Prompt evals over MCP: run a prompt on your dataset, score each output 1-5 with an LLM judge.
Maintain this server? Claim it
Call this server's tools from your browser. Nothing to install.
Your CompletionKit API token as: Bearer <token>. Create one under Settings > API tokens.
Requests go through mcptry because browsers can't call most MCP servers directly. Header values are sent to the server for that request only and are never stored or logged.
Tool list from our last check. Connect live to refresh it.
Upsert an agreement for (run, response, metric, created_by). Verdict is one of agree, disagree, borderline. corrected_score (1..5) is required when verdict is 'disagree'.
This tool isn't marked read-only, so it may change something on the server or in an account you connected.
Upsert an agreement for (run, response, metric, created_by). Verdict is one of agree, disagree, borderline. corrected_score (1..5) is required when verdict is 'disagree'.
| Argument | Type | Description |
|---|---|---|
| note | string | |
| run_id* | integer | |
| verdict* | string | |
| metric_id* | integer | |
| created_by | string | |
| response_id* | integer | |
| corrected_score | number |
{
"type": "object",
"required": [
"run_id",
"response_id",
"metric_id",
"verdict"
],
"properties": {
"note": {
"type": "string"
},
"run_id": {
"type": "integer"
},
"verdict": {
"enum": [
"agree",
"disagree",
"borderline"
],
"type": "string"
},
"metric_id": {
"type": "integer"
},
"created_by": {
"type": "string"
},
"response_id": {
"type": "integer"
},
"corrected_score": {
"type": "number"
}
}
}List agreements. Filter by run_id, response_id, metric_id, or created_by.
| Argument | Type | Description |
|---|---|---|
| run_id | integer | |
| metric_id | integer | |
| created_by | string | |
| response_id | integer |
{
"type": "object",
"required": [],
"properties": {
"run_id": {
"type": "integer"
},
"metric_id": {
"type": "integer"
},
"created_by": {
"type": "string"
},
"response_id": {
"type": "integer"
}
}
}Create a dataset with CSV data. First row is the header. Two column names are recognized specially: "expected_output" is each row's answer key (ground truth) given to the judge and to checks that compare against the row's expected value, and "actual_output" is a pre-made output to score in a prompt-less run. Both are overridable per run (expected_column / output_column). Every column is also available to the prompt as a variable.
| Argument | Type | Description |
|---|---|---|
| name* | string | |
| csv_data* | string | |
| tag_names | array |
{
"type": "object",
"required": [
"name",
"csv_data"
],
"properties": {
"name": {
"type": "string"
},
"csv_data": {
"type": "string"
},
"tag_names": {
"type": "array",
"items": {
"type": "string"
}
}
}
}Create a dataset by downloading CSV from a URL instead of inlining it. Use this for large datasets: pass a public http(s) URL and the server fetches the CSV directly, so the data never has to pass through the tool-call arguments. The URL is SSRF-checked and the download is capped at 10MB. First row is the header; the "expected_output" (answer key) and "actual_output" (pre-made output) columns are recognized specially, overridable per run.
| Argument | Type | Description |
|---|---|---|
| url* | string | Public http(s) URL of the CSV file to download. |
| name* | string | |
| tag_names | array |
{
"type": "object",
"required": [
"name",
"url"
],
"properties": {
"url": {
"type": "string",
"description": "Public http(s) URL of the CSV file to download."
},
"name": {
"type": "string"
},
"tag_names": {
"type": "array",
"items": {
"type": "string"
}
}
}
}Delete a dataset
| Argument | Type | Description |
|---|---|---|
| id* | integer |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
}
}
}Get a dataset by ID
| Argument | Type | Description |
|---|---|---|
| id* | integer |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
}
}
}List all datasets
No arguments.
