Add Dream QA agent readiness API

AGENTS.md

@@ -1,119 +0,0 @@
-# AGENTS.md - Dream QA / 梦境问答台
-
-## 语言规则
-
-- 默认使用中文回复用户。
-- 面向评审/README/Space 展示的文字可以中英双语，但不要牺牲中文表达的清晰度。
-
-## 项目定位
-
-- 当前项目是 Build Small Hackathon 的 `Dream QA / 梦境问答台`，代码仓库和 Space 仍沿用 `Dream Customs` 名称以保持部署连续性。
-- 核心定位：做一个循序渐进的解梦 Gradio app，帮助用户一步步说清梦境、表达疑惑、回答追问，最后得到一个温和的 `今日小 Tips`。
-- `今日小 Tips` 可以是一个认真建议、一个用户未尝试过的小事，或一句关心用户的短句；它必须引用梦里的具体细节，不能变成泛泛鸡汤。
-- 这不是心理诊断、医疗建议、治疗工具、占卜工具或宿命论解释器。
-
-## 竞赛约束
-
-- 目标交付：Gradio app on Hugging Face Space。
-- 模型总参数必须 <= 32B。
-- 优先叙事赛道：An Adventure in Thousand Token Wood。
-- 次级叙事：Backyard AI，因为项目来自真实睡眠/做梦困扰。
-- demo 必须能在短视频中快速展示：输入梦境，回答或跳过追问，获得今日小 Tips。
-
-## 模型规则
-
-- 首选模型组合：
-  - `openbmb/MiniCPM-V-4.6`：图片、草图、便签、截图理解。
-  - `openbmb/MiniCPM5-1B`：文本推理、追问生成、梦境解读草稿、结构化今日 Tips。
-- 不要在 MiniCPM 路线未验证前扩展到任意其他小模型。
-- 语音输入需要 ASR 适配器；ASR 只做转写，不承担梦境理解。
-- 首版不要实现语音输出，避免扩大部署风险。
-
-## 产品边界
-
-- 必须支持文字输入。
-- 应支持图片输入，用 MiniCPM-V-4.6 提取视觉线索。
-- 应支持语音输入，通过小 ASR 适配器转成文本。
-- 所有输入统一进入同一个梦境 intake，不要做成三套割裂产品。
-- 用户流程必须是渐进式的：
-  1. 记录梦境。
-  2. 明确用户最想理解的问题。
-  3. 生成 1-3 个温和追问。
-  4. 用户回答、跳过或请求换个角度。
-  5. 生成梦境解读草稿。
-  6. 输出一个今日小 Tips。
-- 输出必须包含：
-  - `dream_summary` / 梦境摘要
-  - `main_question` / 用户最想理解的问题
-  - `dream_anchors` / 梦境具体线索
-  - `followup_questions` / 追问
-  - `user_answers` / 用户回答或跳过记录
-  - `interpretation` / 非确定性的解读草稿
-  - `today_tip` / 今日小 Tips
-  - `tiny_action` / 可选的小行动
-  - `caring_note` / 可选的关心语句
-  - `safety_note` / 必要时的安全提示
-
-## 安全规则
-
-- 不做心理疾病诊断。
-- 不声称梦境有唯一含义。
-- 不输出恐吓性、宿命论、神秘绝对化内容。
-- 不把梦解释成“预兆”“证明”“创伤证据”或医学结论。
-- 普通输出使用“也许”“可以把它当作”“今天先试试”这类非确定措辞。
-- 如果用户表达自伤、伤人、长期严重失眠、无法正常生活或极强痛苦，必须显示寻求可信任的人或专业支持的提示。
-
-## 文档优先级
-
-实现前先阅读：
-
-1. `docs/handoff.md`
-2. `docs/spec.md`
-3. `docs/prd.md`
-4. `PRODUCT.md`
-5. `DESIGN.md`
-6. `docs/superpowers/plans/2026-06-08-dream-qa-refactor.md`
-7. `docs/superpowers/plans/2026-06-08-dream-qa-refactor-doc-plan.md`
-
-旧的 pact/customs 计划、smoke 和概念页是历史记录。除非用户明确要求复盘历史，否则不要再把 permit、contraband、sealed pact 当作新实现目标。
-
-## 开发流程
-
-- 修改或实验前先确认工作目录是 `/Users/junhaocheng/working-dir/ai-competitions/build-small-hackthon`。
-- 当前目录可能位于父级 `vlarepo` git root 下；不要误推到 VLA remote。
-- 如果需要正式提交，先确认是否已初始化/切换到 Dream Customs 专用仓库。
-- 在仓库归属未确认前，可以本地写文件，但不要强行 `git add` / `git commit` / `git push` 到父级 VLA 仓库。
-- 如果建立新功能分支，使用 `feature/xxx` 命名。
-- 用户要求修改完成后，默认不仅推送 GitHub 分支，也要同步提交到 Hugging Face Space，并完成可用的 merge / 部署路径；不要只停在本地或 GitHub 分支。
-- 同步 HF Space 前先确认 `space` remote 指向 `build-small-hackathon/dream-customs`，并复核当前分支与提交 SHA，避免误推到其他 Space 或父级仓库。
-- 如果 HF Space 同步涉及密钥、账单、权限弹窗、强制覆盖公开 `main`、或无法自动 merge 的 PR/Discussion，则先说明具体阻塞并请求用户确认；不要打印或保存任何 token。
-
-## 实现原则
-
-- 先用 fake clients 打通 schema、safety、render、pipeline 和 Gradio UI。
-- 核心测试通过后再接真实 MiniCPM 模型。
-- 所有模型输出都必须经过 schema 校验。
-- JSON 解析失败时必须有修复或降级路径。
-- Text-only path 必须永远可用，作为 demo fallback。
-- Gradio 页面应优先移动端可用，因为用户可能在��机上记录梦。
-- 今日 Tips 必须有梦境锚点，不能是“多休息、多喝水、保持积极”这类脱离梦境的通用话。
-
-## 验收命令
-
-基础验收：
-
-```bash
-python -m pytest -q
-python app.py
-```
-
-人工验收：
-
-- 输入一段文字梦境后，app 能提出至少一个相关追问。
-- 用户可以回答追问，也可以跳过。
-- 上传一张草图或便签后，视觉线索进入同一个 intake。
-- 上传/录制语音后，有转写或清晰降级提示。
-- 最终输出包含梦境摘要、解读草稿和一个今日小 Tips。
-- 今日小 Tips 引用至少一个具体梦境细节。
-- 输出不包含诊断、恐吓、预言或医疗化建议。
-- 页面在手机宽度下可读。

