Mid-run anchor check
curl --request POST \
--url https://gateway.errorbar.ai/v1/reward/sessions/{id}/anchor \
--header 'Authorization: Bearer <token>' \
--header 'Content-Type: application/json' \
--data '
{
"step": "<string>",
"items": [
{
"requestId": "<string>",
"conversation": "<string>",
"response": "<string>",
"steps": [
{
"content": "<string>",
"toolName": "<string>"
}
],
"sessionId": "<string>"
}
]
}
'import requests
url = "https://gateway.errorbar.ai/v1/reward/sessions/{id}/anchor"
payload = {
"step": "<string>",
"items": [
{
"requestId": "<string>",
"conversation": "<string>",
"response": "<string>",
"steps": [
{
"content": "<string>",
"toolName": "<string>"
}
],
"sessionId": "<string>"
}
]
}
headers = {
"Authorization": "Bearer <token>",
"Content-Type": "application/json"
}
response = requests.post(url, json=payload, headers=headers)
print(response.text)const options = {
method: 'POST',
headers: {Authorization: 'Bearer <token>', 'Content-Type': 'application/json'},
body: JSON.stringify({
step: '<string>',
items: [
{
requestId: '<string>',
conversation: '<string>',
response: '<string>',
steps: [{content: '<string>', toolName: '<string>'}],
sessionId: '<string>'
}
]
})
};
fetch('https://gateway.errorbar.ai/v1/reward/sessions/{id}/anchor', options)
.then(res => res.json())
.then(res => console.log(res))
.catch(err => console.error(err));<?php
$curl = curl_init();
curl_setopt_array($curl, [
CURLOPT_URL => "https://gateway.errorbar.ai/v1/reward/sessions/{id}/anchor",
CURLOPT_RETURNTRANSFER => true,
CURLOPT_ENCODING => "",
CURLOPT_MAXREDIRS => 10,
CURLOPT_TIMEOUT => 30,
CURLOPT_HTTP_VERSION => CURL_HTTP_VERSION_1_1,
CURLOPT_CUSTOMREQUEST => "POST",
CURLOPT_POSTFIELDS => json_encode([
'step' => '<string>',
'items' => [
[
'requestId' => '<string>',
'conversation' => '<string>',
'response' => '<string>',
'steps' => [
[
'content' => '<string>',
'toolName' => '<string>'
]
],
'sessionId' => '<string>'
]
]
]),
CURLOPT_HTTPHEADER => [
"Authorization: Bearer <token>",
"Content-Type: application/json"
],
]);
$response = curl_exec($curl);
$err = curl_error($curl);
curl_close($curl);
if ($err) {
echo "cURL Error #:" . $err;
} else {
echo $response;
}package main
import (
"fmt"
"strings"
"net/http"
"io"
)
func main() {
url := "https://gateway.errorbar.ai/v1/reward/sessions/{id}/anchor"
payload := strings.NewReader("{\n \"step\": \"<string>\",\n \"items\": [\n {\n \"requestId\": \"<string>\",\n \"conversation\": \"<string>\",\n \"response\": \"<string>\",\n \"steps\": [\n {\n \"content\": \"<string>\",\n \"toolName\": \"<string>\"\n }\n ],\n \"sessionId\": \"<string>\"\n }\n ]\n}")
req, _ := http.NewRequest("POST", url, payload)
req.Header.Add("Authorization", "Bearer <token>")
req.Header.Add("Content-Type", "application/json")
res, _ := http.DefaultClient.Do(req)
defer res.Body.Close()
body, _ := io.ReadAll(res.Body)
fmt.Println(string(body))
}HttpResponse<String> response = Unirest.post("https://gateway.errorbar.ai/v1/reward/sessions/{id}/anchor")
.header("Authorization", "Bearer <token>")
.header("Content-Type", "application/json")
.body("{\n \"step\": \"<string>\",\n \"items\": [\n {\n \"requestId\": \"<string>\",\n \"conversation\": \"<string>\",\n \"response\": \"<string>\",\n \"steps\": [\n {\n \"content\": \"<string>\",\n \"toolName\": \"<string>\"\n }\n ],\n \"sessionId\": \"<string>\"\n }\n ]\n}")
.asString();require 'uri'
require 'net/http'
url = URI("https://gateway.errorbar.ai/v1/reward/sessions/{id}/anchor")
http = Net::HTTP.new(url.host, url.port)
http.use_ssl = true
request = Net::HTTP::Post.new(url)
request["Authorization"] = 'Bearer <token>'
request["Content-Type"] = 'application/json'
request.body = "{\n \"step\": \"<string>\",\n \"items\": [\n {\n \"requestId\": \"<string>\",\n \"conversation\": \"<string>\",\n \"response\": \"<string>\",\n \"steps\": [\n {\n \"content\": \"<string>\",\n \"toolName\": \"<string>\"\n }\n ],\n \"sessionId\": \"<string>\"\n }\n ]\n}"
response = http.request(request)
puts response.read_body{
"session_id": "<string>",
"held": true,
"state": "clear",
"reason": "<string>",
"pinned_step": "<string>",
"point": {
"step": "<string>",
"at": "2023-11-07T05:31:56Z",
"n": 123,
"reward_rate": 123,
"anchor_rate": 123,
"anchor_ci": [
123
],
