mirror of
https://github.com/discourse/discourse.git
synced 2026-08-12 03:37:13 +08:00
67f8c017e0Previously, streamed structured-output chunks in the AI helper were appended only when `present?`, so whitespace-only deltas (e.g. an escaped `\n\n` between paragraphs) were consumed from the parser buffer and silently dropped, collapsing multi-paragraph proofread results into a single paragraph (regressed in #42442; the same pattern existed in `ReportRunner` and `ChatThreadTitler`). This change appends any non-nil, non-empty chunk, preserving paragraph breaks and spaces, and adds a regression spec streaming whitespace-only deltas.
116 lines
3.1 KiB
Ruby
Vendored
116 lines
3.1 KiB
Ruby
Vendored
# frozen_string_literal: true
|
|
|
|
module DiscourseAi
|
|
module Completions
|
|
class StructuredOutput
|
|
def initialize(json_schema_properties)
|
|
@property_names = json_schema_properties.keys.map(&:to_sym)
|
|
@property_cursors =
|
|
json_schema_properties.reduce({}) do |m, (k, prop)|
|
|
m[k.to_sym] = 0 if prop[:type] == "string"
|
|
m
|
|
end
|
|
|
|
@tracked = {}
|
|
|
|
@raw_response = +""
|
|
@raw_cursor = 0
|
|
|
|
@partial_json_tracker = JsonStreamingTracker.new(self)
|
|
|
|
@type_map = {}
|
|
json_schema_properties.each { |name, prop| @type_map[name.to_sym] = prop[:type].to_sym }
|
|
|
|
@done = false
|
|
end
|
|
|
|
def to_s
|
|
# we may want to also normalize the JSON here for the broken case
|
|
@raw_response.to_s
|
|
end
|
|
|
|
# require for any implicity string conversions
|
|
def to_str
|
|
to_s
|
|
end
|
|
|
|
attr_reader :last_chunk_buffer
|
|
|
|
def <<(raw)
|
|
raise "Cannot append to a completed StructuredOutput" if @done
|
|
@raw_response << raw
|
|
@partial_json_tracker << raw
|
|
end
|
|
|
|
def finish
|
|
@done = true
|
|
end
|
|
|
|
def finished?
|
|
@done
|
|
end
|
|
|
|
def broken?
|
|
@partial_json_tracker.broken?
|
|
end
|
|
|
|
def read_buffered_property(prop_name)
|
|
if @partial_json_tracker.broken?
|
|
if @done
|
|
return nil if @type_map[prop_name.to_sym].nil?
|
|
log_broken_stream
|
|
return(
|
|
DiscourseAi::Utils::BestEffortJsonParser.extract_key(
|
|
@raw_response,
|
|
@type_map[prop_name.to_sym],
|
|
prop_name,
|
|
)
|
|
)
|
|
else
|
|
return nil
|
|
end
|
|
end
|
|
|
|
# Maybe we haven't read that part of the JSON yet.
|
|
return nil if @tracked[prop_name].nil?
|
|
|
|
# This means this property is a string and we want to return unread chunks.
|
|
if @property_cursors[prop_name].present?
|
|
unread = @tracked[prop_name][@property_cursors[prop_name]..]
|
|
@property_cursors[prop_name] = @tracked[prop_name].length
|
|
unread
|
|
else
|
|
# Ints and bools, and arrays are always returned as is.
|
|
@tracked[prop_name]
|
|
end
|
|
end
|
|
|
|
# Yields the unread chunk of a string property, if any. Reading consumes
|
|
# the buffer, so unlike a presence check at the call site this never
|
|
# drops whitespace-only chunks (e.g. a "\n\n" delta between paragraphs).
|
|
def read_buffered_property_chunk(prop_name)
|
|
chunk = read_buffered_property(prop_name)
|
|
yield chunk if !chunk.nil? && !chunk.empty?
|
|
chunk
|
|
end
|
|
|
|
def notify_progress(key, value)
|
|
key_sym = key.to_sym
|
|
return if !@property_names.include?(key_sym)
|
|
|
|
@tracked[key_sym] = value
|
|
end
|
|
|
|
private
|
|
|
|
def log_broken_stream
|
|
return if @broken_logged
|
|
@broken_logged = true
|
|
Rails.logger.warn(
|
|
"Discourse AI: structured output response was not valid JSON, " \
|
|
"falling back to best-effort parsing (#{@raw_response.bytesize} bytes)",
|
|
)
|
|
end
|
|
end
|
|
end
|
|
end
|
