Regression: Streaming broken in Conversational Retrieval QA Chain after upgrade from 3.0.13 to 3.1.x

[estebanjosse]

Describe the bug

Streaming no longer works in a Chatflow using the Conversational Retrieval QA Chain after upgrading Flowise from v3.0.13 to v3.1.1.

The response is returned only once fully completed (no progressive streaming), whereas streaming was working correctly in v3.0.13.

The issue occurs across multiple Chat Models, so it does not appear to be provider-specific. This suggests a regression in the Conversational Retrieval QA Chain implementation.

To Reproduce

1. Run Flowise v3.1.1
2. Create a Chatflow with:
   • A Conversational Retrieval QA Chain
   • A Chat Model with streaming enabled
   • A retriever (e.g., vector store)
3. Open the chat UI
4. Send any query

Expected behavior

The response should be streamed progressively in the UI (token-by-token or chunked output), as in v3.0.13.

Screenshots

No response

Flow

No response

Use Method

Docker

Flowise Version

3.1.1

Operating System

None

Browser

None

Additional context

• ✅ Streaming works in:
  • Agentflow
  • Chatflow using Conversation Chain
• ❌ Streaming does NOT work in:
  • Chatflow using Conversational Retrieval QA Chain
• Tested with multiple LLM providers → same behavior
• No specific error in logs (streaming silently not triggered)

Possible regression source:
PR #5893 — “Feat/Reasoning Support, Migrate LangChain to v1” (Flowise 3.1.0)

This PR explicitly migrated:

• ConversationalRetrievalQAChain.ts
  from langchain/chains → @langchain/classic/chains
  and langchain/memory → @langchain/classic/memory

Since:

• streaming still works in Agentflow
• streaming still works with Conversation Chain
• the issue is isolated to Conversational Retrieval QA Chain

it seems likely the regression is related to how streaming callbacks / handlers are wired after the LangChain v1 migration for this specific chain.

[estebanjosse]

I’d be happy to work on a fix for this if it’s not already being handled.

[mittayrmalak-boop]

Thanks for the detailed report and debugging. We'll investigate this regression.

…

On Fri, Mar 27, 2026, 5:29 PM Esteban Josse ***@***.***> wrote: *estebanjosse* left a comment (FlowiseAI/Flowise#6070) <#6070 (comment)> I’d be happy to work on a fix for this if it’s not already being handled. — Reply to this email directly, view it on GitHub <#6070?email_source=notifications&email_token=CAXJWKW6MPCDFUI7NNN7LB34S36DDA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTIMJUGU3DSNRRGIZ2M4TFMFZW63VKON2WE43DOJUWEZLEUVSXMZLOOSWGM33PORSXEX3DNRUWG2Y#issuecomment-4145696123>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/CAXJWKWTAX7B4FNTJANDFNT4S36DDAVCNFSM6AAAAACXCTKVPSVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DCNBVGY4TMMJSGM> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[estebanjosse]

Thanks @HenryHengZJ

[mittayrmalak-boop]

You're welcome!

…

On Thu, Apr 2, 2026, 4:21 AM Esteban Josse ***@***.***> wrote: *estebanjosse* left a comment (FlowiseAI/Flowise#6070) <#6070 (comment)> Thanks @HenryHengZJ <https://github.com/HenryHengZJ> — Reply to this email directly, view it on GitHub <#6070?email_source=notifications&email_token=CAXJWKTA525Q6OTEF6NFP6T4TYWK3A5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTIMJXGU4DONZVG4YKM4TFMFZW63VHMNXW23LFNZ2KKZLWMVXHJLDGN5XXIZLSL5RWY2LDNM#issuecomment-4175877570>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/CAXJWKTNMZ35Y5LCV5NUPGD4TYWK3AVCNFSM6AAAAACXCTKVPSVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DCNZVHA3TONJXGA> . You are receiving this because you commented.Message ID: ***@***.***>