Thê-Minh TRINH

Commented on Upcoming Events and News: SF & Paris Gatherings Th...·Posted inAnnouncements

Thê-Minh TRINH

The series of articles on Doctolib’s AI adoption is lit!

Commented on New Issue of 'Another One Bites the Dust' Now Live...·Posted inAnnouncements

Thê-Minh TRINH

Loved the issue! First time I here of this GTM eng role, looks like a super cool role

Posted in Feedback·

Thê-Minh TRINH

Testing LLMs and Evaluating Outputs with Prompt Forge Tooling

Prompt / agent tooling for testing various LLMs and evals Hey long time I hadn’t posted here 😛 I’d love to be able to: - Iterate and validate the output of my assistants a bit more systematically, a bit like a QA eng - Compare various models I came across this demo of an open source project that does exactly that!

3Comments

Commented on New Issue of 'Another One Bites the Dust' Newslett...·Posted inAnnouncements

Thê-Minh TRINH

Great issue!

Commented on Collaborative Dust Conversations: A Multi-Player A...·Posted inFeedback

Thê-Minh TRINH

Another idea to ease teamwork: Who are the top users in my company / in my team? How do they use Dust? Which assistant? E.g.: My boss told me that the Chief of staff used her market watch assistant and that gave her some great insights

→ public profile (public sharing permissions for some “example conversations”), assistants used / created

Commented on Page Loading Issues During Long Conversations in S...·Posted inFeedback

Thê-Minh TRINH

It looks more like page rendering issues. Having lots of visualisations loaded in the conv but not even displayed

Posted in Feedback·

Thê-Minh TRINH

Page Loading Issues During Long Conversations in Slack

Hey team! I’m having a reaaaaally long conversation, making it super slow to charge the page and then interacting with it 🐌 Starting a fresh conv is totally OK though, just letting you know 😉

3Comments

Commented on Recurrent Error with Claude 3.7 Assistants in Tabl...·Posted inHelp

Thê-Minh TRINH

Hey! I’ve had the same issue these days ERROR: Error running TablesQuery app: Error querying `anthropic:claude-3-7-sonnet-20250219`: error=Too many retries (1): [model_error(retryable=, request_id=req_01BqYJ6A7oukPBH37m9nkofmtrue)] AnthropicError: [invalid_request_error] input length and `max_tokens` exceed context limit: 141123 + 64000 > 200000, decrease input length or `max_tokens` and try again (code: tables_query_error) I’m manipulating an imported csv and asking for analyses and updates. Same observations as YoiHito:

The error happens after a series of messages in the conversation
Refreshing doesn’t work
Workaround: switching to Claude 3.5 or creating a fresh conversation with 3.7

Commented on Collaborative Dust Conversations: A Multi-Player A...·Posted inFeedback

Thê-Minh TRINH

OK first shot, not just for multi-player mode:

1.
Is this assistant well-prompted?

a.
Before using an assistant I haven't built myself, I always check how it's built. At least, how it's prompted.
b.
→ Show an overview of the assistant directly from the conversation?

2.
What were the previous (human) interactions?

a.
When sb starts a conv and passes it to you, it's quite painful to scroll through and read it all to understand what's up.
b.
→ Button to collapse the LLM answers?

3.
What specifically in the conv is worth looking at?

a.
When I share a conv, I may want to point to a specific sentence.
b.
→ Highlight + summary of the highlights? (like in a reader, e.g. Pocket)
c.
→ Link to a paragraph?

Commented on Collaborative Dust Conversations: A Multi-Player A...·Posted inFeedback

Thê-Minh TRINH

Vast topic, it’s still something I’m trying to figure out