every other text it would not answer me so I send two text then it answers
in progress
S
Student a3d1f821-08f
Category:
Conversations & ChatType:
User submissionUser role:
studentS
Sage Investigator
Re-verification (2026-07-26):
Bug confirmed still present at commit f7e6ecf6.Root cause:
The LLM streaming client in react_agent.py:103-106
creates httpx.AsyncClient
without an explicit timeout, defaulting to httpx's 5-second read timeout. This is far too short for streaming LLM responses (where first-token latency can exceed 5s), causing intermittent ReadTimeout
errors that silently terminate the SSE stream — manifesting as the "every other message doesn't get a response" symptom.Fix:
Open PR #1391 sets explicit timeouts (10s connect, 600s read, 10s write). Not yet merged.Status:
Awaiting PR merge. Re-verified and commented on the PR.S
Sage Investigator
Re-verification (2026-07-26, run 30189028061):
Bug confirmed still present at HEAD (commit f7e6ecf6). httpx.AsyncClient()
at react_agent.py:103
and builder_llm_service.py:62
still lack explicit timeouts. Fix PR #1391 remains open. No new PR created (dedupe protocol).S
Sage Investigator
Bug re-verification (2026-07-26, run 30185778749)
Bug confirmed still present at HEAD (commit f7e6ecf6). Same root cause as prior verifications —
httpx.AsyncClient()
at react_agent.py:103
and builder_llm_service.py:62
lack explicit timeouts.Open fix PR: https://github.com/momentiq-ai/sage3c/pull/1391 — awaiting review/merge.
S
Sage Investigator
Bug re-verification (2026-07-26, run 30181963969)
Bug confirmed still present at HEAD (commit f7e6ecf6).
Root cause:
httpx.AsyncClient()
at react_agent.py:103
and builder_llm_service.py:62
lack explicit timeouts. In httpx 0.28+, the default timeout behavior can cause LLM streaming responses to hang or drop silently, producing the intermittent "no response" behavior this post describes.Fix PR:
#1391 remains open and awaiting review/merge. No new PR created — dedupe protocol followed.S
Sage Investigator
Bug re-verification (2026-07-25)
Bug confirmed still present at HEAD (commit f7e6ecf6). Root cause: the LLM streaming client in react_agent.py creates an httpx.AsyncClient without explicit timeout configuration. When the upstream LLM provider stalls, the connection hangs indefinitely and the student message goes unanswered.
An open fix PR exists: https://github.com/momentiq-ai/sage3c/pull/1391 — sets explicit httpx timeout for LLM streaming clients.
Status remains in progress — awaiting PR merge.
S
Sage Investigator
Re-verification (2026-07-25, commit f7e6ecf6, run 30173470223):
Bug confirmed still present. httpx.AsyncClient()
at react_agent.py:103
and builder_llm_service.py:62
still lack explicit timeouts. Fix PR #1391 remains open and awaiting review.S
Sage Investigator
Re-verification (2026-07-25, commit f7e6ecf6, run 30169565979):
Bug confirmed still present. httpx.AsyncClient()
at react_agent.py:103
and builder_llm_service.py:62
still lack explicit timeouts. Fix PR #1391 remains open and awaiting review. Added re-verification comment to PR.S
Sage Investigator
Re-verification (2026-07-25, commit f7e6ecf6, run 30165539506):
Bug confirmed still present. httpx.AsyncClient()
at react_agent.py:103
and builder_llm_service.py:62
still lack explicit timeouts. Fix PR #1391 remains open and awaiting review. Re-verified comment posted to the PR.S
Sage Investigator
Re-verification (2026-07-25, commit f7e6ecf6, run 30161925679):
Bug confirmed still present. httpx.AsyncClient()
at react_agent.py:103
and builder_llm_service.py:62
still lack explicit timeouts. Fix PR #1391 remains open and awaiting review. No new findings — root cause unchanged.S
Sage Investigator
Re-verification (2026-07-25, commit f7e6ecf6, run 30151720441):
Bug confirmed still present. httpx.AsyncClient()
at react_agent.py:103
and builder_llm_service.py:62
still lack explicit timeouts. Fix PR #1391 remains open and awaiting review/merge.Load More
→