snapeval 2.0.0 → 2.1.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +144 -104
- package/bin/snapeval.ts +39 -1
- package/dist/bin/snapeval.js +33 -0
- package/dist/bin/snapeval.js.map +1 -1
- package/dist/src/adapters/copilot-sdk-client.js +3 -1
- package/dist/src/adapters/copilot-sdk-client.js.map +1 -1
- package/dist/src/adapters/harness/copilot-sdk.d.ts +11 -0
- package/dist/src/adapters/harness/copilot-sdk.js +101 -0
- package/dist/src/adapters/harness/copilot-sdk.js.map +1 -0
- package/dist/src/adapters/harness/resolve.js +10 -2
- package/dist/src/adapters/harness/resolve.js.map +1 -1
- package/dist/src/adapters/inference/copilot-sdk.js +4 -1
- package/dist/src/adapters/inference/copilot-sdk.js.map +1 -1
- package/dist/src/adapters/report/terminal.js +89 -9
- package/dist/src/adapters/report/terminal.js.map +1 -1
- package/dist/src/commands/eval.d.ts +3 -0
- package/dist/src/commands/eval.js +146 -17
- package/dist/src/commands/eval.js.map +1 -1
- package/dist/src/commands/review.d.ts +1 -0
- package/dist/src/commands/review.js.map +1 -1
- package/dist/src/config.js +2 -1
- package/dist/src/config.js.map +1 -1
- package/dist/src/engine/grader.js +67 -9
- package/dist/src/engine/grader.js.map +1 -1
- package/dist/src/engine/runner.d.ts +1 -0
- package/dist/src/engine/runner.js +15 -12
- package/dist/src/engine/runner.js.map +1 -1
- package/dist/src/errors.d.ts +6 -0
- package/dist/src/errors.js +21 -3
- package/dist/src/errors.js.map +1 -1
- package/dist/src/types.d.ts +3 -0
- package/package.json +4 -1
- package/plugin.json +1 -1
- package/skills/snapeval/SKILL.md +132 -39
- package/src/adapters/copilot-sdk-client.ts +3 -1
- package/src/adapters/harness/copilot-sdk.ts +126 -0
- package/src/adapters/harness/resolve.ts +13 -2
- package/src/adapters/inference/copilot-sdk.ts +5 -1
- package/src/adapters/report/terminal.ts +99 -10
- package/src/commands/eval.ts +183 -31
- package/src/commands/review.ts +1 -1
- package/src/config.ts +2 -1
- package/src/engine/grader.ts +59 -8
- package/src/engine/runner.ts +16 -13
- package/src/errors.ts +24 -3
- package/src/types.ts +3 -0
|
@@ -0,0 +1,101 @@
|
|
|
1
|
+
import * as fs from 'node:fs';
|
|
2
|
+
import * as path from 'node:path';
|
|
3
|
+
import { getClient, isSDKInstalled } from '../copilot-sdk-client.js';
|
|
4
|
+
export class CopilotSDKHarness {
|
|
5
|
+
name = 'copilot-sdk';
|
|
6
|
+
async run(options) {
|
|
7
|
+
const startMs = Date.now();
|
|
8
|
+
const client = await getClient();
|
|
9
|
+
fs.mkdirSync(options.outputDir, { recursive: true });
|
|
10
|
+
// Dynamically import SDK for approveAll
|
|
11
|
+
// @ts-ignore — module may not be installed (optional dep)
|
|
12
|
+
const { approveAll } = await import('@github/copilot-sdk');
|
|
13
|
+
// Build session config
|
|
14
|
+
const sessionConfig = {