{
"type": "object",
"required": [],
"properties": {}
}Update a dataset
| Argument | Type | Description |
|---|---|---|
| id* | integer | |
| name | string | |
| csv_data | string | |
| tag_names | array |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
},
"name": {
"type": "string"
},
"csv_data": {
"type": "string"
},
"tag_names": {
"type": "array",
"items": {
"type": "string"
}
}
}
}Compare two versions of one metric's agreement stats side by side. Requires metric_id, metric_version_a_id, and metric_version_b_id (both versions must belong to that metric). Unavailable for check metrics.
| Argument | Type | Description |
|---|---|---|
| metric_id* | integer | |
| metric_version_a_id* | integer | |
| metric_version_b_id* | integer |
{
"type": "object",
"required": [
"metric_id",
"metric_version_a_id",
"metric_version_b_id"
],
"properties": {
"metric_id": {
"type": "integer"
},
"metric_version_a_id": {
"type": "integer"
},
"metric_version_b_id": {
"type": "integer"
}
}
}Create a scoring run for the current judge over a dataset's existing outputs (wraps runs_create with prompt_id omitted and output_column supplied). This only sets up the run; call runs_generate to actually re-judge the outputs so you can compare against human verdicts.
| Argument | Type | Description |
|---|---|---|
| name* | string | |
| metric_id* | integer | |
| dataset_id* | integer | |
| judge_model* | string | |
| output_column | string | Dataset column with the existing outputs to grade. Defaults to actual_output. |
{
"type": "object",
"required": [
"name",
"metric_id",
"dataset_id",
"judge_model"
],
"properties": {
"name": {
"type": "string"
},
"metric_id": {
"type": "integer"
},
"dataset_id": {
"type": "integer"
},
"judge_model": {
"type": "string"
},
"output_column": {
"type": "string",
"description": "Dataset column with the existing outputs to grade. Defaults to actual_output."
}
}
}Create a metric group
| Argument | Type | Description |
|---|---|---|
| name* | string | |
| tag_names | array | |
| metric_ids | array | |
| description | string |
{
"type": "object",
"required": [
"name"
],
"properties": {
"name": {
"type": "string"
},
"tag_names": {
"type": "array",
"items": {
"type": "string"
}
},
"metric_ids": {
"type": "array",
"items": {
"type": "integer"
}
},
"description": {
"type": "string"
}
}
}Delete a metric group
| Argument | Type | Description |
|---|---|---|
| id* | integer |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
}
}
}Get a metric group by ID
| Argument | Type | Description |
|---|---|---|
| id* | integer |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
}
}
}List all metric groups
No arguments.
{
"type": "object",
"required": [],
"properties": {}
}Update a metric group
| Argument | Type | Description |
|---|---|---|
| id* | integer | |
| name | string | |
| tag_names | array | |
| metric_ids | array | |
| description | string |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
},
"name": {
"type": "string"
},
"tag_names": {
"type": "array",
"items": {
"type": "string"
}
},
"metric_ids": {
"type": "array",
"items": {
"type": "integer"
}
},
"description": {
"type": "string"
}
}
}Destroy a draft MetricVersion (use for either source: 'edit' or source: 'suggestion'). Published versions are refused. To demote a published version, publish a different one as current instead.
| Argument | Type | Description |
|---|---|---|
| metric_version_id* | integer |
{
"type": "object",
"required": [
"metric_version_id"
],
"properties": {
"metric_version_id": {
"type": "integer"
}
}
}List every MetricVersion (drafts + published) for a metric, newest first. Each row carries version_number, state, source, current flag, and timestamps.
| Argument | Type | Description |
|---|---|---|
| metric_id* | integer |
{
"type": "object",
"required": [
"metric_id"
],
"properties": {
"metric_id": {
"type": "integer"
}
}
}Publish a MetricVersion as the live version of its metric. Works for both 'draft → published' and 'revert to an older published version → current'. Transactionally flips current, demotes peers, and writes the version's instruction + rubric_bands back onto the metric so the judge grades against it.
