Surface GPT-5 Reasoning summaries & Tool call Preambles #1164

Closed
opened 2026-02-16 17:29:50 -05:00 by yindo · 2 comments
Owner

Originally created by @brittlewis12 on GitHub (Aug 7, 2025).

Originally assigned to: @thdxr on GitHub.

GPT-5 supports a couple bits of additional context that might be valuable for opencode to surface: namely, reasoning summaries (either “auto” or “detailed”) as well as explanations for each of its tool calls, which OpenAI have named “preambles”.

Surfacing these in the TUI & in web shares of conversation history would help with explaining model behavior as well as understanding latency — & with a streaming implementation, it would improve perception of latency as well.

Vercel’s AI SDK already supports reasoning summaries for responses API models, so this is broadly a matter of enabling & recording this data, and then presenting it. I’m not sure preambles are supported just yet, but I’m sure if not already it’s only a matter of time.

You could imagine supporting other reasoning models like Deepseek R1 to persist & surface their reasoning content in a similar fashion.

Figured a slash command to toggle reasoning detail could make sense, to mirror the current tool details toggle command

I have a lot of energy for this as I’m craving this richer detail myself, so I’d love to take a swing! I know you folks want to continue to own core product development, so I’m eager for any guidance or particular recommendations you might have before I throw an agent at this!

Originally created by @brittlewis12 on GitHub (Aug 7, 2025). Originally assigned to: @thdxr on GitHub. GPT-5 supports a couple bits of additional context that might be valuable for opencode to surface: namely, [reasoning summaries](https://platform.openai.com/docs/guides/reasoning#reasoning-summaries) (either “auto” or “detailed”) as well as explanations for each of its tool calls, which OpenAI have named [“preambles”](https://platform.openai.com/docs/guides/latest-model#preambles). Surfacing these in the TUI & in web shares of conversation history would help with explaining model behavior as well as understanding latency — & with a streaming implementation, it would improve perception of latency as well. Vercel’s AI SDK already [supports reasoning summaries](https://ai-sdk.dev/providers/ai-sdk-providers/openai#:~:text=reasoningSummary%20'auto'%20%7C%20'detailed,the%20reasoning%20field.) for responses API models, so this is broadly a matter of enabling & recording this data, and then presenting it. I’m not sure preambles are supported just yet, but I’m sure if not already it’s only a matter of time. You could imagine supporting other reasoning models like Deepseek R1 to persist & surface their reasoning content in a similar fashion. Figured a slash command to toggle reasoning detail could make sense, to mirror the current tool details toggle command I have a lot of energy for this as I’m craving this richer detail myself, so I’d love to take a swing! I know you folks want to continue to own core product development, so I’m eager for any guidance or particular recommendations you might have before I throw an agent at this!
yindo closed this issue 2026-02-16 17:29:50 -05:00
Author
Owner

@thdxr commented on GitHub (Aug 11, 2025):

we now pass through any additional fields
https://opencode.ai/docs/agents/#additional-options

@thdxr commented on GitHub (Aug 11, 2025): we now pass through any additional fields<br>[https://opencode.ai/docs/agents/#additional-options](https://opencode.ai/docs/agents/#additional-options)
Author
Owner

@brittlewis12 commented on GitHub (Aug 11, 2025):

Fantastic thank you! will give it a shake!

@brittlewis12 commented on GitHub (Aug 11, 2025): Fantastic thank you! will give it a shake!
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: anomalyco/opencode#1164