|
|
15
|
+
model: 'gpt-4.1',
|
|
16
|
+
onPermissionRequest: approveAll,
|
|
17
|
+
workingDirectory: options.outputDir,
|
|
18
|
+
infiniteSessions: { enabled: false },
|
|
19
|
+
};
|
|
20
|
+
// Native skill loading: point skillDirectories at the skill's parent
|
|
21
|
+
if (options.skillPath) {
|
|
22
|
+
sessionConfig.skillDirectories = [options.skillPath];
|
|
23
|
+
}
|
|
24
|
+
const session = await client.createSession(sessionConfig);
|
|
25
|
+
try {
|
|
26
|
+
// Attach input files if provided
|
|
27
|
+
const attachments = [];
|
|
28
|
+
if (options.files) {
|
|
29
|
+
for (const file of options.files) {
|
|
30
|
+
// Copy to outputDir for script assertions, and attach for the model
|
|
31
|
+
const dest = path.join(options.outputDir, path.basename(file));
|
|
32
|
+
fs.copyFileSync(file, dest);
|
|
33
|
+
attachments.push({ type: 'file', path: dest, displayName: path.basename(file) });
|
|
34
|
+
}
|
|
35
|
+
}
|
|
36
|
+
const response = await session.sendAndWait({
|
|
37
|
+
prompt: options.prompt,
|
|
38
|
+
...(attachments.length > 0 ? { attachments } : {}),
|
|
39
|
+
}, 300_000);
|
|
40
|
+
const raw = response?.data?.content ?? '';
|
|
41
|
+
// Collect full transcript from session events
|
|
42
|
+
const events = await session.getMessages();
|
|
43
|
+
const transcript = buildTranscript(events);
|
|
44
|
+
// Extract token count from events if available
|
|
45
|
+
const totalTokens = extractTokenCount(events);
|
|
46
|
+
const durationMs = Date.now() - startMs;
|
|
47
|
+
return {
|
|
48
|
+
raw: raw.trim(),
|
|
49
|
+
transcript,
|
|
50
|
+
files: [],
|
|
51
|
+
total_tokens: totalTokens,
|
|
52
|
+
duration_ms: durationMs,
|
|
53
|
+
};
|
|
54
|
+
}
|
|
55
|
+
finally {
|
|
56
|
+
await session.disconnect();
|
|
57
|
+
}
|
|
58
|
+
}
|
|
59
|
+
async isAvailable() {
|
|
60
|
+
return isSDKInstalled();
|
|
61
|
+
}
|
|
62
|
+
}
|
|
63
|
+
function buildTranscript(events) {
|
|
64
|
+
const lines = [];
|
|
65
|
+
for (const event of events) {
|
|
66
|
+
switch (event.type) {
|
|
67
|
+
case 'user.message':
|
|
68
|
+
lines.push(`[user] ${event.data?.content ?? ''}`);
|
|
69
|
+
break;
|
|
70
|
+
case 'assistant.message':
|
|
71
|
+
lines.push(`[assistant] ${event.data?.content ?? ''}`);
|
|
72
|
+
break;
|
|
73
|
+
case 'tool.execution_start':
|
|
74
|
+
lines.push(`[tool:start] ${event.data?.toolName ?? 'unknown'}(${JSON.stringify(event.data?.arguments ?? {})})`);
|
|
75
|
+
break;
|
|
76
|
+
case 'tool.execution_complete':
|
|
77
|
+
lines.push(`[tool:done] ${event.data?.toolName ?? 'unknown'} → ${truncate(event.data?.result ?? '', 200)}`);
|
|
78
|
+
break;
|
|
79
|
+
case 'skill.invoked':
|
|
80
|
+
lines.push(`[skill] ${event.data?.name ?? 'unknown'} (${event.data?.path ?? ''})`);
|
|
81
|
+
break;
|
|
82
|
+
case 'session.error':
|
|
