picsense
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| VIDEO_FPS | No | Frame sampling rate for video extraction (frames per second). | 1 |
| CHAT_MODEL | No | Model name for the chat provider (e.g. hy4-preview, deepseek-v4-flash). Required when provider is chat. | |
| TIMEOUT_MS | No | Request timeout for the vision model in milliseconds. | 300000 |
| CHAT_API_KEY | No | API key for the Chat Completions protocol provider. Required when provider is chat. | |
| MAX_IMAGE_MB | No | Maximum size per image in MB. | 5 |
| MAX_VIDEO_MB | No | Maximum size per video in MB. | 100 |
| OPENAI_MODEL | No | OpenAI model name (e.g. gpt-5.6-sol). Required when provider is openai. | |
| CHAT_BASE_URL | No | Compatible gateway base URL for the chat provider (e.g. https://copilot.tencent.com/v2). Automatically normalized; version suffixes like /v2 are preserved. | https://api.openai.com/v1 |
| OPENAI_API_KEY | No | OpenAI API Key. Required when provider is openai. | |
| ANTHROPIC_MODEL | No | Model name for the anthropic provider (e.g. claude-sonnet-4, deepseek-v4-flash-vision-exp). Required when provider is anthropic. | |
| OPENAI_BASE_URL | No | Custom base URL for the OpenAI provider (proxy or compatible gateway). Automatically normalized: if it lacks /v1 it is appended. | https://api.openai.com/v1 |
| DEFAULT_PROVIDER | No | Default provider to use for vision requests. One of: openai (OpenAI Responses API), chat (OpenAI Chat Completions), anthropic (Anthropic Messages API). | openai |
| VIDEO_MAX_FRAMES | No | Maximum number of frames to extract from a video. | 30 |
| ANTHROPIC_API_KEY | No | API key for the Anthropic protocol provider. Required when provider is anthropic. | |
| ANTHROPIC_BASE_URL | No | Compatible endpoint base URL for the anthropic provider (e.g. https://api.deepseek.com/anthropic). /v1 is appended automatically before requesting /v1/messages. | https://api.anthropic.com |
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 |
|---|---|
| analyze_imagesA | 识别图片内容。传入一张图片为单图识别，传入多张为批量或对比识别。支持多轮迭代——首次调用创建 session，后续调用传入 session_id 可在已有对话基础上追加提问。 |
| list_sessionsA | 查看当前所有识别会话的列表，包括每个 session 的简介、关联图片和迭代轮数。用于回顾历史识别记录。 |
| analyze_documentB | 解析文档，识别其中所有图片的内容，并在图片位置旁标注描述。返回标注后的完整文档。 |
| analyze_videoA | 识别视频内容。把视频抽帧后发送给视觉模型分析（默认每秒抽 1 帧、最多 30 帧）。支持多轮迭代——首次调用创建 session，后续调用传入 session_id 可在已有对话基础上追加提问。 |
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 4 tools
Each tool targets a distinct input type (video, image, document) or a meta concern (session listing), so an agent can generally pick correctly. The only mild overlap is analyze_document vs analyze_images, since documents contain images, but the annotated-document output distinguishes them.
All four tools use a consistent snake_case verb_noun pattern (analyze_video, analyze_images, analyze_document, list_sessions). No mixing of conventions or vague verbs.
Four tools is a compact, well-scoped surface for a media-analysis server with three input modalities plus session listing. It is a bit thin on session management operations but nothing is redundant.
Core analysis workflows are covered across video, image, and document inputs, plus session listing. However, the session lifecycle is incomplete: there is no way to fetch a single session's detail, delete/close a session, or manage stored sessions, leaving dead ends after iteration.