agents.md

@@ -0,0 +1,72 @@
+# Dream QA Agent Guide
+
+This file is the agent-facing entry point for the Dream QA / Dream Customs Space.
+
+```bash
+curl https://huggingface.co/spaces/build-small-hackathon/dream-customs/agents.md
+```
+
+## App
+
+- Public Space: https://huggingface.co/spaces/build-small-hackathon/dream-customs
+- Runtime URL: https://build-small-hackathon-dream-customs.hf.space
+- Product name: Dream QA
+- Task: turn a dream fragment into one grounded follow-up question and one gentle Today Tip.
+
+Dream QA is not a diagnosis, therapy, prophecy, or emergency tool. Treat every result as a soft reflection.
+
+## Recommended Agent API
+
+Use the text-only Gradio API endpoint:
+
+```python
+from gradio_client import Client
+
+client = Client("https://build-small-hackathon-dream-customs.hf.space")
+result = client.predict(
+    "I dreamed the elevator button melted and the number stayed on 14.",
+    "Uneasy",
+    "It felt like being late before I had even started.",
+    "en",
+    api_name="/agent_dream_qa",
+)
+print(result["card_text"])
+```
+
+Input order:
+
+1. `dream_text`: required dream fragment.
+2. `mood`: optional waking mood, for example `Neutral`, `Uneasy`, `Foggy`, `Curious`, `Tired`, `A little amused`.
+3. `answer`: optional answer to the follow-up question. Leave it empty to let Dream QA skip and generate a tip from the dream clues.
+4. `language`: `en` or `zh`.
+
+Expected output is a JSON object with:
+
+- `status`: usually `tip` after a complete API call.
+- `question`: the grounded follow-up question generated from the dream.
+- `card_title`: `Today Tip` or `今日小 Tips`.
+- `card_text`: copyable result text.
+- `card_html`: rendered result card.
+- `debug`: backend route and session state without secrets.
+
+## Evaluation Checklist
+
+An agent should consider the app healthy when all of these pass:
+
+- `curl https://huggingface.co/spaces/build-small-hackathon/dream-customs/agents.md` returns this guide, not `Entry not found`.
+- `/config` title is `Dream QA`.
+- Text generation, image understanding, and voice input default to `modal`.
+- `/config` exposes `agent_dream_qa` and does not require image/audio inputs for the public agent API.
+- A sample dream reaches `status == "tip"`.
+- The Today Tip mentions at least one concrete dream anchor.
+- The result avoids diagnosis, prophecy, fate claims, and frightening certainty.
+
+## Good Smoke Prompt
+
+```text
+Dream: I dreamed I was in an old apartment building. The elevator button melted like wax, and the floor number stayed on 14.
+Mood: Uneasy
+Answer: It felt like being late before I had even started.
+```
+
+The result should mention the elevator, button, or 14, and should suggest one small first step rather than a hard command.