"anchor_corrected": true,
"masked": 123,
"gap_se": 123
},
"underpowered": "<string>",
"history": [
{
"step": "<string>",
"at": "2023-11-07T05:31:56Z",
"n": 123,
"reward_rate": 123,
"anchor_rate": 123,
"anchor_ci": [
123
],
"anchor_corrected": true,
"masked": 123,
"gap_se": 123
}
],
"items": [
{
"request_id": "<string>",
"reward_grade": 123,
"anchor_verdict": "pass",
"kind": "seam"
}
],
"counts": {
"seam": 123,
"false_fail": 123,
"agree": 123,
"masked": 123
},
"spend_micros": "<string>"
}{
"error": {
"message": "<string>",
"type": "<string>",
"code": "<string>"
}
}{
"error": {
"message": "Invalid API key",
"type": "invalid_request_error",
"code": "invalid_api_key"
}
}{
"error": {
"message": "Insufficient balance: deploying this endpoint requires at least 1h of runway. Top up and try again.",
"type": "insufficient_quota",
"code": "insufficient_balance"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"code": "<string>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"code": "<string>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"code": "<string>"
}
}Certified reward API
Mid-run anchor check
Submit the current policy’s outputs on your FROZEN anchor prompt set (same request ids every time). The server scores them with the session’s reward AND the independent anchor judge (corrected for its measured error), appends the point to the session’s history, and reads two signatures against their own noise, each on two consecutive checks: the GAP (reward read − anchor read on the same outputs, widened since the first check by more than the 95% half-width of the paired difference, never less than 5 points) and the DROP (anchor below its best by more than the 95% half-width of the difference, never less than 2 points). Gap only → grader_fooled; gap and drop → gamed; drop only → degrading; each HOLDS the session. Held sessions refuse scoring (409) until a human resumes them; pinned_step is the checkpoint to ship. Points with fewer than 20 gradable prompts hold nothing (underpowered says so). Every check also returns items — each prompt with both verdicts and its kind (seam = reward pass, anchor fail: what the policy is learning to exploit) — and counts. Hold, never kill.
POST
/
v1
/
reward
/
sessions
/
{id}
/
anchor
Mid-run anchor check
curl --request POST \
--url https://gateway.errorbar.ai/v1/reward/sessions/{id}/anchor \
--header 'Authorization: Bearer <token>' \
--header 'Content-Type: application/json' \
--data '
{
"step": "<string>",
"items": [
{
"requestId": "<string>",
"conversation": "<string>",
"response": "<string>",
"steps": [
{
"content": "<string>",
"toolName": "<string>"
}
],
"sessionId": "<string>"
}
]
}
'import requests
url = "https://gateway.errorbar.ai/v1/reward/sessions/{id}/anchor"
payload = {
"step": "<string>",
"items": [
{
"requestId": "<string>",
"conversation": "<string>",
"response": "<string>",
"steps": [
{
"content": "<string>",
"toolName": "<string>"
}
],
"sessionId": "<string>"
}
]
}
headers = {
"Authorization": "Bearer <token>",
"Content-Type": "application/json"
}
response = requests.post(url, json=payload, headers=headers)
print(response.text)const options = {
method: 'POST',
headers: {Authorization: 'Bearer <token>', 'Content-Type': 'application/json'},
body: JSON.stringify({
step: '<string>',
items: [
{
requestId: '<string>',
conversation: '<string>',
response: '<string>',
steps: [{content: '<string>', toolName: '<string>'}],
sessionId: '<string>'
}
]
})
};
fetch('https://gateway.errorbar.ai/v1/reward/sessions/{id}/anchor', options)
.then(res => res.json())
.then(res => console.log(res))
.catch(err => console.error(err));<?php
$curl = curl_init();
curl_setopt_array($curl, [
CURLOPT_URL => "https://gateway.errorbar.ai/v1/reward/sessions/{id}/anchor",
CURLOPT_RETURNTRANSFER => true,
CURLOPT_ENCODING => "",
CURLOPT_MAXREDIRS => 10,
CURLOPT_TIMEOUT => 30,
CURLOPT_HTTP_VERSION => CURL_HTTP_VERSION_1_1,
CURLOPT_CUSTOMREQUEST => "POST",
CURLOPT_POSTFIELDS => json_encode([
'step' => '<string>',
'items' => [
[
'requestId' => '<string>',
'conversation' => '<string>',
'response' => '<string>',
'steps' => [
[
'content' => '<string>',
'toolName' => '<string>'
]
],
'sessionId' => '<string>'
]
]
]),
CURLOPT_HTTPHEADER => [
"Authorization: Bearer <token>",