| Argument | Type | Description |
|---|---|---|
| metric_version_id* | integer |
{
"type": "object",
"required": [
"metric_version_id"
],
"properties": {
"metric_version_id": {
"type": "integer"
}
}
}Create a metric with evaluation criteria. For a deterministic check set metric_type:"check" and check_config. Per-kind required keys: value (contains/not_contains/equals/list_overlap/numeric_equals), pattern (regex), json_path+expected (json_path_equals), min and/or max (length_bounds/numeric_bounds); valid_json takes no extra keys. target_path is required when target is json_path; a target_path sent with no target sets target to json_path, and one sent with any other target is refused. target_path walks array indices, as in items.0.name; any path may start at the root with $., as in $.items.0.name. list_overlap grades a list-valued field and stores a fraction rather than a verdict: score_by is recall, precision, f1 or jaccard (recall by default), and min is the fraction that counts as a pass (1, a perfect score, by default). numeric_bounds and numeric_equals read a number, ignoring the thousands commas and currency symbol a figure is written with; a comma counts only as a thousands separator, with three digits after it and no decimal point before it, so 1,5 is not a number; numeric_equals takes a tolerance, either a plain number for a fixed amount or a percentage string such as 2% for a share of the expected value. A check with no case_sensitive matches case-insensitively, regex and list_overlap included; send case_sensitive: true to match case exactly. Before 0.29.0 a regex check read a missing case_sensitive as case-sensitive. Set compare_to:"expected" on contains, not_contains, equals, json_path_equals, list_overlap, numeric_equals to grade against each row's own expected_output (ground truth) instead of a constant; drop value (or expected, for json_path_equals) and add expected_path to dig into the answer key when it is JSON.
| Argument | Type | Description |
|---|---|---|
| name* | string | |
| tag_names | array | |
| instruction | string | |
Delete a metric
| Argument | Type | Description |
|---|---|---|
| id* | integer |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
}
}
}Get a metric by ID
| Argument | Type | Description |
|---|---|---|
| id* | integer |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
}
}
}List all metrics
No arguments.
{
"type": "object",
"required": [],
"properties": {}
}Ask the model to rewrite the metric's judge instruction in N variants targeted at the recent disagreements. Each variant is saved as a draft MetricVersion with source="suggestion". Returns the persisted drafts. Stripe-metering hooks fire via ActiveSupport::Notifications under completion_kit.judge_suggestion.generated.
| Argument | Type | Description |
|---|---|---|
| count | integer | How many variants to request (default 1, max 3). One focused rewrite beats five reworded copies. |
| model | string | Override the model used to generate variants. Defaults to the configured judge model or an available judging model. |
| metric_id* | integer |
{
"type": "object",
"required": [
"metric_id"
],
"properties": {
"count": {
"type": "integer",
"description": "How many variants to request (default 1, max 3). One focused rewrite beats five reworded copies."
},
"model": {
"type": "string",
"description": "Override the model used to generate variants. Defaults to the configured judge model or an available judging model."
},
"metric_id": {
"type": "integer"
}
}
}Update a metric. For a deterministic check set metric_type:"check" and check_config. Per-kind required keys: value (contains/not_contains/equals/list_overlap/numeric_equals), pattern (regex), json_path+expected (json_path_equals), min and/or max (length_bounds/numeric_bounds); valid_json takes no extra keys. target_path is required when target is json_path; a target_path sent with no target sets target to json_path, and one sent with any other target is refused. target_path walks array indices, as in items.0.name; any path may start at the root with $., as in $.items.0.name. list_overlap grades a list-valued field and stores a fraction rather than a verdict: score_by is recall, precision, f1 or jaccard (recall by default), and min is the fraction that counts as a pass (1, a perfect score, by default). numeric_bounds and numeric_equals read a number, ignoring the thousands commas and currency symbol a figure is written with; a comma counts only as a thousands separator, with three digits after it and no decimal point before it, so 1,5 is not a number; numeric_equals takes a tolerance, either a plain number for a fixed amount or a percentage string such as 2% for a share of the expected value. A check with no case_sensitive matches case-insensitively, regex and list_overlap included; send case_sensitive: true to match case exactly. Before 0.29.0 a regex check read a missing case_sensitive as case-sensitive. Set compare_to:"expected" on contains, not_contains, equals, json_path_equals, list_overlap, numeric_equals to grade against each row's own expected_output (ground truth) instead of a constant; drop value (or expected, for json_path_equals) and add expected_path to dig into the answer key when it is JSON.