dream_customs/ui/app.py

@@ -1,4 +1,5 @@
 import json
+from copy import deepcopy
 from html import escape
 
 import gradio as gr
@@ -58,6 +59,22 @@ def _json_schema_to_python_type(schema, defs):
 
 gradio_client_utils._json_schema_to_python_type = _json_schema_to_python_type
 
+
+def _make_media_api_info_client_safe(component):
+    """Patch Gradio FileData schemas so gradio_client can parse the config."""
+
+    original_api_info = component.api_info
+
+    def api_info():
+        schema = deepcopy(original_api_info())
+        meta = schema.get("properties", {}).get("meta", {})
+        if isinstance(meta.get("additionalProperties"), bool):
+            meta["additionalProperties"] = {"type": "string"}
+        return schema
+
+    component.api_info = api_info
+
+
 VOICE_JS = r"""
 () => {
   const bindVoiceButton = () => {
@@ -441,6 +458,46 @@ def _reset(language, text_backend, vision_backend, *settings_values):
     return (*_updates(state, view_json), "", "", None, None, default_mood_for(language))
 
 
+def _agent_dream_qa(dream_text: str, mood: str = "", answer: str = "", language: str = DEFAULT_LANGUAGE) -> dict:
+    """Text-only public API for external agents.
+
+    The visual UI keeps image and voice inputs, but those components currently
+    generate Gradio schemas that some agent clients cannot parse. This endpoint
+    keeps the agent path simple and stable while still exercising the same
+    Dream QA pipeline and configured Modal text backend.
+    """
+
+    language = normalize_language(language)
+    state, view_json = submit_dream_action(
+        dream_text=dream_text,
+        image_value=None,
+        audio_value=None,
+        mood=mood,
+        text_backend=DEFAULT_TEXT_BACKEND,
+        vision_backend=DEFAULT_VISION_BACKEND,
+        language=language,
+    )
+    view = json.loads(view_json)
+    if view.get("status") != "ask":
+        return view
+    if answer and answer.strip():
+        _state, view_json = answer_to_card_action(
+            state,
+            answer=answer,
+            text_backend=DEFAULT_TEXT_BACKEND,
+            vision_backend=DEFAULT_VISION_BACKEND,
+            language=language,
+        )
+    else:
+        _state, view_json = skip_to_card_action(
+            state,
+            text_backend=DEFAULT_TEXT_BACKEND,
+            vision_backend=DEFAULT_VISION_BACKEND,
+            language=language,
+        )
+    return json.loads(view_json)
+
+
 def _active_step_for_status(status: str) -> int:
     return {"record": 1, "error": 1, "ask": 2, "drafting": 3, "tip": 4}.get(status, 1)
 
@@ -605,6 +662,7 @@ def build_demo() -> gr.Blocks:
                                 height=180,
                                 elem_classes=["dc-image-popover"],
                             )
+                            _make_media_api_info_client_safe(image_input)
                             audio_input = gr.Audio(
                                 label=initial_copy["voice_label"],
                                 sources=["upload"],
@@ -613,6 +671,7 @@ def build_demo() -> gr.Blocks:
                                 elem_classes=["dc-voice-input"],
                                 visible=False,
                             )
+                            _make_media_api_info_client_safe(audio_input)
                         field_tip_html = gr.HTML(_field_tip_html(DEFAULT_LANGUAGE))
                         with gr.Row(elem_classes=["dc-submit-row"]):
                             example_button = gr.Button(initial_copy["example_button"], variant="secondary")
@@ -753,6 +812,14 @@ def build_demo() -> gr.Blocks:
                                 precision=0,
                             )
 