83
|
+
lines.push(`[error] ${event.data?.message ?? ''}`);
|
|
84
|
+
break;
|
|
85
|
+
}
|
|
86
|
+
}
|
|
87
|
+
return lines.join('\n');
|
|
88
|
+
}
|
|
89
|
+
function extractTokenCount(events) {
|
|
90
|
+
let total = 0;
|
|
91
|
+
for (const event of events) {
|
|
92
|
+
if (event.type === 'assistant.usage') {
|
|
93
|
+
total += (event.data?.inputTokens ?? 0) + (event.data?.outputTokens ?? 0);
|
|
94
|
+
}
|
|
95
|
+
}
|
|
96
|
+
return total;
|
|
97
|
+
}
|
|
98
|
+
function truncate(str, max) {
|
|
99
|
+
return str.length > max ? str.slice(0, max) + '...' : str;
|
|
100
|
+
}
|
|
101
|
+
//# sourceMappingURL=copilot-sdk.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"copilot-sdk.js","sourceRoot":"","sources":["../../../../src/adapters/harness/copilot-sdk.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,MAAM,SAAS,CAAC;AAC9B,OAAO,KAAK,IAAI,MAAM,WAAW,CAAC;AAElC,OAAO,EAAE,SAAS,EAAE,cAAc,EAAE,MAAM,0BAA0B,CAAC;AAErE,MAAM,OAAO,iBAAiB;IACnB,IAAI,GAAG,aAAa,CAAC;IAE9B,KAAK,CAAC,GAAG,CAAC,OAKT;QACC,MAAM,OAAO,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;QAC3B,MAAM,MAAM,GAAG,MAAM,SAAS,EAAE,CAAC;QAEjC,EAAE,CAAC,SAAS,CAAC,OAAO,CAAC,SAAS,EAAE,EAAE,SAAS,EAAE,IAAI,EAAE,CAAC,CAAC;QAErD,wCAAwC;QACxC,0DAA0D;QAC1D,MAAM,EAAE,UAAU,EAAE,GAAG,MAAM,MAAM,CAAC,qBAAqB,CAAC,CAAC;QAE3D,uBAAuB;QACvB,MAAM,aAAa,GAA4B;YAC7C,KAAK,EAAE,SAAS;YAChB,mBAAmB,EAAE,UAAU;YAC/B,gBAAgB,EAAE,OAAO,CAAC,SAAS;YACnC,gBAAgB,EAAE,EAAE,OAAO,EAAE,KAAK,EAAE;SACrC,CAAC;QAEF,qEAAqE;QACrE,IAAI,OAAO,CAAC,SAAS,EAAE,CAAC;YACtB,aAAa,CAAC,gBAAgB,GAAG,CAAC,OAAO,CAAC,SAAS,CAAC,CAAC;QACvD,CAAC;QAED,MAAM,OAAO,GAAG,MAAM,MAAM,CAAC,aAAa,CAAC,aAAa,CAAC,CAAC;QAE1D,IAAI,CAAC;YACH,iCAAiC;YACjC,MAAM,WAAW,GAAgE,EAAE,CAAC;YACpF,IAAI,OAAO,CAAC,KAAK,EAAE,CAAC;gBAClB,KAAK,MAAM,IAAI,IAAI,OAAO,CAAC,KAAK,EAAE,CAAC;oBACjC,oEAAoE;oBACpE,MAAM,IAAI,GAAG,IAAI,CAAC,IAAI,CAAC,OAAO,CAAC,SAAS,EAAE,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAC,CAAC;oBAC/D,EAAE,CAAC,YAAY,CAAC,IAAI,EAAE,IAAI,CAAC,CAAC;oBAC5B,WAAW,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,IAAI,EAAE,WAAW,EAAE,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;gBACnF,CAAC;YACH,CAAC;YAED,MAAM,QAAQ,GAAG,MAAM,OAAO,CAAC,WAAW,CACxC;gBACE,MAAM,EAAE,OAAO,CAAC,MAAM;gBACtB,GAAG,CAAC,WAAW,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,EAAE,WAAW,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;aACnD,EACD,OAAO,CACR,CAAC;YAEF,MAAM,GAAG,GAAG,QAAQ,EAAE,IAAI,EAAE,OAAO,IAAI,EAAE,CAAC;YAE1C,8CAA8C;YAC9C,MAAM,MAAM,GAAG,MAAM,OAAO,CAAC,WAAW,EAAE,CAAC;YAC3C,MAAM,UAAU,GAAG,eAAe,CAAC,MAAM,CAAC,CAAC;YAE3C,+CAA+C;YAC/C,MAAM,WAAW,GAAG,iBAAiB,CAAC,MAAM,CAAC,CAAC;YAE9C,MAAM,UAAU,GAAG,IAAI,CAAC,GAAG,EAAE,GAAG,OAAO,CAAC;YAExC,OAAO;gBACL,GAAG,EAAE,GAAG,CAAC,IAAI,EAAE;gBACf,UAAU;gBACV,KAAK,EAAE,EAAE;gBACT,YAAY,EAAE,WAAW;gBACzB,WAAW,EAAE,UAAU;aACxB,CAAC;QACJ,CAAC;