mirrors/pikasTech-agentrun
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix: 默认使用 gpt-5.6-sol medium

Browse Source
This commit is contained in:
AgentRun Codex
2026-07-13 04:33:40 +02:00
parent 1c15c4aaba
commit 7674c495e3
4 changed files with 32 additions and 25 deletions
Download Patch File Download Diff File Expand all files Collapse all files
config/aipods/artificer.yaml
+2 -2
View File
@@ -16,8 +16,8 @@ spec:
providerId: NC01
backendProfile: gpt-pika
model:
model: gpt-5.6-terra
reasoningEffort: xhigh
model: gpt-5.6-sol
reasoningEffort: medium
imageRef:
kind: env-image-dockerfile
repoUrl: git@github.com:pikasTech/agentrun.git
src/selftest/cases/76-aipod-spec.ts
+11 -11
View File
@@ -49,8 +49,8 @@ const selfTest: SelfTestCase = async (context) => {
assert.equal(shownItem.backendProfile, "gpt-pika");
assert.equal(shownItem.providerId, "NC01");
assert.equal(shownItem.lane, "v0.2");
assert.equal(((shownItem.model as JsonRecord).model), "gpt-5.6-terra");
assert.equal(((shownItem.model as JsonRecord).reasoningEffort), "xhigh");
assert.equal(((shownItem.model as JsonRecord).model), "gpt-5.6-sol");
assert.equal(((shownItem.model as JsonRecord).reasoningEffort), "medium");
assert.deepEqual(shownItem.gitIdentity, {
name: "AgentRun Codex",
email: "agentrun-codex@users.noreply.github.com",
@@ -91,13 +91,13 @@ const selfTest: SelfTestCase = async (context) => {
assert.equal(((runnerDefaults.imageRef as JsonRecord).dockerfilePath), "deploy/container/Containerfile");
assert.equal(runnerDefaults.namespace, "agentrun-v02");
const taskPayload = task.payload as JsonRecord;
assert.equal(taskPayload.model, "gpt-5.6-terra");
assert.equal(taskPayload.reasoningEffort, "xhigh");
assert.equal(taskPayload.model, "gpt-5.6-sol");
assert.equal(taskPayload.reasoningEffort, "medium");
assert.equal((taskPayload.modelConfig as JsonRecord).modelSource, "aipod-spec-default");
assert.equal((taskPayload.modelConfig as JsonRecord).reasoningEffortSource, "aipod-spec-default");
const modelResolution = rendered.modelResolution as JsonRecord;
assert.equal(modelResolution.model, "gpt-5.6-terra");
assert.equal(modelResolution.reasoningEffort, "xhigh");
assert.equal(modelResolution.model, "gpt-5.6-sol");
assert.equal(modelResolution.reasoningEffort, "medium");
assert.equal(modelResolution.modelSource, "aipod-spec-default");
assert.equal(modelResolution.reasoningEffortSource, "aipod-spec-default");
assert.equal(modelResolution.backendProfile, "gpt-pika");
@@ -146,11 +146,11 @@ const selfTest: SelfTestCase = async (context) => {
assert.deepEqual((modelOverrideResolution.providerCredential as JsonRecord).secretRef, disclosedProviderCredential.secretRef);
assert.deepEqual(((modelOverrideTask.executionPolicy as JsonRecord).secretScope as JsonRecord).providerCredentials, providerCredentials);
const reasoningOverride = await client.post("/api/v1/aipod-specs/Artificer/render", { prompt: "override reasoning", reasoningEffort: "medium" }) as JsonRecord;
const reasoningOverride = await client.post("/api/v1/aipod-specs/Artificer/render", { prompt: "override reasoning", reasoningEffort: "high" }) as JsonRecord;
const reasoningOverridePayload = (reasoningOverride.queueTask as JsonRecord).payload as JsonRecord;
const reasoningOverrideResolution = reasoningOverride.modelResolution as JsonRecord;
assert.equal(reasoningOverridePayload.model, "gpt-5.6-terra");
assert.equal(reasoningOverridePayload.reasoningEffort, "medium");
assert.equal(reasoningOverridePayload.model, "gpt-5.6-sol");
assert.equal(reasoningOverridePayload.reasoningEffort, "high");
assert.equal(reasoningOverrideResolution.modelSource, "aipod-spec-default");
assert.equal(reasoningOverrideResolution.reasoningEffortSource, "render-input-override");
