feat: support openAI sdk retrieve videos

[feitianbubu]

支持openai sdk获取Sora视频
参考: https://platform.openai.com/docs/api-reference/videos/retrieve

import OpenAI from 'openai';
const client = new OpenAI();
const video = await client.videos.retrieve('video_123');
console.log(video.id);

    const video = await client.videos.retrieve(config.get('taskId'));

    console.log('✅ 视频查询成功!');
    console.log('📋 视频任务详情:', video);
    console.log(`   ID: ${video.id}`);
    console.log(`   状态: ${video.status}`);
    console.log(`   创建时间: ${new Date(video.created_at * 1000).toLocaleString()}`);

    if (video.completed_at) {
      console.log(`   完成时间: ${new Date(video.completed_at * 1000).toLocaleString()}`);
    }

Summary by CodeRabbit

• New Features
  • Video task responses now include the task’s data payload, providing richer information for clients.
  • Video detail endpoints return the original task data directly, improving fidelity and alignment with upstream responses.
  • Response behavior remains consistent for status, progress, and reasons while delivering more complete data.

[coderabbitai]

Walkthrough

Propagates video task response data into local task objects and adjusts relay response construction. For /v1/videos/ URIs, the relay now returns originTask.Data directly; otherwise it marshals a success DTO. Adds t.Data assignment during task update. No exported APIs changed.

Changes

Cohort / File(s)	Summary of Changes
Video task data propagation and relay response path controller/task_video.go, relay/relay_task.go	controller: In updateVideoSingleTask (new API parsing), assign t.Data to task.Data alongside TaskID/Status/Url/Progress/Reason. relay: Reworked videoFetchByIDRespBodyBuilder to early-return when respBody exists; for URIs starting with /v1/videos/ return originTask.Data directly, else marshal TaskModel2Dto(originTask); updated error path label on marshal failure.

Sequence Diagram(s)

sequenceDiagram
  autonumber
  actor Client
  participant Relay as Relay (relay_task.go)
  participant Controller as Controller (task_video.go)
  participant Origin as Origin Service

  Client->>Relay: GET /v1/videos/{id}
  Relay->>Origin: Fetch task by ID
  Origin-->>Relay: originTask (includes Data)
  alt URI starts with /v1/videos/
    Relay-->>Client: originTask.Data (JSON)
  else Other URIs
    Relay->>Relay: Marshal TaskModel2Dto(originTask)
    Relay-->>Client: Success response (DTO)
  end

  note over Controller,Origin: During updates (new API), Controller assigns t.Data -> task.Data

Loading

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly related PRs

• veo #1792 — Similar edits to propagate task.Data and to return originTask.Data for /v1/videos/ in the relay.
• feat: vertex veo (#1450) #1659 — Related handling for video task Data, including propagation and specialized relay response logic.

Suggested reviewers

• seefs001
• Calcium-Ion

Poem

I nudge the frames with twitchy care,
Pack Data snug in tasks we share—
The relay hops a shorter lane,
For /videos/ it sends plain grain.
Two files touched, a tidy feat—
Thump-thump! My review is swift and sweet. 🐇✨

Pre-merge checks and finishing touches

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	You can run @coderabbitai generate docstrings to improve docstring coverage.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title Check	✅ Passed	The title succinctly describes the new feature of supporting video retrieval via the OpenAI SDK, which matches the core change introduced in the code and documented in the PR objectives. It is specific, concise, and free of extraneous or vague language.

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 2

📜 Review details

Configuration used: CodeRabbit UI

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between 9b551d9 and ff9f9fb.

📒 Files selected for processing (2)

• controller/task_video.go (1 hunks)
• relay/relay_task.go (1 hunks)

🧰 Additional context used

🧠 Learnings (1)

📚 Learning: 2025-08-26T09:59:00.337Z

Learnt from: Sh1n3zZ
PR: QuantumNous/new-api#1659
File: relay/relay_task.go:285-305
Timestamp: 2025-08-26T09:59:00.337Z
Learning: In controller/task_video.go, the redactVideoResponseBody function sanitizes video task responses by removing bytesBase64Encoded fields and truncating base64 strings to 256 characters to prevent large binary data from being stored in task.Data.