| Argument | Type | Description |
|---|---|---|
| id* | integer | |
| name | string | |
| tag_names | array | |
Import a promptfooconfig.yaml. Creates a prompt, a dataset from the test vars, and metrics from the assert blocks (llm-rubric/g-eval become judge metrics; contains/equals/regex/is-json become deterministic check metrics). Returns a summary of what mapped and what was skipped and why; nothing is dropped silently.
| Argument | Type | Description |
|---|---|---|
| config* | string | The full promptfooconfig.yaml contents. |
{
"type": "object",
"required": [
"config"
],
"properties": {
"config": {
"type": "string",
"description": "The full promptfooconfig.yaml contents."
}
}
}Create a prompt
| Argument | Type | Description |
|---|---|---|
| name* | string | |
| template* | string | |
| llm_model* | string | |
| tag_names | array | |
| description | string |
{
"type": "object",
"required": [
"name",
"template",
"llm_model"
],
"properties": {
"name": {
"type": "string"
},
"template": {
"type": "string"
},
"llm_model": {
"type": "string"
},
"tag_names": {
"type": "array",
"items": {
"type": "string"
}
},
"description": {
"type": "string"
}
}
}Delete a prompt
| Argument | Type | Description |
|---|---|---|
| id* | integer |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
}
}
}Get a prompt by ID
| Argument | Type | Description |
|---|---|---|
| id* | integer | Prompt ID |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer",
"description": "Prompt ID"
}
}
}List all prompts
No arguments.
{
"type": "object",
"required": [],
"properties": {}
}Publish a prompt version, making it the current version
| Argument | Type | Description |
|---|---|---|
| id* | integer |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
}
}
}Suggest an improved version of a prompt, grounded in a run's test results and judge feedback. Analyzes the run's responses, scores, and reviews, then returns reasoning plus a rewritten template (preserving {{variables}}) and persists it as a Suggestion. Requires a run that has a prompt (not a scoring-only run).
| Argument | Type | Description |
|---|---|---|
| run_id* | integer | The run whose results ground the improvement. |
{
"type": "object",
"required": [
"run_id"
],
"properties": {
"run_id": {
"type": "integer",
"description": "The run whose results ground the improvement."
}
}
}Update a prompt. If the prompt already has runs, this creates a new DRAFT version (current=false) rather than editing in place or publishing. Promote it with prompts_publish, so an agent's edits don't go live without a gate. If it has no runs, it is updated in place.
| Argument | Type | Description |
|---|---|---|
| id* | integer | |
| name | string | |
| template | string | |
| llm_model | string | |
| tag_names | array | |
| description | string |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
},
"name": {
"type": "string"
},
"template": {
"type": "string"
},
"llm_model": {
"type": "string"
},
"tag_names": {
"type": "array",
"items": {
"type": "string"
}
},
"description": {
"type": "string"
}
}
}Create a provider credential
| Argument | Type | Description |
|---|---|---|
| api_key* | string | |
| provider* | string | |
| api_version | string | |
| api_endpoint | string |
{
"type": "object",
"required": [
"provider",
"api_key"
],
"properties": {
"api_key": {
"type": "string"
},
"provider": {
"enum": [
"openai",
"anthropic",
"ollama",
"openrouter",
"azure_foundry"
],
"type": "string"
},
"api_version": {
"type": "string"
},
"api_endpoint": {
"type": "string"
}
}
}Delete a provider credential
| Argument | Type | Description |
|---|---|---|
| id* | integer |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
}
}
}Get a provider credential by ID (API key is not exposed)
| Argument | Type | Description |
|---|---|---|
| id* | integer |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
}
}
}List all provider credentials (API keys are not exposed)
No arguments.