"Content-Type: application/json"
],
]);
$response = curl_exec($curl);
$err = curl_error($curl);
curl_close($curl);
if ($err) {
echo "cURL Error #:" . $err;
} else {
echo $response;
}package main
import (
"fmt"
"strings"
"net/http"
"io"
)
func main() {
url := "https://gateway.errorbar.ai/v1/reward/sessions/{id}/anchor"
payload := strings.NewReader("{\n \"step\": \"<string>\",\n \"items\": [\n {\n \"requestId\": \"<string>\",\n \"conversation\": \"<string>\",\n \"response\": \"<string>\",\n \"steps\": [\n {\n \"content\": \"<string>\",\n \"toolName\": \"<string>\"\n }\n ],\n \"sessionId\": \"<string>\"\n }\n ]\n}")
req, _ := http.NewRequest("POST", url, payload)
req.Header.Add("Authorization", "Bearer <token>")
req.Header.Add("Content-Type", "application/json")
res, _ := http.DefaultClient.Do(req)
defer res.Body.Close()
body, _ := io.ReadAll(res.Body)
fmt.Println(string(body))
}HttpResponse<String> response = Unirest.post("https://gateway.errorbar.ai/v1/reward/sessions/{id}/anchor")
.header("Authorization", "Bearer <token>")
.header("Content-Type", "application/json")
.body("{\n \"step\": \"<string>\",\n \"items\": [\n {\n \"requestId\": \"<string>\",\n \"conversation\": \"<string>\",\n \"response\": \"<string>\",\n \"steps\": [\n {\n \"content\": \"<string>\",\n \"toolName\": \"<string>\"\n }\n ],\n \"sessionId\": \"<string>\"\n }\n ]\n}")
.asString();require 'uri'
require 'net/http'
url = URI("https://gateway.errorbar.ai/v1/reward/sessions/{id}/anchor")
http = Net::HTTP.new(url.host, url.port)
http.use_ssl = true
request = Net::HTTP::Post.new(url)
request["Authorization"] = 'Bearer <token>'
request["Content-Type"] = 'application/json'
request.body = "{\n \"step\": \"<string>\",\n \"items\": [\n {\n \"requestId\": \"<string>\",\n \"conversation\": \"<string>\",\n \"response\": \"<string>\",\n \"steps\": [\n {\n \"content\": \"<string>\",\n \"toolName\": \"<string>\"\n }\n ],\n \"sessionId\": \"<string>\"\n }\n ]\n}"
response = http.request(request)
puts response.read_body{
"session_id": "<string>",
"held": true,
"state": "clear",
"reason": "<string>",
"pinned_step": "<string>",
"point": {
"step": "<string>",
"at": "2023-11-07T05:31:56Z",
"n": 123,
"reward_rate": 123,
"anchor_rate": 123,
"anchor_ci": [
123
],
"anchor_corrected": true,
"masked": 123,
"gap_se": 123
},
"underpowered": "<string>",
"history": [
{
"step": "<string>",
"at": "2023-11-07T05:31:56Z",
"n": 123,
"reward_rate": 123,
"anchor_rate": 123,
"anchor_ci": [
123
],
"anchor_corrected": true,
"masked": 123,
"gap_se": 123
}
],
"items": [
{
"request_id": "<string>",
"reward_grade": 123,
"anchor_verdict": "pass",
"kind": "seam"
}
],
"counts": {
"seam": 123,
"false_fail": 123,
"agree": 123,
"masked": 123
},
"spend_micros": "<string>"
}{
"error": {
"message": "<string>",
"type": "<string>",
"code": "<string>"
}
}{
"error": {
"message": "Invalid API key",
"type": "invalid_request_error",
"code": "invalid_api_key"
}
}{
"error": {
"message": "Insufficient balance: deploying this endpoint requires at least 1h of runway. Top up and try again.",
"type": "insufficient_quota",
"code": "insufficient_balance"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"code": "<string>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"code": "<string>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"code": "<string>"
}
}Authorizations
Your workspace API key, e.g. sk_sovereign_..., sent as Authorization: Bearer <key>.
Path Parameters
Body
application/json
Response
The verdict for this check.
clear — no hold. grader_fooled — the gap (reward read − anchor read on the same outputs) widened beyond its paired noise since the first check while the anchor stayed within its own: the reward passes what the anchor fails, before true quality moves. gamed — the gap widened AND the anchor fell beyond its noise from its best. degrading — the anchor fell and the gap did not widen: both instruments see the fall. Every non-clear state holds.
Available options:
clear, grader_fooled, gamed, degrading Show child attributes
Show child attributes
Show child attributes
Show child attributes
Every item of this check with both verdicts and its kind — the seam list. Not stored server-side.
Show child attributes
Show child attributes
Show child attributes
Show child attributes