Evally
Start free
Agents

Context relevance filter

Decide what an agent keeps in context: keep, compress or drop each chunk.

Scores each tool result, retrieved passage or old message for how useful it is to the current task, and drops stale results such as errors that were already fixed. Smaller contexts mean cheaper, faster and more accurate agents.

Use this judgeCustomize it

Try it live

Real calls to Jev. Pick an example or paste your own.

Examples
33
79
No verdict yet

Pick an example or paste your own input, then run it.

Under the hood

Every question and rule, in the open. Fork it to change any of them.

What it asks

Reads task and chunk, then asks 3 questions in one call.

  1. Q1UsefulnessScore

    How useful is the chunk for completing the task?

    1. 0Irrelevant
    2. 1Background only
    3. 2Useful
    4. 3Essential
  2. Q2Contains what's neededYes / no

    Does the chunk directly contain a fact, code or instruction the task needs?

  3. Q3StaleYes / no

    Is the chunk a superseded or outdated result (an old error that was since fixed, or an earlier version of a file)?

How it decides

Rules run in code, top to bottom. The first one that matches gives the verdict.

  1. 1
    If Stale ≥ 0.7Drop
  2. 2
    If Contains what's needed ≥ 0.7Keep
  3. 3
    If Usefulness ≥ 2 of 3Keep
  4. 4
    If Usefulness ≥ 1 of 3Compress
  5. ·
    DefaultDrop

Call it from your code

Every template runs as-is with an API key, no setup: POST /api/v1/judges/context-filter/run.

curl -X POST https://tryevally.com/api/v1/judges/context-filter/run \  -H "Authorization: Bearer $EVALLY_API_KEY" \  -H "Content-Type: application/json" \  -d '{  "input": {    "task": "Add pagination to the /orders API",    "chunk": "Error: Cannot find module '\''zod'\'' (resolved by running npm install zod in step 3)"  }}'

Seen in the wild

Community builds on Jev that use this pattern. Not affiliated with Evally.

More in Agents