{
"type": "object",
"required": [],
"properties": {}
}Update a provider credential. Changing provider or api_endpoint on a credential with a stored key requires api_key in the same call, so a stored key never moves to a new provider or endpoint.
| Argument | Type | Description |
|---|---|---|
| id* | integer | |
| api_key | string | Required when provider or api_endpoint changes and a key is stored |
| provider | string | |
| api_version | string | |
| api_endpoint | string |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
},
"api_key": {
"type": "string",
"description": "Required when provider or api_endpoint changes and a key is stored"
},
"provider": {
"type": "string"
},
"api_version": {
"type": "string"
},
"api_endpoint": {
"type": "string"
}
}
}Get a specific response
| Argument | Type | Description |
|---|---|---|
| id* | integer | |
| run_id* | integer |
{
"type": "object",
"required": [
"run_id",
"id"
],
"properties": {
"id": {
"type": "integer"
},
"run_id": {
"type": "integer"
}
}
}List responses for a run, in row order. Returns {total, limit, offset, returned, responses}. Defaults to 50 rows because full payloads are large: use "fields" to drop the bodies, "min_score"/"max_score" to isolate low scorers, and sort "score_asc" to read the worst rows first. For per-metric averages of the whole run use runs_get instead of aggregating here.
| Argument | Type | Description |
|---|---|---|
| sort | string | Row order; defaults to "id". |
| limit | integer | Rows to return; defaults to 50, capped at 500. |
| fields | array | Only return these keys, keeping the payload small. Response keys: id, run_id, input_data, response_text, expected_output, created_at, score, reviewed, reviews, status, attempts, row_index, error. Prefix with "reviews." to trim each review, e.g. ["score", "reviews.metric_name", "reviews.ai_score"]. id is always included. |
| offset | integer | Rows to skip before returning results. |
| run_id* | integer | |
| status | string | Filter by row status: pending, retrying, succeeded or failed. |
| max_score | number | Only rows whose average judge score is at most this. Use with sort "score_asc" for failure-mode analysis. |
| min_score |
Create a run. Omit prompt_id and provide output_column to score existing outputs by grading a pre-existing dataset column instead of generating new ones.
| Argument | Type | Description |
|---|---|---|
| name* | string | |
| prompt_id | integer | |
| tag_names | array | |
| dataset_id | integer | |
| max_tokens | integer | Cap on generated tokens per row. Leave unset to use the provider client's default, which is what silently truncates long outputs and makes the judge score malformed JSON. Set it to whatever the prompt uses in production so the eval matches. |
| metric_ids | array | |
| judge_model | string | |
| temperature | number | Sampling temperature for generation, 0 to 1. Leave it unset, which is the default, and no temperature is sent at all, so the model applies its own. Most current frontier models refuse the parameter outright; set it only when you are targeting a model that honours it, such as anything served locally through Ollama. A refused value is re-sent without one and the run is flagged temperature_ignored. |
Delete a run
| Argument | Type | Description |
|---|---|---|
| id* | integer |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
}
}
}Start a run. Required for every run, including score-only runs (no prompt): generates responses with the prompt when there is one, otherwise copies the graded dataset column and grades it.
| Argument | Type | Description |
|---|---|---|
| id* | integer |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
}
}
}Get a run by ID, including "metric_averages": a per-metric breakdown with each metric's average score (or pass rate for checks), how many rows it graded, and how many scored low. Use this to find the metric dragging a prompt down without listing responses.
| Argument | Type | Description |
|---|---|---|
| id* | integer |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
}
}
}List all runs
No arguments.
{
"type": "object",
"required": [],
"properties": {}
}Re-grade a run's existing responses with its currently attached metrics, without regenerating. Use after attaching or editing metrics on an already-generated run.
| Argument | Type | Description |
|---|---|---|
| id* | integer |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
}
}
}Create and start a fresh copy of a run with the same prompt, dataset, metrics, and settings. Use when the judge changed and you want a clean run instead of mixing versions.
| Argument | Type | Description |
|---|---|---|
| id* | integer |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
}
}
}Re-run only the failed responses of a run, optionally limited to specific response ids via "only".
| Argument | Type | Description |
|---|---|---|
| id* | integer | |
| only | array |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
},
"only": {
"type": "array",
"items": {
"type": "integer"
}
}
}
}Update a run
| Argument | Type | Description |
|---|---|---|
| id* | integer | |
| name | string | |
| tag_names | array | |
| dataset_id | integer | |
| max_tokens | integer | Cap on generated tokens per row. Leave unset to use the provider client's default, which is what silently truncates long outputs and makes the judge score malformed JSON. Set it to whatever the prompt uses in production so the eval matches. |
| metric_ids | array | |
| judge_model | string | |
| temperature | number | Sampling temperature for generation, 0 to 1. Leave it unset, which is the default, and no temperature is sent at all, so the model applies its own. Most current frontier models refuse the parameter outright; set it only when you are targeting a model that honours it, such as anything served locally through Ollama. A refused value is re-sent without one and the run is flagged temperature_ignored. |
| output_column | string |
Create a tag. Color is auto-assigned.
| Argument | Type | Description |
|---|---|---|
| name* | string |
{
"type": "object",
"required": [
"name"
],
"properties": {
"name": {
"type": "string"
}
}
}Delete a tag. Removes the tag from every linked metric, prompt, run, and dataset.
| Argument | Type | Description |
|---|---|---|
| id* | integer |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
}
}
}Get a tag by ID
| Argument | Type | Description |
|---|---|---|
| id* | integer |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
}
}
}List all tags
No arguments.
{
"type": "object",
"required": [],
"properties": {}
}Rename a tag.
| Argument | Type | Description |
|---|---|---|
| id* | integer | |
| name | string |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
},
"name": {
"type": "string"
}
}
}Get this organization's plan usage and limits for the current billing period: runs and prompt fetches used, their limits, how many remain, and when the period resets. Call this to pre-check quota before starting runs. Runs are hard-blocked once the run limit is reached (with a small grace band), so a run over the limit will fail with run_limit_reached.
No arguments.
{
"type": "object",
"required": [],
"properties": {}
}| metric_type |
| string |
| check_config | object |
| rubric_bands | array |
{
"type": "object",
"required": [
"name"
],
"properties": {
"name": {
"type": "string"
},
"tag_names": {
"type": "array",
"items": {
"type": "string"
}
},
"instruction": {
"type": "string"
},
"metric_type": {
"enum": [
"llm_judge",
"check"
],
"type": "string"
},
"check_config": {
"type": "object",
"properties": {
"max": {
"type": "number"
},
"min": {
"type": "number"
},
"trim": {
"type": "boolean"
},
"value": {
"type": [
"string",
"number",
"array"
],
"items": {
"type": [
"string",
"number"
]
},
"description": "Only list_overlap takes a list, as a JSON array or as comma-separated text; every other rule refuses one"
},
"target": {
"enum": [
"response_text",
"input_data",
"json_path"
],
"type": "string"
},
"pattern": {
"type": "string"
},
"expected": {},
"score_by": {
"enum": [
"recall",
"precision",
"f1",
"jaccard"
],
"type": "string"
},
"json_path": {
"type": "string"
},
"multiline": {
"type": "boolean"
},
"tolerance": {
"type": [
"number",
"string"
],
"description": "A plain number for a fixed amount, or a percentage string such as \"2%\" for a share of the expected value"
},
"check_kind": {
"enum": [
"contains",
"not_contains",
"equals",
"regex",
"valid_json",
"json_path_equals",
"length_bounds",
"list_overlap",
"numeric_bounds",
"numeric_equals"
],
"type": "string"
},
"compare_to": {
"enum": [
"constant",
"expected"
],
"type": "string"
},
"target_path": {
"type": "string"
},
"expected_path": {
"type": "string"
},
"case_sensitive": {
"type": "boolean",
"description": "A check with no case_sensitive matches case-insensitively, regex and list_overlap included; send case_sensitive: true to match case exactly. Before 0.29.0 a regex check read a missing case_sensitive as case-sensitive."