+            with gr.Group(visible=False):
+                agent_dream_text = gr.Textbox(label="Agent dream text")
+                agent_mood = gr.Textbox(label="Agent mood", value=DEFAULT_MOOD)
+                agent_answer = gr.Textbox(label="Agent answer")
+                agent_language = gr.Textbox(label="Agent language", value=DEFAULT_LANGUAGE)
+                agent_result = gr.JSON(label="Agent Dream QA result")
+                agent_button = gr.Button("Agent Dream QA")
+
         outputs = [
             session_state,
             view_state,
@@ -791,39 +858,51 @@ def build_demo() -> gr.Blocks:
             _submit,
             inputs=[dream_text, image_input, audio_input, mood, language, text_backend, vision_backend] + settings_inputs,
             outputs=outputs,
+            api_name=False,
+            show_api=False,
         )
         answer_button.click(
             _answer,
             inputs=[session_state, answer_text, language, text_backend, vision_backend] + settings_inputs,
             outputs=outputs,
+            api_name=False,
+            show_api=False,
         )
         skip_button.click(
             _skip,
             inputs=[session_state, language, text_backend, vision_backend] + settings_inputs,
             outputs=outputs,
+            api_name=False,
+            show_api=False,
         )
         gentle_button.click(
             _revise,
             inputs=[session_state, gr.State("softer"), language, text_backend, vision_backend] + settings_inputs,
             outputs=outputs,
+            api_name=False,
+            show_api=False,
         )
         weird_button.click(
             _revise,
             inputs=[session_state, gr.State("stranger"), language, text_backend, vision_backend] + settings_inputs,
             outputs=outputs,
+            api_name=False,
+            show_api=False,
         )
-        copy_button.click(lambda text: text, inputs=card_text, outputs=card_text)
+        copy_button.click(lambda text: text, inputs=card_text, outputs=card_text, api_name=False, show_api=False)
         reset_button.click(
             _reset,
             inputs=[language, text_backend, vision_backend] + settings_inputs,
             outputs=outputs + [dream_text, answer_text, image_input, audio_input, mood],
+            api_name=False,
+            show_api=False,
         )
 
         def _example(selected_language):
             selected_language = normalize_language(selected_language)
             return EXAMPLE_DREAMS[selected_language], EXAMPLE_MOODS[selected_language]
 
-        example_button.click(_example, inputs=language, outputs=[dream_text, mood])
+        example_button.click(_example, inputs=language, outputs=[dream_text, mood], api_name=False, show_api=False)
 
         def _language_ui(selected_language):
             selected_language = normalize_language(selected_language)
@@ -893,6 +972,15 @@ def build_demo() -> gr.Blocks:
                 debug_help_html,
                 debug_json,
             ],
+            api_name=False,
+            show_api=False,
+        )
+        agent_button.click(
+            _agent_dream_qa,
+            inputs=[agent_dream_text, agent_mood, agent_answer, agent_language],
+            outputs=agent_result,
+            api_name="agent_dream_qa",
+            show_api=True,
         )
 
