Skip to content

Phase 3: reasoning effort, image input, structured outputs - #25

Merged
kkdai merged 2 commits into
masterfrom
copilot/issue-20-phase-3-tasks
Oct 4, 2026
Merged

kkdai merged 2 commits into
masterfrom
copilot/issue-20-phase-3-tasks

Conversation

Copilot AI commented Oct 4, 2026 •

Copy link
Copy Markdown
Contributor

Roadmap Phase 3 (OpenAI new features) is still open. This PR implements the three items that need no new API surface or confirmation UX; the rest (tools, /draw, audio, RAG, web search) are left for follow-ups.

  • Reasoning effort

    • New --reasoning-effort flag (env OPENAI_REASONING_EFFORT), passed through as reasoning_effort.
    • The value is not validated client-side. Models that don't support it will get an API error.
  • Vision (/image <file>)

    • Queues a local png/jpg/gif/webp (20 MiB cap) as a base64 data URL.
    • Queued images go out with the next question as multipart content.
    • They are cleared only once an answer is received, so a failed request keeps them queued.
    • Multipart messages are persisted by /save, and /load accepts them.
  • Structured Outputs (--json-schema)

    • Accepts inline JSON or @file, validated as JSON up front.
    • Sent as response_format: json_schema with strict: true and the fixed name response.
    • strict: true means the schema must follow OpenAI's strict-mode subset (e.g. additionalProperties: false, all properties required).
chatgpt -m o4-mini --reasoning-effort high -p "..."
chatgpt --json-schema @schema.json -p "extract the fields from: ..."
# interactive
/image shot.png
what is wrong with this UI?

README is updated. One unit test covers image attachment and message construction.

Co-authored-by: kkdai <2252691+kkdai@users.noreply.github.com>
@kkdai

kkdai commented Oct 4, 2026

Copy link
Copy Markdown
Owner

@copilot 需要加上 README_zh.md 的相對應修改

Co-authored-by: kkdai <2252691+kkdai@users.noreply.github.com>

Copilot AI commented Oct 4, 2026

Copy link
Copy Markdown
Contributor Author

@copilot 需要加上 README_zh.md 的相對應修改

已在 0b30bc4 補上 README_zh.md:指令清單加入 /image <file>,並新增與英文版對應的「Reasoning、圖片輸入與結構化輸出」段落。

Copilot AI requested a review from kkdai October 4, 2026 06:48
@kkdai
kkdai marked this pull request as ready for review October 4, 2026 06:48
@kkdai
kkdai merged commit ed78f8f into master Oct 4, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants