ton/pi_harness
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

add usage info to branch summary, compaction and tool result entries (#6671)

Browse Source 
* add usage info to branch summary entries

* add usage to compaction entries

* allow custom tools to report llm usage in tool results

* allow observing and patching usage in tool_result hooks

* agent-harness: save usage in entries

for compaction, branch summaries and tool results
This commit is contained in:
David Brailovsky
2026-07-20 16:41:43 +02:00
committed by GitHub
parent c179395218
commit 2fd3868401
29 changed files with 844 additions and 122 deletions
Download Patch File Download Diff File Expand all files Collapse all files
packages/coding-agent/test/suite/agent-session-compaction.test.ts
+37
View File
@@ -97,6 +97,14 @@ describe("AgentSession compaction characterization", () => {
});
it("manually compacts using an extension-provided summary", async () => {
const summaryUsage = {
input: 10,
output: 20,
cacheRead: 30,
cacheWrite: 40,
totalTokens: 100,
cost: { input: 0.1, output: 0.2, cacheRead: 0.3, cacheWrite: 0.4, total: 1 },
};
const harness = await createHarness({
settings: { compaction: { keepRecentTokens: 1 } },
extensionFactories: [
@@ -106,6 +114,7 @@ describe("AgentSession compaction characterization", () => {
summary: "summary from extension",
firstKeptEntryId: event.preparation.firstKeptEntryId,
tokensBefore: event.preparation.tokensBefore,
usage: summaryUsage,
details: { source: "extension" },
},
}));
@@ -116,14 +125,26 @@ describe("AgentSession compaction characterization", () => {
await harness.session.prompt("one");
await harness.session.prompt("two");
const statsBefore = harness.session.getSessionStats();
const result = await harness.session.compact();
const compactionEntries = harness.sessionManager.getEntries().filter((entry) => entry.type === "compaction");
const estimatedTokensAfter = harness.session.messages.reduce((sum, message) => sum + estimateTokens(message), 0);
expect(result.summary).toBe("summary from extension");
expect(result.usage).toEqual(summaryUsage);
expect(result.estimatedTokensAfter).toBe(estimatedTokensAfter);
expect(compactionEntries).toHaveLength(1);
const compactionEntry = compactionEntries[0];
if (compactionEntry?.type === "compaction") {
expect(compactionEntry.usage).toEqual(summaryUsage);
}
const statsAfter = harness.session.getSessionStats();
expect(statsAfter.tokens.input).toBe(statsBefore.tokens.input + summaryUsage.input);
expect(statsAfter.tokens.output).toBe(statsBefore.tokens.output + summaryUsage.output);
expect(statsAfter.tokens.cacheRead).toBe(statsBefore.tokens.cacheRead + summaryUsage.cacheRead);
expect(statsAfter.tokens.cacheWrite).toBe(statsBefore.tokens.cacheWrite + summaryUsage.cacheWrite);
expect(statsAfter.cost).toBe(statsBefore.cost + summaryUsage.cost.total);
expect(harness.session.messages[0]?.role).toBe("compactionSummary");
});
@@ -154,6 +175,22 @@ describe("AgentSession compaction characterization", () => {
expect(getStreamCallCount()).toBe(1);
});
it("persists usage from pi-generated manual compaction", async () => {
const harness = await createHarness({ withConfiguredAuth: false });
harnesses.push(harness);
seedCompactableSession(harness);
useSummaryStreamFn(harness, "summary from custom stream");
const result = await harness.session.compact();
const compactionEntries = harness.sessionManager.getEntries().filter((entry) => entry.type === "compaction");
expect(result.usage).toEqual(createUsage(10));
expect(compactionEntries).toHaveLength(1);
expect(compactionEntries[0]?.type === "compaction" ? compactionEntries[0].usage : undefined).toEqual(
createUsage(10),
);
});
it("auto-compacts with a custom streamFn when registry auth is absent", async () => {
const harness = await createHarness({ withConfiguredAuth: false });
harnesses.push(harness);