
|
end
|
|
|
|
def normalize_model_params(model_params)
|
|
# max_tokens, temperature, stop_sequences are already supported
|
|
model_params
|
|
end
|
|
|
|
attr_reader :responses, :completions, :dialect
|
|
|
|
def prompt_messages
|
|
dialect.prompt.messages
|
|
end
|
|
|
|
def perform_completion!(
|
|
dialect,
|
|
_user,
|
|
_model_params,
|
|
feature_name: nil,
|
|
feature_context: nil,
|
|
partial_tool_calls: false
|
|
)
|
|
@dialect = dialect
|
|
response = responses[completions]
|
|
if response.nil?
|
|
raise CANNED_RESPONSE_ERROR,
|
|
"The number of completions you requested exceed the number of canned responses"
|
|
end
|
|
|
|
raise response if response.is_a?(StandardError)
|
|
|
|
@completions += 1
|
|
if block_given?
|
|
cancelled = false
|
|
cancel_fn = lambda { cancelled = true }
|
|
|
|
# We buffer and return tool invocations in one go.
|
|
as_array = response.is_a?(Array) ? response : [response]
|
|
as_array.each do |response|
|
|
if is_tool?(response)
|
|
yield(response, cancel_fn)
|
|
else
|
|
response.each_char do |char|
|
|
break if cancelled
|
|
yield(char, cancel_fn)
|
|
end
|
|
end
|
|
end
|
|
end
|
|
|
|
response = response.first if response.is_a?(Array) && response.length == 1
|
|
response
|
|
end
|
|
|
|
def tokenizer
|
|
DiscourseAi::Tokenizer::OpenAiTokenizer
|
|
end
|
|
|
|
private
|
|
|
|
def is_tool?(response)
|
|
response.is_a?(DiscourseAi::Completions::ToolCall)
|
|
end
|
|
end
|
|
end
|
|
end
|
|
end
|