gBAAS,CAAC;YACT,MAAM,OAAO,CAAC,UAAU,EAAE,CAAC;QAC7B,CAAC;IACH,CAAC;IAED,KAAK,CAAC,WAAW;QACf,OAAO,cAAc,EAAE,CAAC;IAC1B,CAAC;CACF;AAED,SAAS,eAAe,CAAC,MAAa;IACpC,MAAM,KAAK,GAAa,EAAE,CAAC;IAC3B,KAAK,MAAM,KAAK,IAAI,MAAM,EAAE,CAAC;QAC3B,QAAQ,KAAK,CAAC,IAAI,EAAE,CAAC;YACnB,KAAK,cAAc;gBACjB,KAAK,CAAC,IAAI,CAAC,UAAU,KAAK,CAAC,IAAI,EAAE,OAAO,IAAI,EAAE,EAAE,CAAC,CAAC;gBAClD,MAAM;YACR,KAAK,mBAAmB;gBACtB,KAAK,CAAC,IAAI,CAAC,eAAe,KAAK,CAAC,IAAI,EAAE,OAAO,IAAI,EAAE,EAAE,CAAC,CAAC;gBACvD,MAAM;YACR,KAAK,sBAAsB;gBACzB,KAAK,CAAC,IAAI,CAAC,gBAAgB,KAAK,CAAC,IAAI,EAAE,QAAQ,IAAI,SAAS,IAAI,IAAI,CAAC,SAAS,CAAC,KAAK,CAAC,IAAI,EAAE,SAAS,IAAI,EAAE,CAAC,GAAG,CAAC,CAAC;gBAChH,MAAM;YACR,KAAK,yBAAyB;gBAC5B,KAAK,CAAC,IAAI,CAAC,eAAe,KAAK,CAAC,IAAI,EAAE,QAAQ,IAAI,SAAS,MAAM,QAAQ,CAAC,KAAK,CAAC,IAAI,EAAE,MAAM,IAAI,EAAE,EAAE,GAAG,CAAC,EAAE,CAAC,CAAC;gBAC5G,MAAM;YACR,KAAK,eAAe;gBAClB,KAAK,CAAC,IAAI,CAAC,WAAW,KAAK,CAAC,IAAI,EAAE,IAAI,IAAI,SAAS,KAAK,KAAK,CAAC,IAAI,EAAE,IAAI,IAAI,EAAE,GAAG,CAAC,CAAC;gBACnF,MAAM;YACR,KAAK,eAAe;gBAClB,KAAK,CAAC,IAAI,CAAC,WAAW,KAAK,CAAC,IAAI,EAAE,OAAO,IAAI,EAAE,EAAE,CAAC,CAAC;gBACnD,MAAM;QACV,CAAC;IACH,CAAC;IACD,OAAO,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;AAC1B,CAAC;AAED,SAAS,iBAAiB,CAAC,MAAa;IACtC,IAAI,KAAK,GAAG,CAAC,CAAC;IACd,KAAK,MAAM,KAAK,IAAI,MAAM,EAAE,CAAC;QAC3B,IAAI,KAAK,CAAC,IAAI,KAAK,iBAAiB,EAAE,CAAC;YACrC,KAAK,IAAI,CAAC,KAAK,CAAC,IAAI,EAAE,WAAW,IAAI,CAAC,CAAC,GAAG,CAAC,KAAK,CAAC,IAAI,EAAE,YAAY,IAAI,CAAC,CAAC,CAAC;QAC5E,CAAC;IACH,CAAC;IACD,OAAO,KAAK,CAAC;AACf,CAAC;AAED,SAAS,QAAQ,CAAC,GAAW,EAAE,GAAW;IACxC,OAAO,GAAG,CAAC,MAAM,GAAG,GAAG,CAAC,CAAC,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,GAAG,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC;AAC5D,CAAC"}
|
|
@@ -1,9 +1,17 @@
|
|
|
1
1
|
import { CopilotCLIHarness } from './copilot-cli.js';
|
|
2
|
-
import {
|
|
2
|
+
import { CopilotSDKHarness } from './copilot-sdk.js';
|
|
3
|
+
import { AdapterNotAvailableError, SnapevalError } from '../../errors.js';
|
|
4
|
+
import { isSDKInstalled } from '../copilot-sdk-client.js';
|
|
3
5
|
export function resolveHarness(name) {
|
|
6
|
+
if (name === 'copilot-sdk') {
|
|
7
|
+
if (!isSDKInstalled()) {
|
|
8
|
+
throw new AdapterNotAvailableError('copilot-sdk', '@github/copilot-sdk is not installed. Install with: npm install @github/copilot-sdk');
|
|
9
|
+
}
|
|
10
|
+
return new CopilotSDKHarness();
|
|
11
|
+
}
|
|
4
12
|
if (name === 'copilot-cli') {
|
|
5
13
|
return new CopilotCLIHarness();
|
|
6
14
|
}
|
|
7
|
-
throw new SnapevalError(`Unknown harness "${name}". Built-in options: copilot-cli.`);
|
|
15
|
+
throw new SnapevalError(`Unknown harness "${name}". Built-in options: copilot-sdk, copilot-cli.`);
|
|
8
16
|
}
|
|
9
17
|