}
}
},
"rubric_bands": {
"type": "array",
"items": {
"type": "object",
"properties": {
"stars": {
"type": "integer"
},
"description": {
"type": "string"
}
}
}
}
}
}| instruction |
| string |
| metric_type | string |
| check_config | object |
| rubric_bands | array |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
},
"name": {
"type": "string"
},
"tag_names": {
"type": "array",
"items": {
"type": "string"
}
},
"instruction": {
"type": "string"
},
"metric_type": {
"enum": [
"llm_judge",
"check"
],
"type": "string"
},
"check_config": {
"type": "object",
"properties": {
"max": {
"type": "number"
},
"min": {
"type": "number"
},
"trim": {
"type": "boolean"
},
"value": {
"type": [
"string",
"number",
"array"
],
"items": {
"type": [
"string",
"number"
]
},
"description": "Only list_overlap takes a list, as a JSON array or as comma-separated text; every other rule refuses one"
},
"target": {
"enum": [
"response_text",
"input_data",
"json_path"
],
"type": "string"
},
"pattern": {
"type": "string"
},
"expected": {},
"score_by": {
"enum": [
"recall",
"precision",
"f1",
"jaccard"
],
"type": "string"
},
"json_path": {
"type": "string"
},
"multiline": {
"type": "boolean"
},
"tolerance": {
"type": [
"number",
"string"
],
"description": "A plain number for a fixed amount, or a percentage string such as \"2%\" for a share of the expected value"
},
"check_kind": {
"enum": [
"contains",
"not_contains",
"equals",
"regex",
"valid_json",
"json_path_equals",
"length_bounds",
"list_overlap",
"numeric_bounds",
"numeric_equals"
],
"type": "string"
},
"compare_to": {
"enum": [
"constant",
"expected"
],
"type": "string"
},
"target_path": {
"type": "string"
},
"expected_path": {
"type": "string"
},
"case_sensitive": {
"type": "boolean",
"description": "A check with no case_sensitive matches case-insensitively, regex and list_overlap included; send case_sensitive: true to match case exactly. Before 0.29.0 a regex check read a missing case_sensitive as case-sensitive."
}
}
},
"rubric_bands": {
"type": "array",
"items": {
"type": "object",
"properties": {
"stars": {
"type": "integer"
},
"description": {
"type": "string"
}
}
}
}
}
}| number |
| Only rows whose average judge score is at least this. |
{
"type": "object",
"required": [
"run_id"
],
"properties": {
"sort": {
"enum": [
"id",
"score_asc",
"score_desc"
],
"type": "string",
"description": "Row order; defaults to \"id\"."
},
"limit": {
"type": "integer",
"description": "Rows to return; defaults to 50, capped at 500."
},
"fields": {
"type": "array",
"items": {
"type": "string"
},
"description": "Only return these keys, keeping the payload small. Response keys: id, run_id, input_data, response_text, expected_output, created_at, score, reviewed, reviews, status, attempts, row_index, error. Prefix with \"reviews.\" to trim each review, e.g. [\"score\", \"reviews.metric_name\", \"reviews.ai_score\"]. id is always included."
},
"offset": {
"type": "integer",
"description": "Rows to skip before returning results."
},
"run_id": {
"type": "integer"
},
"status": {
"type": "string",
"description": "Filter by row status: pending, retrying, succeeded or failed."
},
"max_score": {
"type": "number",
"description": "Only rows whose average judge score is at most this. Use with sort \"score_asc\" for failure-mode analysis."
},
"min_score": {
"type": "number",
"description": "Only rows whose average judge score is at least this."
