Files
paperclip/server/src/__tests__/gemini-local-adapter.test.ts
T
Sherif Kozman b8fb81dee9 fix(gemini-local): treat token-overflow as a fresh-session signal (#4932)
## Thinking Path

The same 2026-04-30 audit that produced PR #4118 (`Invalid session`
regex extension) and the ENOTFOUND classifier (#4931) identified a third
stuck-session pattern: **13 failures in 7 days, all on a single agent
(Ernest)**, with stderr matching:

```
_ApiError: {"error":{"code":400,"message":"The input token count exceeds
the maximum number of tokens allowed 1048576","status":"INVALID_ARGUMENT"}}
  at ChatCompressionService.compress
```

The root cause is that gemini-cli's `ChatCompressionService` blew the 1M
token context limit **during its compression step itself**. Resuming the
same session ID will hit the same wall on the next attempt — the session
is effectively dead the same way it is when "Invalid session identifier"
fires (PR #4118).

## What Changed

Extends the `isGeminiUnknownSessionError` regex in `parse.ts` with two
phrases:
- `exceeds\s+the\s+maximum\s+number\s+of\s+tokens`
- `input\s+token\s+count\s+exceeds`

Both trigger the **existing** fresh-session retry path in
`execute.ts:596` — no new code path. Same extension pattern as PR #4118.

## Verification

- `npx vitest run --project @paperclipai/adapter-gemini-local` → 14/14
pass (11 in `parse.test.ts` + 3 existing in `execute.remote.test.ts`)
- 2 new tests cover the token-overflow patterns
- `pnpm --filter @paperclipai/adapter-gemini-local typecheck` → clean
- Audit query against `heartbeat_runs.stderr_excerpt` confirms regex
matches all 13 occurrences

## Stacking

This PR is stacked on top of #4931 (the ENOTFOUND classifier) which adds
the `parse.test.ts` file. If #4931 merges first, this PR's diff is just
the regex + 2 tests. If this PR is reviewed first, please merge #4931
first to avoid touching the same test scaffolding twice.

## Risks

- **Low.** Single-line regex extension. No new code paths.
- The session-reset path is well-trodden (PR #4118 in flight).
- If a non-Gemini caller produces a stderr containing "exceeds the
maximum number of tokens" by coincidence, they would trigger one
unnecessary fresh-session retry. Not plausible in the gemini-cli output
context where this stderr is sourced.

## Model Used

Claude Opus 4.7 (1M context), Anthropic SDK via Claude Code CLI.

## Checklist

- [x] Thinking path traces from audit data to single-line regex change
- [x] Model specified
- [x] No duplicate of planned core work
- [x] Tests pass locally
- [x] Tests added (2 new)
- [x] N/A — server-side regex
- [x] Internal pattern; no docs change
- [x] Risks documented
- [x] Will address Greptile + reviewer comments before merge
- [x] I searched the GitHub PR list for similar PRs and confirmed this
is not a duplicate (related: #4118 covers the "Invalid session
identifier" regex; this PR extends the same regex with token-overflow
phrases)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Devin Foley <devin@paperclip.ing>
Co-authored-by: Paperclip <noreply@paperclip.ing>
2026-06-10 19:10:29 -07:00

