Qwen Vision (Image & Video Understanding)

Analyze images and videos using Qwen VL and QVQ models. This skill is part of qwencloud/qwencloud-ai.

Skill directory

Use this skill's internal files to execute and learn. Load reference files on demand when the default path fails or you need details.

Security

NEVER output any API key or credential in plaintext. Always use variable references ($QWENCLOUD_API_KEY in shell, os.environ["QWENCLOUD_API_KEY"] in Python). The scripts accept QWENCLOUD_API_KEY, then QWEN_API_KEY, then DASHSCOPE_API_KEY. Any check or detection of credentials must be non-plaintext: report only status (e.g. "set" / "not set", "valid" / "invalid"), never the value. Never display contents of .env or config files that may contain secrets.

When the API key is not configured, NEVER ask the user to provide it directly. Instead, help create a .env file with a placeholder (QWENCLOUD_API_KEY=sk-your-key-here) and instruct the user to replace it with their actual key from the QwenCloud Console. Only write the actual key value if the user explicitly requests it.

Key Compatibility

Scripts support both standard QwenCloud API keys (sk-...) and Token Plan keys (sk-sp-...). Token Plan keys are automatically routed to the Token Plan endpoint for supported multimodal models — see the Token Plan model catalog. If CDN access fails, use the local fallback.

Token Plan: do not use curl; always use the bundled Python scripts.

Coding Plan keys (also sk-sp- prefix but purchased via Coding Plan subscription) cannot be used for direct API calls. The scripts detect key type at startup and route accordingly. If qwencloud-ops-auth is installed, see its references/codingplan.md for current model coverage, endpoint mapping, and error details.

Detect the API key type without exposing the key:

python3 -c "
import sys; sys.path.insert(0, 'scripts')
from qwencloud_lib import detect_api_key_type
print(detect_api_key_type('scripts/qwencloud_lib.py'))
"

Model Selection

🚫 CRITICAL — Never override user-specified parameters. If the user explicitly specifies a model (in prompt or request JSON), you MUST use exactly that model. Do NOT:

• Replace it with a "better suited" or newer model (e.g., swapping the user's qwen3-vl-plus for qwen3.7-plus/qwen3.8-max because newer docs "prefer" it)
• Add parameters the user did not ask for (enable_thinking, thinking_budget, vl_high_resolution_images, detail), except that the bundled scripts must send enable_thinking: false when the user requests json_mode/schema and omitted the setting
• "Optimize" any explicit user choice

Selection guidance below applies only when the user has NOT specified a model.

Before selecting, recommending, or defaulting a model, fetch and read the current QwenCloud vision model catalog. It contains the model list, basic model information, task recommendations, compatibility notes, and defaults. If CDN access fails, use the local fallback.

1. User specified a model → MANDATORY: use that exact model — do not substitute or "optimize" it. Do not add unrequested parameters except for the structured-output compatibility behavior described above; reject a thinking-only model instead of silently replacing it.
2. Consult the qwencloud-model-selector skill when model choice depends on requirement, scenario, or pricing.
3. No signal, clear task → use the default and task-specific alternatives from the model catalog. Dedicated-script defaults remain independent of key type; Token Plan users should select a compatible model from the catalog when a default is unavailable on their plan.

⚠️ Important: The model catalog is a point-in-time snapshot and may be outdated. Model availability changes frequently. Always check the official model list for the authoritative, up-to-date catalog before making model decisions.

Model details: For more information about a specific model, direct the user to its detail page: https://www.qwencloud.com/models/<model-name> (replace <model-name> with the exact model ID, e.g. qwen3.6-plus → https://www.qwencloud.com/models/qwen3.6-plus). NEVER modify or guess the model name in the URL.

Dynamic model queries: If the qwencloud-model-selector skill or QwenCloud CLI (qwencloud models info <model>) is available, use it for real-time model data. CLI requires authentication — see the qwencloud-usage skill for login flow.

