vision-eyes
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
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": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| vision_eyesA | LOCAL image eyes for AI agents. No cloud, no API keys, no payments — Pillow+numpy only. action='analyze' gives full breakdown (size, brightness, RGB stats, dominant colors, 3x3 grid, brightness histogram, homogeneous regions with bboxes, sharpness, edge density). action='pixel' returns exact color at x,y + neighborhood. action='crop' saves an enlarged crop so the agent can look closer with its own vision. action='grid' saves a coordinate-grid overlay. action='compare' diffs two images. action='classify' gives the FINAL name of the object via local CLIP zero-shot (needs optional local-ai install, still no cloud; pass labels='a, b, c' for exact candidates, e.g. labels='T-34 tank, M46 Patton tank, Tiger tank'). Give image_path OR image_url OR image_base64 (chat attachments as base64/data-URL work). Coordinates: x 0..W-1 left->right, y 0..H-1 top->bottom. |
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 selecting the wrong tool. The internal action modes are clearly separated and described, making each use case easy to distinguish.
With a single tool, there is no naming inconsistency to penalize. 'vision_eyes' is a clear, snake_case, descriptive name that matches the server's purpose.
A single tool is on the low end of tool countsainer, but the tool is not trivial since it packs many actions into an action parameter. The unified design is reasonable, though splitting it into separate tools might improve clarity.
The single tool covers a wide range of local image needs: analysis, exact pixel lookup, cropping, grid overlays, image comparison, and classification. The main gap is that it relies on optional local-ai installation for classification and lacks some advanced image operations, but core 'eyes' functionality is well covered.