Marisol Okafor
@marisol
New · Member since Nov 2025
Backend engineer. Mostly here for streaming questions.
Location: Lagos
0 followers · 0 following
- Topics created
- 4
- Posts created
- 10
- Likes given
- 21
- Likes received
- 30
- Days visited
- 0
- Solutions
- 0
Days visited count across the whole community, not just this forum.
Most active topics
- Streaming responses stop after exactly 60 seconds behind nginx
5 replies · 0 votes
- Which embedding dimension should I use for 2M short product titles?
3 replies · 0 votes
- Temperature 0 still gives different answers — is that expected?
3 replies · 0 votes
- Voice mode cuts me off mid-sentence
3 replies · 0 votes
Most liked posts
- A checklist before you give an agent write access to anything
A kill switch that does not require a deploy.
6 likes
- Which embedding dimension should I use for 2M short product titles?
We are embedding about two million product titles (5–15 words each) for semantic search. vx-embed-3 supports 256, 1024 and 3072 dimensions. Storage is a concern
6 likes
- Voice mode cuts me off mid-sentence
When I pause to think, voice mode decides I am done and starts answering. Is there a setting for how long it waits?
5 likes
- Temperature 0 still gives different answers — is that expected?
Makes sense. With seed I get the same answer 5/5 now.
3 likes
- Streaming responses stop after exactly 60 seconds behind nginx
Without proxy_buffering off nginx holds the chunks until its buffer fills, so the stream looks stuck and then arrives all at once. That was it. Both lines, relo
2 likes
Top categories
Top tags
- #embeddings 1
- #evals 1
- #python-sdk 1
- #search 1
- #streaming 1
- Voice mode cuts me off mid-sentence
Liked by Beatrix Lund
Patient mode is perfect. Thank you!
Likes received · 10/12/2026
- Voice mode cuts me off mid-sentence
Liked by Tamsin Rowe
Patient mode is perfect. Thank you!
Likes received · 10/12/2026
- Voice mode cuts me off mid-sentence
There is: Settings → Voice → Pause before replying. 'Patient' waits about twice as long. We are also testing a hold-to-talk mode.
Likes given · 10/12/2026
- Voice mode cuts me off mid-sentence
Liked by Ivo Petrov
When I pause to think, voice mode decides I am done and starts answering. Is there a setting for how long it waits?
Likes received · 10/12/2026
- Voice mode cuts me off mid-sentence
Liked by Tamsin Rowe
When I pause to think, voice mode decides I am done and starts answering. Is there a setting for how long it waits?
Likes received · 10/12/2026
- Voice mode cuts me off mid-sentence
Liked by Beatrix Lund
When I pause to think, voice mode decides I am done and starts answering. Is there a setting for how long it waits?
Likes received · 10/12/2026
- Voice mode cuts me off mid-sentence
Liked by Dario Ferri
When I pause to think, voice mode decides I am done and starts answering. Is there a setting for how long it waits?
Likes received · 10/12/2026
- Voice mode cuts me off mid-sentence
Liked by Maya Chen
When I pause to think, voice mode decides I am done and starts answering. Is there a setting for how long it waits?
Likes received · 10/12/2026
- Tracking per-user cost in a multi-tenant app
How are people attributing API cost to their own customers? We bill per seat but want to spot the 2% of users who cost 40% of the bill.
Likes given · 10/12/2026
- A checklist before you give an agent write access to anything
My expenses agent tried to file a coffee as 'client entertainment'. Item 2 saved me.
Likes given · 10/12/2026
- A checklist before you give an agent write access to anything
Liked by Quillstone Team
A kill switch that does not require a deploy.
Likes received · 10/12/2026
- A checklist before you give an agent write access to anything
Liked by Fennwick Hale
A kill switch that does not require a deploy.
Likes received · 10/12/2026
- A checklist before you give an agent write access to anything
Liked by Sofia Marchetti
A kill switch that does not require a deploy.
Likes received · 10/12/2026
- A checklist before you give an agent write access to anything
Liked by Grace Adeyemi
A kill switch that does not require a deploy.
Likes received · 10/12/2026
- A checklist before you give an agent write access to anything
Liked by Felix Ortega
A kill switch that does not require a deploy.
Likes received · 10/12/2026
- A checklist before you give an agent write access to anything
Liked by Noah Lindgren
A kill switch that does not require a deploy.
Likes received · 10/12/2026
- A checklist before you give an agent write access to anything
After answering a dozen threads about agents doing surprising things, here is the list I now send people. Dry-run mode. Every write tool has a dry_run flag that
Likes given · 10/12/2026
- Python SDK retries a non-idempotent request after a timeout
That is the answer. Pass idempotency_key per logical operation and the server returns the first result for a duplicate instead of re-running it: client.response
Likes given · 10/12/2026
- Python SDK retries a non-idempotent request after a timeout
We call client.responses.create with a tool that charges a card. When the request times out client-side, the SDK retries — and the server had actually finished
Likes given · 10/12/2026
- Temperature 0 still gives different answers — is that expected?
Liked by Beatrix Lund
Makes sense. With seed I get the same answer 5/5 now.
Likes received · 10/12/2026