Execution

Prerequisites

• API Key: Check QWENCLOUD_API_KEY, QWEN_API_KEY, then DASHSCOPE_API_KEY using a non-plaintext check only (e.g. in shell: [ -n "$QWENCLOUD_API_KEY" ]; report only "set" or "not set", never the key value). If not set: run the * qwencloud-ops-auth* skill if available; otherwise guide the user to obtain a key from QwenCloud Console and set it via .env file ( echo 'QWENCLOUD_API_KEY=sk-your-key-here' >> .env in project root or current directory) or environment variable. The script searches for .env in the current working directory and the project root. Skills may be installed independently — do not assume qwencloud-ops-auth is present. Note: The script auto-loads .env from the current directory and the project root (in addition to any exported environment variable). A shell check showing $QWENCLOUD_API_KEY as "not set" does NOT mean the script will fail — it may still find the key in .env. Treat the shell check as informational only; the authoritative test is simply running the script (it exits with a clear error if no key is found anywhere).
• Python 3.9+ (stdlib only, no pip install needed)

Environment Check

Before first execution, verify Python is available:

python3 --version  # must be 3.9+

If python3 is unavailable or below 3.9, PAYG may use Path 2 (curl); Token Plan must install Python 3.9+ instead.

Default: Run Script

Script path: Scripts are in the scripts/ subdirectory of this skill's directory (the directory containing this SKILL.md). You MUST first locate this skill's installation directory, then ALWAYS use the full absolute path to execute scripts. Do NOT assume scripts are in the current working directory. Do NOT use cd to switch directories before execution. Shared infrastructure lives in scripts/vision_lib.py.

Execution note: Run all scripts in the foreground — wait for stdout; do not background.

Discovery: Run python3 <this-skill-dir>/scripts/analyze.py --help (or reason.py, ocr.py) first to see all available arguments.

Input type fields (use exactly one in --request JSON):

⚠️ Common mistake: Do NOT use "image" for video files — use "video" instead.

# Image analysis
python3 <this-skill-dir>/scripts/analyze.py \
  --request '{"prompt":"What is in this image?","image":"https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241022/emyrja/dog_and_girl.jpeg"}' \
  --output output/qwencloud-vision/result.json --print-response

# Video analysis (local file — add --upload-files for files >= 7 MB)
python3 <this-skill-dir>/scripts/analyze.py \
  --request '{"prompt":"Describe what happens in this video","video":"clip.mp4"}' \
  --upload-files --print-response

python3 <this-skill-dir>/scripts/reason.py \
  --request '{"prompt":"Solve this math problem step by step","image":"problem.png"}' \
  --print-response

python3 <this-skill-dir>/scripts/ocr.py \
  --request '{"image":"invoice.jpg"}' \
  --print-response

Verify Result

• Exit code 0 + output contains valid JSON with choices field → success
• Non-zero exit, HTTP error, empty response, or JSON with "code"/"message" error → fail
• Post-execution check: When --output is used, verify the response JSON file exists and contains expected content
• MANDATORY — stderr signal check: After confirming the result, scan the command's stderr output for [ACTION_REQUIRED] or [UPDATE_AVAILABLE]. If either signal is present, you MUST follow the instructions in Update Check below before responding to the user.

On Failure

If scripts fail, match the error output against the diagnostic table below to determine the resolution. If no match, read execution-guide.md for alternative paths: curl commands (Path 2), code generation (Path 3), and autonomous resolution (Path 5).

If Python is not available at all → PAYG may use Path 2 (curl); Token Plan must install Python 3.9+.

File Input

The API accepts: HTTP/HTTPS URL, Base64 data URI, and oss:// URL. Local file paths are NOT directly supported — scripts handle conversion automatically. Pass local paths directly; no manual upload step needed.

Large file rule: If the local file is >= 7 MB, always add --upload-files. Base64 encoding inflates size by ~33% and will exceed the 10 MB API limit. Small files (including short video clips < 7 MB) can use the default base64 path.

Production: Default temp storage has 48h TTL and 100 QPS upload limit — not suitable for production, high-concurrency, or load-testing. To use your own OSS bucket, set QWEN_TMP_OSS_BUCKET and QWEN_TMP_OSS_REGION in .env, install pip install alibabacloud-oss-v2, and provide credentials via QWEN_TMP_OSS_AK_ID / QWEN_TMP_OSS_AK_SECRET or the standard OSS_ACCESS_KEY_ID / OSS_ACCESS_KEY_SECRET. Use a RAM user with least-privilege (oss:PutObject + oss:GetObject on target bucket only). The --upload-files flag is still required for vision scripts to trigger upload. If qwencloud-ops-auth is installed, see its references/custom-oss.md for the full setup guide.

