test(benchmark): full post-CLI gemini e2e + GEMINI_API_KEY auth path

Exercise resultFromGeminiStream against current stream-json fixtures
(including empty-success hardening). Recognize GEMINI_API_KEY and map
IneligibleTierError to auth — personal OAuth free-tier is no longer
supported by gemini CLI.

Co-authored-by: Cursor <cursoragent@cursor.com>
This commit is contained in:
Sina
2026-07-10 09:38:06 -07:00
parent f155481b78
commit 5d93ce278d
2 changed files with 92 additions and 23 deletions
+45 -2
View File
@@ -1,7 +1,8 @@
import { describe, test, expect } from 'bun:test';
import { parseGeminiStreamJson } from './gemini';
import { parseGeminiStreamJson, resultFromGeminiStream } from './gemini';
// Current CLI shape (gemini ≥ ~1.58): content + role, tokens under result.stats
// Current CLI shape (gemini ≥ ~0.11 / stream-json): content + role, tokens under result.stats
// Documented at https://geminicli.com/docs/cli/headless/ and gemini-cli PR #10883.
const CURRENT_CLI_FIXTURE = [
'{"type":"init","timestamp":"2026-03-20T15:14:46.455Z","session_id":"test-session-123","model":"gemini-2.5-pro"}',
'{"type":"message","role":"user","content":"Reply with exactly the word: PONG"}',
@@ -93,3 +94,45 @@ describe('parseGeminiStreamJson', () => {
expect(parseGeminiStreamJson(raw).modelUsed).toBe('from-result');
});
});
describe('resultFromGeminiStream (adapter post-CLI e2e)', () => {
test('current CLI fixture → success row with PONG + tokens (not $0)', () => {
const result = resultFromGeminiStream(CURRENT_CLI_FIXTURE, { durationMs: 42 });
expect(result.error).toBeUndefined();
expect(result.output).toBe('PONG');
expect(result.tokens.input).toBe(24946);
expect(result.tokens.output).toBe(32);
expect(result.modelUsed).toBe('gemini-2.5-pro');
expect(result.durationMs).toBe(42);
// Cost signal: non-zero tokens means estimateCost will not be $0.
expect(result.tokens.input + result.tokens.output).toBeGreaterThan(0);
});
test('legacy text-only success still works', () => {
const result = resultFromGeminiStream(LEGACY_FIXTURE, { durationMs: 10 });
expect(result.error).toBeUndefined();
expect(result.output).toBe('hello');
expect(result.toolCalls).toBe(1);
});
test('empty exit-0 stream → error row, never silent $0 success (#2159)', () => {
const emptySuccess = [
'{"type":"init","model":"gemini-2.5-pro"}',
'{"type":"message","role":"user","content":"hi"}',
'{"type":"result","status":"success","stats":{"input_tokens":0,"output_tokens":0}}',
].join('\n');
const result = resultFromGeminiStream(emptySuccess, { durationMs: 5 });
expect(result.error).toBeDefined();
expect(result.error!.code).toBe('unknown');
expect(result.error!.reason).toContain('empty output');
expect(result.output).toBe('');
expect(result.modelUsed).toBe('gemini-2.5-pro');
});
test('pre-fix bug shape (text field missing, only content) would have been empty — now succeeds', () => {
// This is the exact failure mode from #2159: content present, no text.
const result = resultFromGeminiStream(CURRENT_CLI_FIXTURE);
expect(result.error).toBeUndefined();
expect(result.output).toBe('PONG');
});
});
+47 -21
View File
@@ -62,6 +62,37 @@ export function parseGeminiStreamJson(raw: string): GeminiStreamParse {
return { output, tokens: { input, output: out }, toolCalls, modelUsed };
}
/**
* Map a raw stream-json dump to a RunResult, including the empty-success
* hardening from #2159. Exported so adapter e2e can exercise the full
* post-CLI path without a live gemini binary.
