Problem
Captain's direct OpenAI API backend currently goes through Genkit's Chat Completions transport (/v1/chat/completions). OpenAI recommends the Responses API for new projects and as the future direction for agentic applications.
This is already causing a capability gap. For GPT-5.4 and later, Chat Completions does not support function tools with a reasoning_effort other than none. PR #87 therefore has to disable reasoning effort for tool-enabled GPT-5.6 requests. That is a tactical compatibility workaround, not the desired product behavior.
Codex runtimes already use the Responses protocol and are not part of this migration.
https://developers.openai.com/api/docs/guides/migrate-to-responses#responses-benefits
Problem
Captain's direct OpenAI API backend currently goes through Genkit's Chat Completions transport (
/v1/chat/completions). OpenAI recommends the Responses API for new projects and as the future direction for agentic applications.This is already causing a capability gap. For GPT-5.4 and later, Chat Completions does not support function tools with a
reasoning_effortother thannone. PR #87 therefore has to disable reasoning effort for tool-enabled GPT-5.6 requests. That is a tactical compatibility workaround, not the desired product behavior.Codex runtimes already use the Responses protocol and are not part of this migration.