//# sourceMappingURL=resolve.js.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"resolve.js","sourceRoot":"","sources":["../../../../src/adapters/harness/resolve.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,iBAAiB,EAAE,MAAM,kBAAkB,CAAC;AACrD,OAAO,EAAE,aAAa,EAAE,MAAM,iBAAiB,CAAC;
|
|
1
|
+
{"version":3,"file":"resolve.js","sourceRoot":"","sources":["../../../../src/adapters/harness/resolve.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,iBAAiB,EAAE,MAAM,kBAAkB,CAAC;AACrD,OAAO,EAAE,iBAAiB,EAAE,MAAM,kBAAkB,CAAC;AACrD,OAAO,EAAE,wBAAwB,EAAE,aAAa,EAAE,MAAM,iBAAiB,CAAC;AAC1E,OAAO,EAAE,cAAc,EAAE,MAAM,0BAA0B,CAAC;AAE1D,MAAM,UAAU,cAAc,CAAC,IAAY;IACzC,IAAI,IAAI,KAAK,aAAa,EAAE,CAAC;QAC3B,IAAI,CAAC,cAAc,EAAE,EAAE,CAAC;YACtB,MAAM,IAAI,wBAAwB,CAChC,aAAa,EACb,qFAAqF,CACtF,CAAC;QACJ,CAAC;QACD,OAAO,IAAI,iBAAiB,EAAE,CAAC;IACjC,CAAC;IACD,IAAI,IAAI,KAAK,aAAa,EAAE,CAAC;QAC3B,OAAO,IAAI,iBAAiB,EAAE,CAAC;IACjC,CAAC;IACD,MAAM,IAAI,aAAa,CAAC,oBAAoB,IAAI,gDAAgD,CAAC,CAAC;AACpG,CAAC"}
|
|
@@ -3,6 +3,8 @@ export class CopilotSDKInference {
|
|
|
3
3
|
name = 'copilot-sdk';
|
|
4
4
|
async chat(messages, _options) {
|
|
5
5
|
const client = await getClient();
|
|
6
|
+
// @ts-ignore — module may not be installed (optional dep)
|
|
7
|
+
const { approveAll } = await import('@github/copilot-sdk');
|
|
6
8
|
const systemMessages = messages.filter((m) => m.role === 'system');
|
|
7
9
|
const nonSystemMessages = messages.filter((m) => m.role !== 'system');
|
|
8
10
|
const systemContent = systemMessages.map((m) => m.content).join('\n');
|
|
@@ -12,7 +14,8 @@ export class CopilotSDKInference {
|
|
|
12
14
|
...(systemContent
|
|
13
15
|
? { systemMessage: { content: systemContent } }
|
|
14
16
|
: {}),
|
|
15
|
-
onPermissionRequest:
|
|
17
|
+
onPermissionRequest: approveAll,
|
|
18
|
+
infiniteSessions: { enabled: false },
|
|
16
19
|
});
|
|
17
20
|
try {
|
|
18
21
|
const response = await session.sendAndWait({ prompt: userPrompt });
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"copilot-sdk.js","sourceRoot":"","sources":["../../../../src/adapters/inference/copilot-sdk.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,SAAS,EAAE,MAAM,0BAA0B,CAAC;AAErD,MAAM,OAAO,mBAAmB;IACrB,IAAI,GAAG,aAAa,CAAC;IAE9B,KAAK,CAAC,IAAI,CAAC,QAAmB,EAAE,QAAsB;QACpD,MAAM,MAAM,GAAG,MAAM,SAAS,EAAE,CAAC;QAEjC,MAAM,cAAc,GAAG,QAAQ,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,KAAK,QAAQ,CAAC,CAAC;QACnE,MAAM,iBAAiB,GAAG,QAAQ,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,KAAK,QAAQ,CAAC,CAAC;QACtE,MAAM,aAAa,GAAG,cAAc,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QACtE,MAAM,UAAU,GAAG,iBAAiB,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QAEtE,MAAM,OAAO,GAAG,MAAM,MAAM,CAAC,aAAa,CAAC;YACzC,KAAK,EAAE,SAAS;YAChB,GAAG,CAAC,aAAa;gBACf,CAAC,CAAC,EAAE,aAAa,EAAE,EAAE,OAAO,EAAE,aAAa,EAAE,EAAE;gBAC/C,CAAC,CAAC,EAAE,CAAC;YACP,mBAAmB,EAAE,
|
|
1
|
+