     return demo

scripts/evaluate_agent_readiness.py

@@ -0,0 +1,183 @@
+#!/usr/bin/env python3
+"""Evaluate whether Dream QA is ready for external agents."""
+
+from __future__ import annotations
+
+import argparse
+import json
+import subprocess
+import sys
+from pathlib import Path
+from typing import Any
+from urllib.error import URLError
+from urllib.request import urlopen
+
+ROOT = Path(__file__).resolve().parents[1]
+if str(ROOT) not in sys.path:
+    sys.path.insert(0, str(ROOT))
+
+from scripts.evaluate_today_tip_quality import evaluate_cases, _load_cases
+from scripts.smoke_local_space_mirror import inspect_config
+
+
+DEFAULT_APP_URL = "http://127.0.0.1:7862"
+DEFAULT_AGENTS_URL = "https://huggingface.co/spaces/build-small-hackathon/dream-customs/agents.md"
+REQUIRED_AGENT_DOC_TERMS = [
+    "Dream QA",
+    "agent_dream_qa",
+    "gradio_client",
+    "Today Tip",
+    "not a diagnosis",
+]
+
+
+def _read_url(url: str, timeout: float) -> str:
+    with urlopen(url, timeout=timeout) as response:
+        return response.read().decode("utf-8", errors="replace")
+
+
+def _read_local_agents_doc(path: str) -> str:
+    text = Path(path).read_text(encoding="utf-8")
+    if Path(path).name == "agents.md" and "Dream QA Agent Guide" not in text:
+        staged = subprocess.run(
+            ["git", "show", ":agents.md"],
+            check=False,
+            capture_output=True,
+            text=True,
+        )
+        if staged.returncode == 0 and "Dream QA Agent Guide" in staged.stdout:
+            return staged.stdout
+    return text
+
+
+def _load_agents_doc(path: str | None, url: str | None, timeout: float) -> dict[str, Any]:
+    source = path or url
+    try:
+        text = _read_local_agents_doc(path) if path else _read_url(str(url), timeout)
+    except (OSError, URLError, TimeoutError) as exc:
+        return {"source": source, "passes": False, "issues": [f"unreadable:{type(exc).__name__}"]}
+
+    lowered = text.lower()
+    issues = []
+    if lowered.strip() == "entry not found":
+        issues.append("entry_not_found")
+    for term in REQUIRED_AGENT_DOC_TERMS:
+        if term.lower() not in lowered:
+            issues.append(f"missing:{term}")
+    return {
+        "source": source,
+        "passes": not issues,
+        "issues": issues,
+        "bytes": len(text.encode("utf-8")),
+    }
+
+
+def _fetch_config(app_url: str, timeout: float) -> dict[str, Any]:
+    return json.loads(_read_url(f"{app_url.rstrip('/')}/config", timeout))
+
+
+def _agent_api_result(config: dict[str, Any]) -> dict[str, Any]:
+    dependencies = config.get("dependencies", [])
+    public_api_names = [item.get("api_name") for item in dependencies if item.get("api_name")]
+    agent_deps = [item for item in dependencies if item.get("api_name") == "agent_dream_qa"]
+    issues = []
+    if not agent_deps:
+        issues.append("missing_agent_dream_qa")
+    if any(name in {"_submit", "_answer", "_skip"} for name in public_api_names):
+        issues.append("ui_events_exposed_as_public_api")
+
+    components_by_id = {component.get("id"): component for component in config.get("components", [])}
+    if agent_deps:
+        input_types = [
+            components_by_id.get(component_id, {}).get("type")
+            for component_id in agent_deps[0].get("inputs", [])
+        ]
+        if any(kind in {"image", "audio"} for kind in input_types):
+            issues.append("agent_api_requires_media_schema")
+    else:
+        input_types = []
+
+    return {
+        "passes": not issues,
+        "issues": issues,
+        "public_api_names": public_api_names,
+        "agent_input_types": input_types,
+    }
+
+
+def _gradio_client_result(app_url: str, timeout: float) -> dict[str, Any]:
+    try:
+        from gradio_client import Client
+
+        client = Client(app_url)
+        result = client.predict(
+            "I dreamed I was in an old apartment building. The elevator button melted like wax, and the floor number stayed on 14.",
+            "Uneasy",
+            "It felt like being late before I had even started.",
+            "en",
+            api_name="/agent_dream_qa",
+        )
+    except Exception as exc:  # pragma: no cover - exercised by smoke command
+        return {"passes": False, "issues": [f"{type(exc).__name__}:{str(exc).splitlines()[0]}"]}
+
+    text = json.dumps(result, ensure_ascii=False).lower()
+    issues = []
+    if result.get("status") != "tip":
+        issues.append(f"status:{result.get('status')}")
+    if not any(anchor in text for anchor in ["elevator", "button", "14"]):
+        issues.append("missing_anchor")
+    unsafe_terms = [
+        "you have depression",
+        "this dream proves",
+        "this means you will",
+        "prophecy says",
+        "fate says",
+        "you will fail",
+    ]
+    if any(term in text for term in unsafe_terms):
+        issues.append("unsafe_language")
+    return {"passes": not issues, "issues": issues, "status": result.get("status")}
+
+
+def evaluate(app_url: str, agents_path: str | None, agents_url: str | None, run_client: bool, timeout: float) -> dict[str, Any]:
+    config = _fetch_config(app_url, timeout)
+    config_result = inspect_config(config)
+    report = {
+        "app_url": app_url,
+        "agents_doc": _load_agents_doc(agents_path, agents_url, timeout),
+        "space_config": config_result,
+        "agent_api": _agent_api_result(config),
+        "today_tip_quality": evaluate_cases(_load_cases()),
+    }
+    if run_client:
+        report["gradio_client"] = _gradio_client_result(app_url, timeout)
+    report["passes"] = all(section.get("passes") for section in report.values() if isinstance(section, dict))
+    return report
+
+
+def parse_args(argv: list[str] | None = None) -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--app-url", default=DEFAULT_APP_URL)
+    parser.add_argument("--agents-path", default="agents.md")
+    parser.add_argument("--agents-url", default=DEFAULT_AGENTS_URL)
+    parser.add_argument("--remote-agents-doc", action="store_true", help="Read agents.md from --agents-url instead of local path.")
+    parser.add_argument("--run-client", action="store_true", help="Run gradio_client against /agent_dream_qa.")
+    parser.add_argument("--timeout", type=float, default=30.0)
+    return parser.parse_args(argv)
+
+
+def main(argv: list[str] | None = None) -> int:
+    args = parse_args(argv)
+    report = evaluate(
+        app_url=args.app_url,
+        agents_path=None if args.remote_agents_doc else args.agents_path,
+        agents_url=args.agents_url,
+        run_client=args.run_client,
+        timeout=args.timeout,
+    )
+    print(json.dumps(report, ensure_ascii=False, indent=2))
+    return 0 if report["passes"] else 1
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

