getcompanion-ai/co-mono
1
0
Fork 0
mirror of https://github.com/getcompanion-ai/co-mono.git synced 2026-04-15 12:03:49 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

fix(plan-mode): track step completion via tool_result events

Browse source 
- No longer relies on agent outputting [STEP N DONE] tags
- Each successful tool_result marks the next uncompleted step done
- Much more reliable than expecting LLM to follow tag format
- Simplified execution context (no special instructions needed)
This commit is contained in:
Helmut Januschka 2026-01-03 22:20:07 +01:00 • committed by Mario Zechner
parent 361292d4de
commit fdcc044491
1 changed files with 19 additions and 40 deletions
Download patch file Download diff file Expand all files Collapse all files

59
packages/coding-agent/examples/hooks/plan-mode.ts
View file

@ -168,21 +168,7 @@ function extractTodoItems(message: string): TodoItem[] {
return items;
}
/**
* Find [STEP N DONE] or [DONE N] tags in text and return step numbers.
*/
function findDoneSteps(text: string): number[] {
const steps: number[] = [];
// Match [STEP 1 DONE], [STEP 2 DONE], etc.
for (const match of text.matchAll(/\[STEP\s+(\d+)\s+DONE\]/gi)) {
steps.push(parseInt(match[1], 10));
}
// Also match [DONE 1], [DONE 2], etc.
for (const match of text.matchAll(/\[DONE\s+(\d+)\]/gi)) {
steps.push(parseInt(match[1], 10));
}
return steps;
}
export default function planModeHook(pi: HookAPI) {
let planModeEnabled = false;
@ -290,6 +276,19 @@ export default function planModeHook(pi: HookAPI) {
}
});
// Track step completion based on tool results
pi.on("tool_result", async (_event, ctx) => {
if (!executionMode || todoItems.length === 0) return;
// Mark the first uncompleted step as done when any tool succeeds
const nextStep = todoItems.find((t) => !t.completed);
if (nextStep) {
nextStep.completed = true;
console.error(`[plan-mode] Marked step ${nextStep.step} complete: ${nextStep.text}`);
updateStatus(ctx);
}
});
// Filter out stale plan mode context messages from LLM context
// This ensures the agent only sees the CURRENT state (plan mode on/off)
pi.on("context", async (event) => {
@ -355,16 +354,16 @@ Do NOT attempt to make changes - just describe what you would do.`,
if (executionMode && todoItems.length > 0) {
console.error("[plan-mode] before_agent_start: injecting EXECUTING PLAN context");
const remaining = todoItems.filter((t) => !t.completed);
const todoList = remaining.map((t) => `Step ${t.step}: ${t.text}`).join("\n");
const todoList = remaining.map((t) => `${t.step}. ${t.text}`).join("\n");
return {
message: {
customType: "plan-execution-context",
content: `[EXECUTING PLAN - You have FULL tool access]
content: `[EXECUTING PLAN - Full tool access enabled]
Remaining steps:
${todoList}
IMPORTANT: After completing each step, output [STEP N DONE] where N is the step number.
Example: After completing step ${remaining[0]?.step || 1}, write [STEP ${remaining[0]?.step || 1} DONE]`,
Execute each step in order.`,
display: false,
},
};
@ -374,28 +373,8 @@ Example: After completing step ${remaining[0]?.step || 1}, write [STEP ${remaini
// After agent finishes
pi.on("agent_end", async (event, ctx) => {
// Check for done tags in execution mode
// In execution mode, check if all steps complete
if (executionMode && todoItems.length > 0) {
const messages = event.messages;
const lastAssistant = [...messages].reverse().find((m) => m.role === "assistant");
if (lastAssistant && Array.isArray(lastAssistant.content)) {
const textContent = lastAssistant.content
.filter((block): block is { type: "text"; text: string } => block.type === "text")
.map((block) => block.text)
.join("\n");
// Find and mark completed items
const doneSteps = findDoneSteps(textContent);
for (const stepNum of doneSteps) {
const item = todoItems.find((t) => t.step === stepNum);
if (item && !item.completed) {
item.completed = true;
}
}
updateStatus(ctx);
}
// Check if all complete
const allComplete = todoItems.every((t) => t.completed);
if (allComplete) {
// Show final completed list in chat