fix(agent-tars): probabilistic plan anomaly on gpt-4o (Closes #273) (#373)

ulivz · web-flow · commit 97f0f74af21f · 2025-03-26T15:54:08.000+08:00
diff --git a/apps/agent-tars/src/renderer/src/agent/Aware.ts b/apps/agent-tars/src/renderer/src/agent/Aware.ts
@@ -33,6 +33,7 @@ If the task is none or current step is done, you should increment the step numbe
 {
   "reflection": "[your reflection about current environment]",
   "step": "[next step number]",
+  "plan": "[steps array with id and title fields]",
   "status": "[next task description, a complete sentence tell user what to do next]",
 }
 \`\`\`

Original file line number	Diff line number	Diff line change
`@@ -33,6 +33,7 @@ If the task is none or current step is done, you should increment the step numbe`
`33`	`33`	`{`
`34`	`34`	`"reflection": "[your reflection about current environment]",`
`35`	`35`	`"step": "[next step number]",`
	`36`	`+ "plan": "[steps array with id and title fields]",`
`36`	`37`	`"status": "[next task description, a complete sentence tell user what to do next]",`
`37`	`38`	`}`
`38`	`39`	\`\`\`