VisionPower
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| RUN_VISION_MODEL | No | Vision model name. | qwen3-vl-flash |
| RUN_VISION_API_KEY | Yes | API key for the configured vision provider. | |
| RUN_VISION_BASE_URL | No | Base URL without /chat/completions. | DashScope compatible endpoint |
| RUN_VISION_MAX_TOKENS | No | Max response tokens. | 2048 |
| RUN_VISION_TIMEOUT_MS | No | Upstream API timeout. | 60000 |
| RUN_VISION_ALLOWED_DIRS | No | Comma-separated directories local image_path may read from. | unrestricted |
| RUN_VISION_MAX_IMAGE_BYTES | No | Max local image size in bytes. | 20971520 |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| describe_imageA | Image-understanding tool for non-multimodal (text-only) models. Uses an external vision model to read text (OCR), describe scenes, interpret charts and diagrams, compare images, and return the results as text. Models that can understand images directly must not call this tool; use their own vision capability instead. Accepts image_path, image_url, image_base64, image_ref, or images[]. For faster, more useful answers, ask the specific question you need answered (e.g. "read the error text", "what does this chart show") instead of an open-ended "describe everything". |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 1 tool
There is only one tool, so there is no risk of an agent confusing it with another server tool. The description also clearly scopes who should use it, removing ambiguity about when it should be called.
The single tool name follows a clear verb_noun convention (describe_image), so there is no mixed naming style or inconsistency. A larger set would provide more evidence of a pattern, but nothing here violates a predictable convention.
One tool feels thin for a server named VisionPower, even though the single tool is substantial and consolidates several vision tasks. The count is borderline rather than clearly inappropriate.
The tool covers the major image-understanding needs: OCR, scene description, chart/diagram interpretation, and image comparison, with multiple input formats supported. For its stated purpose of serving vision to text-only models, there are no obvious dead ends.