Applied to files:

• relay/relay_task.go
• controller/task_video.go

🧬 Code graph analysis (1)

relay/relay_task.go (1)

service/error.go (1)

• TaskErrorWrapper (139-155)

🔇 Additional comments (2)

relay/relay_task.go (2)

365-367: LGTM: Early return pattern improves code clarity.

The early return when respBody is already populated simplifies control flow and avoids unnecessary processing.

---

373-379: LGTM: Proper error handling for marshaling failures.

The fallback logic correctly marshals a TaskResponse and handles JSON marshaling errors appropriately.

[coderabbitai]

⚠️ Potential issue | 🔴 Critical

🧩 Analysis chain

Verify data sanitization for New API response format.

The Data field is now propagated from the New API response, but unlike the alternative parsing path (line 104) which applies redactVideoResponseBody to sanitize large binary data, this path stores t.Data directly without sanitization. This inconsistency could lead to large base64-encoded video data being stored in the database when processing New API format responses.

Verify whether the New API response format already sanitizes the Data field or if sanitization should be applied here as well:

Based on learnings.

---

🏁 Script executed:

#!/bin/bash
# Description: Check if there are any other locations where New API response data is sanitized
# or if the New API format is expected to return pre-sanitized data

# Search for sanitization patterns in New API response handling
rg -n "dto\.TaskResponse\[model\.Task\]" -A 10 -B 5

# Check if there's documentation or comments about New API data format
rg -n "New API.*Data" -C 3

Length of output: 1122

---

🏁 Script executed:

#!/bin/bash
# Locate the sanitization function
rg -n "func redactVideoResponseBody" -C3

# Inspect the legacy parsing path to see if it applies sanitization
rg -n "func ParseTaskResult" -C10

Length of output: 354

---

🏁 Script executed:

#!/bin/bash
# Show imports in controller/task_video.go to locate adaptor package path
sed -n '1,50p' controller/task_video.go

# Search for all definitions and references of ParseTaskResult
rg -n "ParseTaskResult\(" -C3

Length of output: 6385

---

Sanitize video Data for New API response
The New API path at controller/task_video.go:100 assigns t.Data directly, bypassing redactVideoResponseBody. This can store unbounded base64-encoded video data in the database. Apply redactVideoResponseBody (or equivalent truncation) here as well.

🤖 Prompt for AI Agents

In controller/task_video.go around line 100, the New API path assigns task.Data
= t.Data directly, which bypasses redactVideoResponseBody and can store
unbounded base64 video blobs; replace that direct assignment by passing t.Data
through redactVideoResponseBody (or an equivalent truncation/sanitization
function) and assign the sanitized result to task.Data so only a
truncated/redacted payload is stored.

[coderabbitai]

⚠️ Potential issue | 🟠 Major

Validate originTask.Data before returning.

Directly returning originTask.Data for /v1/videos/ URIs assumes the data is valid, non-nil JSON. However, if originTask.Data is nil, empty, or contains invalid JSON (due to database corruption, incomplete task updates, or sanitization issues), the response will fail with cryptic errors since the Content-Type is set to "application/json" in line 207.

Apply this diff to add validation:

 if strings.HasPrefix(c.Request.RequestURI, "/v1/videos/") {
+  if len(originTask.Data) == 0 {
+    taskResp = service.TaskErrorWrapperLocal(errors.New("task data is empty"), "task_data_empty", http.StatusInternalServerError)
+    return
+  }
+  // Validate it's valid JSON
+  if !json.Valid(originTask.Data) {
+    taskResp = service.TaskErrorWrapperLocal(errors.New("task data is not valid JSON"), "invalid_task_data", http.StatusInternalServerError)
+    return
+  }
   respBody = originTask.Data
   return
 }

🤖 Prompt for AI Agents

In relay/relay_task.go around lines 369 to 372, the code returns originTask.Data
directly for requests under /v1/videos/ without checking it; validate
originTask.Data before returning: if originTask.Data is nil or empty, set
respBody to a safe JSON payload (e.g., []byte("null")) or return a proper JSON
error response; otherwise use json.Valid(originTask.Data) to ensure it is valid
JSON and if not valid, log a warning and again set respBody to a safe JSON
payload or return a 5xx JSON error; ensure you do not change the existing
Content-Type and avoid panic by never returning raw nil/invalid bytes.