Skip to main content
Glama
mosaic-probe-harness
by 4ltena

Server Configuration

Describes the environment variables required to run the server.

NameRequiredDescriptionDefault
RECHECKNoSet to 1 to enable crop-based re-evaluation by the Pi agent for borderline scores.0
MAX_EDGENoMaximum long-edge pixels for VLM input (only downscales, never upscales).2048
VLM_MODELNoModel name passed to llama-server.qwen3.8-flash-next
MAX_PIXELSNoMaximum total pixel count when decoding images (decompression bomb protection).100000000
RECHECK_HINoUpper bound of p_ok for triggering re-evaluation.0.9
RECHECK_LONoLower bound of p_ok for triggering re-evaluation.0.2
MAX_RENDERSNoMaximum number of renders (judgments) per file.8
MOSAIC_BINDNoListen IP address. Non-loopback addresses require MOSAIC_API_KEY.127.0.0.1
MOSAIC_PORTNoListen port number.8790
VLM_API_KEYNoOnly set if llama-server requires Bearer authentication.
MAX_IMAGE_MBNoMaximum received image size in MiB. Exceeding this returns 413 without buffering in memory.15
VLM_BASE_URLNoBase URL of llama-server (default is http://127.0.0.1:8080).http://127.0.0.1:8080
VLM_BASE_URLSNoComma-separated base URLs for multiple llama-server instances. If unset, uses VLM_BASE_URL. Trailing / and /v1 are removed and duplicates are eliminated.
MOSAIC_API_KEYNoBearer authentication API key. Required when listening on a non-loopback address.
VLM_EXTRA_BODYNoAdditional settings merged at the top level of the request JSON (JSON object string, e.g. to override chat_template_kwargs).
VLM_TIMEOUT_MSNoTimeout for each llama-server call in milliseconds.120000
QUEUE_TIMEOUT_SNoMaximum wait time in the execution queue (seconds). Exceeding returns 503.60
RECHECK_MAX_CROPSNoMaximum number of crop regions per image for re-evaluation.3
MOSAIC_ALLOW_CIDRSNoComma-separated CIDR list of allowed source IPs.127.0.0.0/8,::1/128,10.0.0.0/8,172.16.0.0/12,192.168.0.0/16
MOSAIC_CORS_ORIGINSNo`*` or comma-separated list of allowed Origins. Default: `*` when API key is set, no origins allowed when no key is set.
CLASSIFY_TEMPERATURENoTemperature parameter used to divide classification logits. Values greater than 1 smooth the probability distribution.1
MOSAIC_LAYER_PATTERNNoRegular expression for layer names considered mosaic layers (case-insensitive).モザイク|mosaic|pixel|ぼかし|blur|censor|修正|規制
VLM_SLOTS_PER_BACKENDNoConcurrent executions per backend. Should match llama-server --parallel (total slots = number of backends × this value; maximum concurrent requests).1

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

Server capabilities have not been inspected yet.

Tools

Functions exposed to the LLM to take actions

NameDescription

No tools

Prompts

Interactive templates invoked by user choice

NameDescription

No prompts

Resources

Contextual data attached and managed by the client

NameDescription

No resources