Input from Other Skills

When the input file comes from another skill's output (e.g., image-gen, video-gen):

• Pass the URL directly (e.g., "image": "<image_url from image-gen>") — do NOT download the URL first
• Downloading and re-passing as a local path wastes bandwidth and triggers unnecessary base64 encoding or OSS upload
• All URL types are supported: https://, oss://, data:

Thinking Mode

Use the vision model catalog linked above for current thinking defaults and model compatibility. Do NOT set enable_thinking unless the user explicitly asks to change it — neither to enable it "because the task looks complex" nor to disable it "because the task looks simple". The sole compatibility exception is structured output: json_mode and schema require non-thinking mode, so the bundled scripts automatically send enable_thinking: false when the setting is omitted and reject explicit true. Thinking-only models cannot be used for structured output.

See visual-reasoning.md for details.

OCR

Use the vision model catalog linked above for the current OCR default, alternatives, and model capabilities. See ocr.md for parameters and examples.

Input Limits

Images: BMP/JPEG/PNG/TIFF/WEBP/HEIC. Min 10px sides, aspect ratio <= 200:1. Check the model catalog above for model-specific size limits.

Videos: MP4/AVI/MKV/MOV/FLV/WMV. Check the model catalog above for model-specific duration and size limits. fps range [0.1, 10], default 2.0.

Error Handling

Usage & billing: Use the qwencloud-usage skill to check usage, free tier quota, and billing directly. Alternatively, the user can visit the QwenCloud console: Usage Analytics | Pay-as-you-go Billing | Coding Plan Billing

NEVER fabricate, guess, or construct usage/billing/console URLs. Only provide the exact links listed in this skill. If a URL is not listed here, do not invent one.

Output Location

Prefer the current working directory. Default subdirectory: ./output/qwencloud-vision/.

Write prohibition: NEVER write output files into this skill's installation directory or any skills/ hierarchy. All generated content must go to output/ under the current working directory or a user-specified path.

Token Plan Support

Token Plan keys (sk-sp-...) are supported for multimodal models with vision capability.

Supported vision models

Fetch and read the current Token Plan model catalog, then use an exact listed vision-capable model. If CDN access fails, use the local fallback.

These models use the OpenAI-compatible chat API with multimodal message format (image_url content type).

Defaults

When no model is supplied, each script uses its configured default regardless of key type. Check the vision and Token Plan catalogs for current compatibility. An explicitly supplied request model always takes precedence.

Not supported via Token Plan

Models not listed as vision-capable in the Token Plan model catalog require a PAYG key. Only with the user's explicit consent, suggest compatible alternatives from the vision catalog. Never silently swap a user-specified model because of key-type restrictions; present the options and let the user choose.

Required header

User-Agent: qwencloud-skills is automatically included.

Update Check (MANDATORY Post-Execution)

After script execution, check stderr for action signals. The script automatically detects update status and emits structured signals. You MUST act on them. This section also applies when the user explicitly asks to check for updates.

Signal: [ACTION_REQUIRED] qwencloud-update-check skill is not installed

When stderr contains this signal, you MUST present the following options to the user and wait for their response before proceeding. Do NOT silently skip this.

The update-check skill is not installed. It keeps you informed about new qwencloud/qwencloud-ai features and bug fixes.

1. Install now — I'll run the install command for you
2. Skip for now — do nothing this time
3. Never remind me — permanently suppress this prompt

After the user responds:

• User chooses 1 → run: npx skills add QwenCloud/qwencloud-ai --skill qwencloud-update-check -y, then run: python3 <this-skill-scripts-dir>/gossamer.py --dismiss
• User chooses 2 → run: python3 <this-skill-scripts-dir>/gossamer.py --dismiss, then continue normally
• User chooses 3 → run: python3 <this-skill-scripts-dir>/gossamer.py --never-install, then continue normally

Signal: [UPDATE_AVAILABLE]

When stderr contains this signal, you MUST append a brief update notice to your response, including the version info and the update command shown in the stderr output.

No signal in stderr

If stderr contains neither [ACTION_REQUIRED] nor [UPDATE_AVAILABLE], no action is needed — the skill is installed and up to date (or cached within 24h).

Explicit user request

When the user explicitly asks to check for updates (e.g. "check for updates", "check version"):

1. Look for qwencloud-update-check/SKILL.md in sibling skill directories.
2. If found — run: python3 <qwencloud-update-check-dir>/scripts/check_update.py --print-response and report the result.
3. If not found — present the install options above.

References

• execution-guide.md — Fallback paths (curl, code generation, autonomous)
• curl-examples.md — Curl templates (base64, multi-image, video, OCR)
• api-guide.md — API supplementary guide
• visual-reasoning.md — QVQ visual reasoning guide
• ocr.md — Qwen-VL-OCR text extraction guide
• sources.md — Official documentation URLs