{"version":3,"file":"copilot-sdk.js","sourceRoot":"","sources":["../../../../src/adapters/inference/copilot-sdk.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,SAAS,EAAE,MAAM,0BAA0B,CAAC;AAErD,MAAM,OAAO,mBAAmB;IACrB,IAAI,GAAG,aAAa,CAAC;IAE9B,KAAK,CAAC,IAAI,CAAC,QAAmB,EAAE,QAAsB;QACpD,MAAM,MAAM,GAAG,MAAM,SAAS,EAAE,CAAC;QAEjC,0DAA0D;QAC1D,MAAM,EAAE,UAAU,EAAE,GAAG,MAAM,MAAM,CAAC,qBAAqB,CAAC,CAAC;QAE3D,MAAM,cAAc,GAAG,QAAQ,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,KAAK,QAAQ,CAAC,CAAC;QACnE,MAAM,iBAAiB,GAAG,QAAQ,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,KAAK,QAAQ,CAAC,CAAC;QACtE,MAAM,aAAa,GAAG,cAAc,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QACtE,MAAM,UAAU,GAAG,iBAAiB,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QAEtE,MAAM,OAAO,GAAG,MAAM,MAAM,CAAC,aAAa,CAAC;YACzC,KAAK,EAAE,SAAS;YAChB,GAAG,CAAC,aAAa;gBACf,CAAC,CAAC,EAAE,aAAa,EAAE,EAAE,OAAO,EAAE,aAAa,EAAE,EAAE;gBAC/C,CAAC,CAAC,EAAE,CAAC;YACP,mBAAmB,EAAE,UAAU;YAC/B,gBAAgB,EAAE,EAAE,OAAO,EAAE,KAAK,EAAE;SACrC,CAAC,CAAC;QAEH,IAAI,CAAC;YACH,MAAM,QAAQ,GAAG,MAAM,OAAO,CAAC,WAAW,CAAC,EAAE,MAAM,EAAE,UAAU,EAAE,CAAC,CAAC;YACnE,OAAO,CAAC,QAAQ,EAAE,IAAI,EAAE,OAAO,IAAI,EAAE,CAAC,CAAC,IAAI,EAAE,CAAC;QAChD,CAAC;gBAAS,CAAC;YACT,MAAM,OAAO,CAAC,UAAU,EAAE,CAAC;QAC7B,CAAC;IACH,CAAC;CACF"}
|
|
@@ -1,25 +1,105 @@
|
|
|
1
|
+
import * as fs from 'node:fs';
|
|
2
|
+
import * as path from 'node:path';
|
|
1
3
|
import chalk from 'chalk';
|
|
4
|
+
function loadPreviousIteration(iterationDir) {
|
|
5
|
+
const workspaceDir = path.dirname(iterationDir);
|
|
6
|
+
const currentName = path.basename(iterationDir);
|
|
7
|
+
const currentNum = parseInt(currentName.replace('iteration-', ''), 10);
|
|
8
|
+
if (isNaN(currentNum) || currentNum <= 1)
|
|
9
|
+
return null;
|
|
10
|
+
const prevDir = path.join(workspaceDir, `iteration-${currentNum - 1}`);
|
|
11
|
+
const prevBenchmarkPath = path.join(prevDir, 'benchmark.json');
|
|
12
|
+
if (!fs.existsSync(prevBenchmarkPath))
|
|
13
|
+
return null;
|
|
14
|
+
try {
|
|
15
|
+
const benchmark = JSON.parse(fs.readFileSync(prevBenchmarkPath, 'utf-8'));
|
|
16
|
+
const gradings = new Map();
|
|
17
|
+
const evalDirs = fs.readdirSync(prevDir).filter(d => d.startsWith('eval-'));
|
|
18
|
+
for (const evalDir of evalDirs) {
|
|
19
|
+
const wsPath = path.join(prevDir, evalDir, 'with_skill', 'grading.json');
|
|
20
|
+
const wosPath = path.join(prevDir, evalDir, 'without_skill', 'grading.json');
|
|
21
|
+
const ws = fs.existsSync(wsPath) ? JSON.parse(fs.readFileSync(wsPath, 'utf-8')) : undefined;
|
|
22
|
+
const wos = fs.existsSync(wosPath) ? JSON.parse(fs.readFileSync(wosPath, 'utf-8')) : undefined;
|
|
23
|
+
gradings.set(evalDir, { withSkill: ws, withoutSkill: wos });
|
|
24
|
+
}
|
|
25
|
+
return { benchmark, gradings };
|
|
26
|
+
}
|
|
27
|
+
catch {
|
|
28
|
+
return null;
|
|
29
|
+
}
|
|
30
|
+
}
|
|
