Hacktoberfest 2026: le issue che i maintainer hanno segnato per ottobre, aperte e adatte ai principianti. Sfoglia le issue Hacktoberfest

BedrockAgentCore: invoke_agent_runtime buffers entire response instead of streaming in real-time

Aperta
  • Qualcosa non va?
#3,348 6 commenti 1 reazione 0 assegnatari Vedi su GitHub

I maintainer di solito rispondono entro 1 giorno

Nessuno ha ancora preso questa issue.

Valutazione

Difficoltà
4/5
Tempo stimato
3-5 giorni
Idoneità per principianti
48/100
Tipo di issue
Bug
Chiarezza
Abbastanza chiara
Stato di attività
Tranquilla
Stack tecnologico
aws, ruby
Ambito
api, backend

Direzione di ricerca

Inizia dal punto di ingresso invoke_agent_runtime descritto nell’issue e confrontalo con invoke_model_with_response_stream di BedrockRuntime e il relativo supporto per event_stream_handler. Riproduci il comportamento temporale con lo script Ruby fornito, quindi traccia come resp.response.each riceve i chunks. Il lavoro è completato quando i chunks vengono forniti progressivamente durante la risposta, invece che dopo il completamento della risposta completa.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

documentation p3 queued
Describe the bug

The invoke_agent_runtime method buffers the entire EventStream response before yielding any chunks to resp.response.each, completely defeating the purpose of streaming. All chunks arrive simultaneously after the full response is generated, making real-time streaming UX impossible.

Regression Issue
  • Select this option if this issue appears to be a regression.
Expected Behavior

Chunks should arrive progressively as the AWS agent generates them, similar to BedrockRuntime's invoke_model_with_response_stream:

  • Chunk 1 arrives at ~150ms
  • Chunk 2 arrives at ~450ms
  • Chunk 3 arrives at ~1200ms
  • etc.

This enables real-time UX like typing indicators and progressive responses in chatbot applications.

Current Behavior

All chunks arrive within 1-2ms after waiting for the complete response:

D, [2026-01-21T01:15:34.902128] [AwsBedrockService] Chunk 1 at 25520.21ms, length=31
D, [2026-01-21T01:15:34.902602] [AwsBedrockService] Chunk 3 at 25520.78ms, length=21
D, [2026-01-21T01:15:34.902683] [AwsBedrockService] Chunk 5 at 25520.87ms, length=32
D, [2026-01-21T01:15:34.902734] [AwsBedrockService] Chunk 7 at 25520.92ms, length=56

All 4 chunks delivered within 0.75ms at the 25.5-second mark, despite the agent taking 25+ seconds to generate the response. The SDK is clearly buffering internally.

Reproduction Steps
require 'aws-sdk-bedrockagentcore'
require 'securerandom'
require 'json'

# This reproduces buffering issue with BedrockAgentCore streaming
client = Aws::BedrockAgentCore::Client.new(region: 'us-west-2')

# NOTE: Replace with your actual agent runtime ARN
# Example format: arn:aws:bedrock-agentcore:us-west-2:123456789012:runtime/abc123
agent_arn = ENV.fetch('AGENT_RUNTIME_ARN')

start_time = Time.now
puts "Starting request at #{Time.now}"

resp = client.invoke_agent_runtime(
  agent_runtime_arn: agent_arn,
  runtime_session_id: SecureRandom.uuid,
  payload: JSON.generate({ input: { prompt: 'Tell me a long story' } })
)

puts "\nIterating over response stream:"
chunk_count = 0

resp.response.each do |chunk|
  chunk_count += 1
  elapsed_ms = ((Time.now - start_time) * 1000).round(2)
  puts "Chunk ##{chunk_count} at #{elapsed_ms}ms: #{chunk.bytesize} bytes"
end

puts "\nExpected: Chunks arriving progressively (e.g., 100ms, 500ms, 1000ms)"
puts "Actual: All chunks arriving at approximately the same time (buffering)"

Expected: Chunks arriving progressively (100ms, 500ms, 1000ms, etc.)
Actual: All chunks arrive at once after full response completes

Possible Solution

Add event_stream_handler support similar to aws-sdk-bedrockruntime's invoke_model_with_response_stream:

client.invoke_agent_runtime(
  agent_runtime_arn: '...',
  payload: '...',
  event_stream_handler: proc { |stream|
    stream.on_event { |event| puts event }
  }
)

Alternatively, fix resp.response.each to yield chunks as they arrive from the HTTP connection rather than buffering the entire response first.

We've confirmed that raw Net::HTTP with AWS SigV4 signing streams correctly in real-time, proving the service supports it but the SDK buffers.

Additional Information/Context

This issue makes aws-sdk-bedrockagentcore unusable for real-time applications requiring streaming UX (chatbots, live agents, etc.).

Workaround: We've implemented a monkey patch using raw Net::HTTP that successfully streams in real-time by bypassing the SDK's invoke_agent_runtime method entirely and making direct HTTP requests with streaming enabled.

The BedrockRuntime service gem (aws-sdk-bedrockruntime) has proper event streaming support via event_stream_handler, but BedrockAgentCore lacks this functionality.

Gem name ('aws-sdk', 'aws-sdk-resources' or service gems like 'aws-sdk-s3') and its version

aws-sdk-bedrockagentcore version 1.14.0

Environment details (Version of Ruby, OS environment)
  • Ruby version: 3.2.6 - Operating System: macOS (Darwin 23.1.0) - Rails: 8.0.2.1
Lingua principale
Ruby
Stelle
3.7k
Fork
1.2k
Merge medio
1g 16h
PR unite (30g)
25

Preparare l'ambiente

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di aws/aws-sdk-ruby

Tutte le issue di aws/aws-sdk-ruby

Issue simili

Altre issue su Ruby

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.