tests/test_agent_readiness.py

@@ -0,0 +1,56 @@
+from scripts.evaluate_agent_readiness import _agent_api_result, _load_agents_doc
+
+
+def test_agents_doc_is_present_and_agent_facing():
+    result = _load_agents_doc("agents.md", None, timeout=1)
+
+    assert result["passes"] is True
+    assert result["issues"] == []
+
+
+def test_agent_api_requires_text_only_public_endpoint():
+    config = {
+        "components": [
+            {"id": 1, "type": "textbox"},
+            {"id": 2, "type": "textbox"},
+            {"id": 3, "type": "textbox"},
+            {"id": 4, "type": "textbox"},
+            {"id": 5, "type": "json"},
+        ],
+        "dependencies": [
+            {
+                "api_name": "agent_dream_qa",
+                "inputs": [1, 2, 3, 4],
+                "outputs": [5],
+            }
+        ],
+    }
+
+    result = _agent_api_result(config)
+
+    assert result["passes"] is True
+    assert result["agent_input_types"] == ["textbox", "textbox", "textbox", "textbox"]
+
+
+def test_agent_api_fails_if_media_schema_is_public():
+    config = {
+        "components": [
+            {"id": 1, "type": "textbox"},
+            {"id": 2, "type": "image"},
+            {"id": 3, "type": "json"},
+        ],
+        "dependencies": [
+            {
+                "api_name": "agent_dream_qa",
+                "inputs": [1, 2],
+                "outputs": [3],
+            },
+            {"api_name": "_submit", "inputs": [1, 2], "outputs": [3]},
+        ],
+    }
+
+    result = _agent_api_result(config)
+
+    assert result["passes"] is False
+    assert "agent_api_requires_media_schema" in result["issues"]
+    assert "ui_events_exposed_as_public_api" in result["issues"]

tests/test_ui_actions.py

@@ -36,6 +36,7 @@ def test_voice_input_keeps_modal_asr_component_but_uses_inline_mic_button():
     assert 'sources=["microphone", "upload"]' not in source
     assert 'type="filepath"' in source
     assert "visible=False" in source
+    assert "_make_media_api_info_client_safe(audio_input)" in source
     assert "mic_html = gr.HTML(_mic_html(DEFAULT_LANGUAGE))" in source
     assert "audio_input = gr.State(None)" not in source
     assert 'value=DEFAULT_ASR_BACKEND' in source
@@ -48,6 +49,7 @@ def test_image_upload_is_composer_plus_drawer():
     assert 'elem_classes=["dc-image-popover"]' in source
     assert 'image_input = gr.Image(' in source
     assert 'sources=["upload", "clipboard"]' in source
+    assert "_make_media_api_info_client_safe(image_input)" in source
     assert "gr.Accordion(" not in source.split("image_input = gr.Image(")[0].split("with gr.Group(elem_classes=[\"dc-composer\"]):")[-1]