*/
export function resultFromGeminiStream(
raw: string,
opts: { model?: string; durationMs?: number } = {},
): RunResult {
const parsed = parseGeminiStreamJson(raw);
const modelUsed = parsed.modelUsed || opts.model || 'gemini-2.5-pro';
const durationMs = opts.durationMs ?? 0;
if (!parsed.output.trim()) {
return {
output: '',
tokens: { input: 0, output: 0 },
durationMs,
toolCalls: 0,
modelUsed,
error: { code: 'unknown', reason: 'empty output from gemini CLI (exit 0)' },
};
}
return {
output: parsed.output,
tokens: parsed.tokens,
durationMs,
toolCalls: parsed.toolCalls,
modelUsed,
};
}
/**
* Gemini adapter — wraps the `gemini` CLI.
*
@@ -83,9 +114,14 @@ export class GeminiAdapter implements ProviderAdapter {
const newCfgDir = path.join(os.homedir(), '.gemini');
const newOauth = path.join(newCfgDir, 'oauth_creds.json');
const hasCfg = fs.existsSync(legacyCfgDir) || fs.existsSync(newOauth);
const hasKey = !!process.env.GOOGLE_API_KEY;
// CLI accepts either name; Google AI Studio keys are usually GEMINI_API_KEY.
const hasKey = !!(process.env.GOOGLE_API_KEY || process.env.GEMINI_API_KEY);
if (!hasCfg && !hasKey) {
return { ok: false, reason: 'No Gemini auth found. Log in via `gemini login` or export GOOGLE_API_KEY.' };
return {
ok: false,
reason:
'No Gemini auth found. Export GEMINI_API_KEY (or GOOGLE_API_KEY) from https://aistudio.google.com/app/apikey — personal OAuth free-tier is no longer supported by gemini CLI.',
};
}
return { ok: true };
}
@@ -104,25 +140,15 @@ export class GeminiAdapter implements ProviderAdapter {
timeout: opts.timeoutMs,
encoding: 'utf-8',
maxBuffer: 32 * 1024 * 1024,
env: {
...process.env,
// Prefer GEMINI_API_KEY when only that is set (CLI reads both).
...(process.env.GEMINI_API_KEY && !process.env.GOOGLE_API_KEY
? { GOOGLE_API_KEY: process.env.GEMINI_API_KEY }
: {}),
},
});
const parsed = parseGeminiStreamJson(out);
const modelUsed = parsed.modelUsed || opts.model || 'gemini-2.5-pro';
// Empty-success is indistinguishable from a healthy cheap run in the
// comparison table — never report it as a $0 success row (#2159).
if (!parsed.output.trim()) {
return this.emptyResult(
Date.now() - start,
{ code: 'unknown', reason: 'empty output from gemini CLI (exit 0)' },
modelUsed,
);
}
return {
output: parsed.output,
tokens: parsed.tokens,
durationMs: Date.now() - start,
toolCalls: parsed.toolCalls,
modelUsed,
};
return resultFromGeminiStream(out, { model: opts.model, durationMs: Date.now() - start });
} catch (err: unknown) {
const durationMs = Date.now() - start;
const e = err as { code?: string; stderr?: Buffer; signal?: string; message?: string };
@@ -130,7 +156,7 @@ export class GeminiAdapter implements ProviderAdapter {
if (e.signal === 'SIGTERM' || e.code === 'ETIMEDOUT') {
return this.emptyResult(durationMs, { code: 'timeout', reason: `exceeded ${opts.timeoutMs}ms` }, opts.model);
}
if (/unauthorized|auth|login|api key/i.test(stderr)) {
if (/unauthorized|auth|login|api key|ineligibletier|no longer supported/i.test(stderr)) {
return this.emptyResult(durationMs, { code: 'auth', reason: stderr.slice(0, 400) }, opts.model);
}
if (/rate[- ]?limit|429|quota/i.test(stderr)) {