}
}
}| output_column |
| string |
| Dataset column to grade when prompt_id is omitted; defaults to "actual_output". |
| expected_column | string | Dataset column holding each row's answer key / ground truth, graded by checks with compare_to "expected" and passed to the judge; defaults to "expected_output". |
| metric_group_id | integer | Attach the metrics belonging to this metric group (its current metric_ids). Ignored when metric_ids is also given. |
| judge_temperature | number | Sampling temperature for the judge, 0 to 1. Defaults to 0 so re-judging the same output gives the same score. Raise it only to measure judge variance on purpose; any value above 0 makes the run's scores irreproducible. |
{
"type": "object",
"required": [
"name"
],
"properties": {
"name": {
"type": "string"
},
"prompt_id": {
"type": "integer"
},
"tag_names": {
"type": "array",
"items": {
"type": "string"
}
},
"dataset_id": {
"type": "integer"
},
"max_tokens": {
"type": "integer",
"description": "Cap on generated tokens per row. Leave unset to use the provider client's default, which is what silently truncates long outputs and makes the judge score malformed JSON. Set it to whatever the prompt uses in production so the eval matches."
},
"metric_ids": {
"type": "array",
"items": {
"type": "integer"
}
},
"judge_model": {
"type": "string"
},
"temperature": {
"type": "number",
"description": "Sampling temperature for generation, 0 to 1. Leave it unset, which is the default, and no temperature is sent at all, so the model applies its own. Most current frontier models refuse the parameter outright; set it only when you are targeting a model that honours it, such as anything served locally through Ollama. A refused value is re-sent without one and the run is flagged temperature_ignored."
},
"output_column": {
"type": "string",
"description": "Dataset column to grade when prompt_id is omitted; defaults to \"actual_output\"."
},
"expected_column": {
"type": "string",
"description": "Dataset column holding each row's answer key / ground truth, graded by checks with compare_to \"expected\" and passed to the judge; defaults to \"expected_output\"."
},
"metric_group_id": {
"type": "integer",
"description": "Attach the metrics belonging to this metric group (its current metric_ids). Ignored when metric_ids is also given."
},
"judge_temperature": {
"type": "number",
"description": "Sampling temperature for the judge, 0 to 1. Defaults to 0 so re-judging the same output gives the same score. Raise it only to measure judge variance on purpose; any value above 0 makes the run's scores irreproducible."
}
}
}| expected_column | string |
| metric_group_id | integer | Replace the run's metrics with those belonging to this metric group. Ignored when metric_ids is also given. |
| judge_temperature | number | Sampling temperature for the judge, 0 to 1. Defaults to 0 so re-judging the same output gives the same score. Raise it only to measure judge variance on purpose; any value above 0 makes the run's scores irreproducible. |
{
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "integer"
},
"name": {
"type": "string"
},
"tag_names": {
"type": "array",
"items": {
"type": "string"
}
},
"dataset_id": {
"type": "integer"
},
"max_tokens": {
"type": "integer",
"description": "Cap on generated tokens per row. Leave unset to use the provider client's default, which is what silently truncates long outputs and makes the judge score malformed JSON. Set it to whatever the prompt uses in production so the eval matches."
},
"metric_ids": {
"type": "array",
"items": {
"type": "integer"
}
},
"judge_model": {
"type": "string"
},
"temperature": {
"type": "number",
"description": "Sampling temperature for generation, 0 to 1. Leave it unset, which is the default, and no temperature is sent at all, so the model applies its own. Most current frontier models refuse the parameter outright; set it only when you are targeting a model that honours it, such as anything served locally through Ollama. A refused value is re-sent without one and the run is flagged temperature_ignored."
},
"output_column": {
"type": "string"
},
"expected_column": {
"type": "string"
},
"metric_group_id": {
"type": "integer",
"description": "Replace the run's metrics with those belonging to this metric group. Ignored when metric_ids is also given."
},
"judge_temperature": {
"type": "number",
"description": "Sampling temperature for the judge, 0 to 1. Defaults to 0 so re-judging the same output gives the same score. Raise it only to measure judge variance on purpose; any value above 0 makes the run's scores irreproducible."
}
}
}