31
|
+
function evalLabel(run) {
|
|
32
|
+
if (run.label)
|
|
33
|
+
return run.label;
|
|
34
|
+
if (run.slug && run.slug !== `${run.evalId}`)
|
|
35
|
+
return run.slug;
|
|
36
|
+
const firstLine = run.prompt.split('\n')[0].slice(0, 60);
|
|
37
|
+
return firstLine;
|
|
38
|
+
}
|
|
2
39
|
export class TerminalReporter {
|
|
3
40
|
name = 'terminal';
|
|
4
41
|
async report(results) {
|
|
5
42
|
const { skillName, evalRuns, benchmark } = results;
|
|
6
43
|
console.log(chalk.bold(`\nsnapeval — ${skillName}`));
|
|
7
|
-
console.log(chalk.dim(
|
|
44
|
+
console.log(chalk.dim(`Baseline = without SKILL.md (raw AI response)`));
|
|
45
|
+
console.log(chalk.dim('─'.repeat(60)));
|
|
46
|
+
const prev = loadPreviousIteration(results.iterationDir);
|
|
8
47
|
for (const run of evalRuns) {
|
|
9
|
-
const
|
|
48
|
+
const wsGrading = run.withSkill.grading;
|
|
49
|
+
const wsRate = wsGrading?.summary.pass_rate;
|
|
10
50
|
const wosRate = run.withoutSkill.grading?.summary.pass_rate;
|
|
11
51
|
const wsLabel = wsRate !== undefined ? `${(wsRate * 100).toFixed(0)}%` : 'n/a';
|
|
12
52
|
const wosLabel = wosRate !== undefined ? `${(wosRate * 100).toFixed(0)}%` : 'n/a';
|
|
13
|
-
const
|
|
14
|
-
const durationS = (run.withSkill.output.duration_ms / 1000).toFixed(
|
|
15
|
-
|
|
16
|
-
|
|
53
|
+
const wsColor = wsRate === 1 ? chalk.green : wsRate === 0 ? chalk.red : chalk.yellow;
|
|
54
|
+
const durationS = (run.withSkill.output.duration_ms / 1000).toFixed(1);
|
|
55
|
+
// Show per-eval delta from previous iteration
|
|
56
|
+
let perEvalDelta = '';
|
|
57
|
+
if (prev) {
|
|
58
|
+
const prevGrading = prev.gradings.get(`eval-${run.slug}`);
|
|
59
|
+
const prevRate = prevGrading?.withSkill?.summary.pass_rate;
|
|
60
|
+
if (prevRate !== undefined && wsRate !== undefined) {
|
|
61
|
+
const change = wsRate - prevRate;
|
|
62
|
+
if (change !== 0) {
|
|
63
|
+
const arrow = change > 0 ? chalk.green('↑') : chalk.red('↓');
|
|
64
|
+
perEvalDelta = ` ${arrow} was ${(prevRate * 100).toFixed(0)}%`;
|
|
65
|
+
}
|
|
66
|
+
}
|
|
67
|
+
}
|
|
68
|
+
console.log(` ${chalk.cyan(`#${run.evalId}`)} ${evalLabel(run)}`);
|
|
69
|
+
console.log(` Skill: ${wsColor(wsLabel)}${perEvalDelta} | Baseline: ${wosLabel} | ${durationS}s`);
|
|
70
|
+
// Show failed assertions inline
|
|
71
|
+
if (wsGrading) {
|
|
72
|
+
const failed = wsGrading.assertion_results.filter((a) => !a.passed);
|
|
73
|
+
for (const f of failed) {
|
|
74
|
+
console.log(chalk.red(` FAIL: ${f.text}`));
|
|
75
|
+
if (f.evidence) {
|
|
76
|
+
console.log(chalk.dim(` ${f.evidence.slice(0, 100)}`));
|
|
77
|
+
}
|
|
78
|
+
}
|
|
79
|
+
}
|
|
17
80
|
}
|
|
18
|
-
console.log(chalk.dim('─'.repeat(
|
|
81
|
+
console.log(chalk.dim('─'.repeat(60)));
|
|
82
|
+
const ws = benchmark.run_summary.with_skill;
|
|
83
|
+