@@ -236,8 +236,8 @@ const selfTest: SelfTestCase = async (context) => {
const changedRender = await client.post("/api/v1/aipod-specs/Artificer/render", { prompt: "new yaml default" }) as JsonRecord;
assert.equal(((changedRender.queueTask as JsonRecord).payload as JsonRecord).model, "gpt-5.7-terra");
const storedTask = await client.get(`/api/v1/queue/tasks/${String(durableTask.id)}`) as JsonRecord;
assert.equal((storedTask.payload as JsonRecord).model, "gpt-5.6-terra");
assert.equal((storedTask.payload as JsonRecord).reasoningEffort, "xhigh");
assert.equal((storedTask.payload as JsonRecord).model, "gpt-5.6-sol");
assert.equal((storedTask.payload as JsonRecord).reasoningEffort, "medium");
assert.deepEqual((storedTask.payload as JsonRecord).modelConfig, taskPayload.modelConfig);
return { name: "aipod-spec", tests: ["aipod-spec-yaml-parser-runtime-compatible", "aipod-spec-artificer-image-ref-render", "aipod-spec-artificer-v02-runtime-authority", "aipod-spec-gpt-pika-secret-namespace", "aipod-spec-model-defaults", "aipod-spec-model-reasoning-overrides", "aipod-spec-model-override-keeps-provider", "aipod-spec-control-override-fail-closed", "aipod-spec-payload-control-bypass-fail-closed", "aipod-spec-missing-gpt-pika-credential", "aipod-spec-durable-task-model-snapshot", "aipod-spec-primary-workspace-contract", "aipod-spec-artificer-github-url-render", "aipod-spec-artificer-github-ssh-required-keys", "aipod-spec-git-mirror-url", "git-fetch-stderr-classification", "resource-bundle-source-authority-validation", "resource-bundle-target-conflict-validation", "queue-submit-aipod-dry-run", "session-send-aipod-model-dry-run", "aipod-cli-help"] };
} finally {
src/selftest/cases/80-artificer-model-selection.ts
+14 -9
View File
@@ -33,8 +33,8 @@ const selfTest: SelfTestCase = async (context) => {
const modelResolution = rendered.modelResolution as JsonRecord;
assert.equal(task.backendProfile, "gpt-pika");
assert.equal(task.providerId, "NC01");
assert.equal(modelResolution.model, "gpt-5.6-terra");
assert.equal(modelResolution.reasoningEffort, "xhigh");
assert.equal(modelResolution.model, "gpt-5.6-sol");
assert.equal(modelResolution.reasoningEffort, "medium");
await assertDurableRetryModel(context.tmp, task as unknown as CreateQueueTaskInput);
const runtimeExecutionPolicy = structuredClone(task.executionPolicy as JsonRecord);
((runtimeExecutionPolicy.secretScope as JsonRecord).toolCredentials) = [];
@@ -63,7 +63,12 @@ const selfTest: SelfTestCase = async (context) => {
codexCommand: context.fakeCodexCommand,
codexArgs: context.fakeCodexArgs,
codexHome: context.codexHome,
env: { CODEX_HOME: context.codexHome, AGENTRUN_FAKE_CODEX_MODE: "require-explicit-model-effort" },
env: {
CODEX_HOME: context.codexHome,
AGENTRUN_FAKE_CODEX_MODE: "require-explicit-model-effort",
AGENTRUN_FAKE_CODEX_EXPECTED_MODEL: "gpt-5.6-sol",
AGENTRUN_FAKE_CODEX_EXPECTED_REASONING_EFFORT: "medium",
},
oneShot: true,
});
assert.equal(result.terminalStatus, "completed");
@@ -71,12 +76,12 @@ const selfTest: SelfTestCase = async (context) => {
const starting = events.items?.find((event) => event.type === "backend_status" && event.payload.phase === "codex-app-server-starting");
const selection = starting?.payload.modelSelection as JsonRecord | undefined;
assert.equal(starting?.payload.backendProfile, "gpt-pika");
assert.equal(selection?.model, "gpt-5.6-terra");
assert.equal(selection?.reasoningEffort, "xhigh");
assert.equal(selection?.model, "gpt-5.6-sol");
assert.equal(selection?.reasoningEffort, "medium");
assert.equal(selection?.reasoningProtocolField, "turn/start.effort");
const turnStart = events.items?.find((event) => event.type === "backend_status" && event.payload.phase === "turn/start:completed");
