Uma
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| UMA_CROSS_ENCODER_MODEL | No | Optional. The cross-encoder model to use for relevance scoring. Default is cross-encoder/ms-marco-MiniLM-L-6-v2. | cross-encoder/ms-marco-MiniLM-L-6-v2 |
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 |
|---|---|
| uma_filterA | Filter retrieved context down to sentences relevant to a query. Runs sentence segmentation, then a local cross-encoder
(cross-encoder/ms-marco-MiniLM-L-6-v2 by default) to score every
sentence's relevance to the query, keeps sentences scoring at or above
|
| uma_scoreA | Return per-sentence cross-encoder relevance scores for a context. Segments |
| uma_statsB | Return cumulative Uma filtering statistics for this server process. |
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 3 tools
Each tool has a clearly distinct purpose: uma_filter applies a threshold to filter context, uma_score returns raw relevance scores for inspection, and uma_stats provides cumulative statistics. There is no overlap in their functionality.
All tools share the 'uma_' prefix and use lowercase snake_case, which is predictable. However, 'stats' is a noun while 'filter' and 'score' are verbs, creating a minor deviation from a purely verb-based pattern.
With only 3 tools, the set is small but well-scoped for a focused context-filtering server. It covers the core operation, inspection, and monitoring without being unnecessarily bloated.
The domain of context filtering is fully covered: uma_filter performs the main filtering, uma_score allows inspecting per-sentence scores, and uma_stats provides the overall usage statistics. There are no obvious missing operations for this purpose.