204 lines
5.9 KiB
TypeScript

import { describe, expect, it, vi } from "vitest";
import {
isGeminiTurnLimitResult,
isGeminiSessionUnrecoverableError,
parseGeminiJsonl,
} from "@paperclipai/adapter-gemini-local/server";
import { parseGeminiStdoutLine } from "@paperclipai/adapter-gemini-local/ui";
import { printGeminiStreamEvent } from "@paperclipai/adapter-gemini-local/cli";
describe("gemini_local parser", () => {
it("extracts session, summary, usage, cost, and terminal error message from v0.38 stream-json output", () => {
const stdout = [
JSON.stringify({ type: "system", subtype: "init", session_id: "gemini-session-1", model: "gemini-2.5-pro" }),
JSON.stringify({
type: "message",
role: "assistant",
content: "hello",
}),
JSON.stringify({
type: "result",
status: "success",
session_id: "gemini-session-1",
stats: {
input_tokens: 12,
cached_input_tokens: 3,
output_tokens: 7,
},
total_cost_usd: 0.00123,
}),
JSON.stringify({ type: "error", message: "model access denied" }),
].join("\n");
const parsed = parseGeminiJsonl(stdout);
expect(parsed.sessionId).toBe("gemini-session-1");
expect(parsed.summary).toBe("hello");
expect(parsed.usage).toEqual({
inputTokens: 12,
cachedInputTokens: 3,
outputTokens: 7,
});
expect(parsed.costUsd).toBeCloseTo(0.00123, 6);
expect(parsed.errorMessage).toBe("model access denied");
});
it("extracts structured questions", () => {
const stdout = [
JSON.stringify({
type: "assistant",
message: {
content: [
{ type: "output_text", text: "I have a question." },
{
type: "question",
prompt: "Which model?",
choices: [
{ key: "pro", label: "Gemini Pro", description: "Better" },
{ key: "flash", label: "Gemini Flash" },
],
},
],
},
}),
].join("\n");
const parsed = parseGeminiJsonl(stdout);
expect(parsed.summary).toBe("I have a question.");
expect(parsed.question).toEqual({
prompt: "Which model?",
choices: [
{ key: "pro", label: "Gemini Pro", description: "Better" },
{ key: "flash", label: "Gemini Flash", description: undefined },
],
});
});
});
describe("gemini_local stale session detection", () => {
it("treats missing session messages as an unknown session error", () => {
expect(isGeminiSessionUnrecoverableError("", "unknown session id abc")).toBe(true);
expect(isGeminiSessionUnrecoverableError("", "checkpoint latest not found")).toBe(true);
});
});
describe("gemini_local turn-limit detection", () => {
it("detects structured turn-limit signals and exit code 53", () => {
expect(isGeminiTurnLimitResult({ status: "turn_limit" })).toBe(true);
expect(isGeminiTurnLimitResult({ stopReason: "max_turns_exhausted" })).toBe(true);
expect(isGeminiTurnLimitResult(null, 53)).toBe(true);
});
it("checks every structured stop field for turn-limit exhaustion", () => {
expect(
isGeminiTurnLimitResult({
status: "success",
stopReason: "turn_limit_exhausted",
}),
).toBe(true);
});
it("does not detect turn-limit exhaustion from unstructured error text", () => {
expect(isGeminiTurnLimitResult({ error: "max_turns reached" })).toBe(false);
});
});
describe("gemini_local ui stdout parser", () => {
it("parses v0.38 assistant message and result events", () => {
const ts = "2026-03-08T00:00:00.000Z";
expect(
parseGeminiStdoutLine(
JSON.stringify({
type: "message",
role: "assistant",
content: "I checked the repo.",
}),
ts,
),
).toEqual([
{ kind: "assistant", ts, text: "I checked the repo." },
]);
expect(
parseGeminiStdoutLine(
JSON.stringify({
type: "result",
status: "success",
text: "Done",
stats: {
input_tokens: 10,
output_tokens: 5,
cached_input_tokens: 2,
},
total_cost_usd: 0.00042,
}),
ts,
),
).toEqual([
{
kind: "result",
ts,
text: "Done",
inputTokens: 10,
outputTokens: 5,
cachedTokens: 2,
costUsd: 0.00042,
subtype: "success",
isError: false,
errors: [],
},
]);
});
});
function stripAnsi(value: string): string {
return value.replace(/\x1b\[[0-9;]*m/g, "");
}
describe("gemini_local cli formatter", () => {
it("prints init, v0.38 assistant, result, and error events", () => {
const spy = vi.spyOn(console, "log").mockImplementation(() => {});
let joined = "";
try {
printGeminiStreamEvent(
JSON.stringify({ type: "system", subtype: "init", session_id: "gemini-session-1", model: "gemini-2.5-pro" }),
false,
);
printGeminiStreamEvent(
JSON.stringify({
type: "message",
role: "assistant",
content: "hello",
}),
false,
);
printGeminiStreamEvent(
JSON.stringify({
type: "result",
status: "success",
stats: {
input_tokens: 10,
output_tokens: 5,
cached_input_tokens: 2,
},
total_cost_usd: 0.00042,
}),
false,
);
printGeminiStreamEvent(
JSON.stringify({ type: "error", message: "boom" }),
false,
);
joined = spy.mock.calls.map((call) => stripAnsi(call.join(" "))).join("\n");
} finally {
spy.mockRestore();
}
expect(joined).toContain("Gemini init");
expect(joined).toContain("assistant: hello");
expect(joined).toContain("tokens: in=10 out=5 cached=2 cost=$0.000420");
expect(joined).toContain("error: boom");
});
});