assert.equal(turnStart?.payload.model, "gpt-5.6-terra");
assert.equal(turnStart?.payload.reasoningEffort, "xhigh");
assert.equal(turnStart?.payload.model, "gpt-5.6-sol");
assert.equal(turnStart?.payload.reasoningEffort, "medium");
assert.equal(turnStart?.payload.reasoningProtocolField, "effort");
assertNoSecretLeak({ rendered, result, events });
return { name: "artificer-model-selection", tests: ["artificer-gpt-pika-default-render", "artificer-durable-retry-model", "artificer-gpt-pika-model-runtime", "artificer-reasoning-effort-turn-start", "artificer-model-selection-visible"] };
@@ -134,8 +139,8 @@ console.log(JSON.stringify({ apiVersion: manifest.apiVersion, kind: manifest.kin
});
const command = retried.command as JsonRecord;
assert.deepEqual(command.payload, task.payload);
assert.equal((command.payload as JsonRecord).model, "gpt-5.6-terra");
assert.equal((command.payload as JsonRecord).reasoningEffort, "xhigh");
assert.equal((command.payload as JsonRecord).model, "gpt-5.6-sol");
assert.equal((command.payload as JsonRecord).reasoningEffort, "medium");
assert.equal(((command.payload as JsonRecord).modelConfig as JsonRecord).modelSource, "aipod-spec-default");
}
src/selftest/fake-codex-app-server.ts
+5 -3
View File
@@ -4,6 +4,8 @@ import { appendFileSync, readFileSync, writeFileSync } from "node:fs";
const rl = readline.createInterface({ input: process.stdin, crlfDelay: Infinity });
const mode = process.env.AGENTRUN_FAKE_CODEX_MODE ?? "success";
const expectedModelEffortModel = process.env.AGENTRUN_FAKE_CODEX_EXPECTED_MODEL ?? "gpt-5.6-terra";
const expectedModelEffort = process.env.AGENTRUN_FAKE_CODEX_EXPECTED_REASONING_EFFORT ?? "xhigh";
if (process.env.AGENTRUN_FAKE_CODEX_START_FILE) appendFileSync(process.env.AGENTRUN_FAKE_CODEX_START_FILE, `${process.pid}\n`);
const fakeProcessAttempt = process.env.AGENTRUN_FAKE_CODEX_START_FILE ? lineCount(process.env.AGENTRUN_FAKE_CODEX_START_FILE) : 1;
let threadCounter = 0;
@@ -37,7 +39,7 @@ for await (const line of rl) {
respond(message.id, null, { code: -32000, message: "thread/start did not include expected model" });
continue;
}
if (mode === "require-explicit-model-effort" && (message.params?.model !== "gpt-5.6-terra" || Object.hasOwn(message.params ?? {}, "effort"))) {
if (mode === "require-explicit-model-effort" && (message.params?.model !== expectedModelEffortModel || Object.hasOwn(message.params ?? {}, "effort"))) {
respond(message.id, null, { code: -32000, message: "thread/start expected model without turn effort" });
continue;
}
@@ -65,7 +67,7 @@ for await (const line of rl) {
respond(message.id, null, { code: -32000, message: "thread/resume did not include expected model" });
continue;
}
if (mode === "require-explicit-model-effort" && (message.params?.model !== "gpt-5.6-terra" || Object.hasOwn(message.params ?? {}, "effort"))) {
if (mode === "require-explicit-model-effort" && (message.params?.model !== expectedModelEffortModel || Object.hasOwn(message.params ?? {}, "effort"))) {
respond(message.id, null, { code: -32000, message: "thread/resume expected model without turn effort" });
continue;
}
@@ -84,7 +86,7 @@ for await (const line of rl) {
respond(message.id, null, { code: -32000, message: "turn/start did not include expected model" });
continue;
}
if (mode === "require-explicit-model-effort" && (message.params?.model !== "gpt-5.6-terra" || message.params?.effort !== "xhigh")) {
if (mode === "require-explicit-model-effort" && (message.params?.model !== expectedModelEffortModel || message.params?.effort !== expectedModelEffort)) {
respond(message.id, null, { code: -32000, message: "turn/start did not include expected model and effort" });
continue;
}