const wos = benchmark.run_summary.without_skill;
|
|
19
84
|
const delta = benchmark.run_summary.delta;
|
|
20
85
|
const deltaColor = delta.pass_rate > 0 ? chalk.green : delta.pass_rate < 0 ? chalk.red : chalk.dim;
|
|
21
|
-
console.log(
|
|
22
|
-
console.log(
|
|
86
|
+
console.log(chalk.bold('Summary:'));
|
|
87
|
+
console.log(` Skill pass rate: ${(ws.pass_rate.mean * 100).toFixed(1)}%`);
|
|
88
|
+
console.log(` Baseline pass rate: ${(wos.pass_rate.mean * 100).toFixed(1)}%`);
|
|
89
|
+
console.log(` Improvement: ${deltaColor(`${delta.pass_rate > 0 ? '+' : ''}${(delta.pass_rate * 100).toFixed(1)}%`)}`);
|
|
90
|
+
if (prev) {
|
|
91
|
+
const prevRate = prev.benchmark.run_summary.with_skill.pass_rate.mean;
|
|
92
|
+
const currRate = ws.pass_rate.mean;
|
|
93
|
+
const change = currRate - prevRate;
|
|
94
|
+
const changeColor = change > 0 ? chalk.green : change < 0 ? chalk.red : chalk.dim;
|
|
95
|
+
console.log(` vs previous: ${changeColor(`${change > 0 ? '+' : ''}${(change * 100).toFixed(1)}%`)} (was ${(prevRate * 100).toFixed(1)}%)`);
|
|
96
|
+
// Note if eval set size changed
|
|
97
|
+
const prevEvalCount = prev.gradings.size;
|
|
98
|
+
const currEvalCount = evalRuns.length;
|
|
99
|
+
if (prevEvalCount !== currEvalCount) {
|
|
100
|
+
console.log(chalk.dim(` Note: eval set changed (${prevEvalCount} → ${currEvalCount} evals)`));
|
|
101
|
+
}
|
|
102
|
+
}
|
|
23
103
|
}
|
|
24
104
|
}
|
|
25
105
|
//# sourceMappingURL=terminal.js.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"terminal.js","sourceRoot":"","sources":["../../../../src/adapters/report/terminal.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,MAAM,OAAO,CAAC;
|
|
1
|
+
{"version":3,"file":"terminal.js","sourceRoot":"","sources":["../../../../src/adapters/report/terminal.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,MAAM,SAAS,CAAC;AAC9B,OAAO,KAAK,IAAI,MAAM,WAAW,CAAC;AAClC,OAAO,KAAK,MAAM,OAAO,CAAC;AAQ1B,SAAS,qBAAqB,CAAC,YAAoB;IACjD,MAAM,YAAY,GAAG,IAAI,CAAC,OAAO,CAAC,YAAY,CAAC,CAAC;IAChD,MAAM,WAAW,GAAG,IAAI,CAAC,QAAQ,CAAC,YAAY,CAAC,CAAC;IAChD,MAAM,UAAU,GAAG,QAAQ,CAAC,WAAW,CAAC,OAAO,CAAC,YAAY,EAAE,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC;IACvE,IAAI,KAAK,CAAC,UAAU,CAAC,IAAI,UAAU,IAAI,CAAC;QAAE,OAAO,IAAI,CAAC;IACtD,MAAM,OAAO,GAAG,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,aAAa,UAAU,GAAG,CAAC,EAAE,CAAC,CAAC;IACvE,MAAM,iBAAiB,GAAG,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,gBAAgB,CAAC,CAAC;IAC/D,IAAI,CAAC,EAAE,CAAC,UAAU,CAAC,iBAAiB,CAAC;QAAE,OAAO,IAAI,CAAC;IACnD,IAAI,CAAC;QACH,MAAM,SAAS,GAAG,IAAI,CAAC,KAAK,CAAC,EAAE,CAAC,YAAY,CAAC,iBAAiB,EAAE,OAAO,CAAC,CAAC,CAAC;QAC1E,MAAM,QAAQ,GAAG,IAAI,GAAG,EAAuE,CAAC;QAChG,MAAM,QAAQ,GAAG,EAAE,CAAC,WAAW,CAAC,OAAO,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,UAAU,CAAC,OAAO,CAAC,CAAC,CAAC;QAC5E,KAAK,MAAM,OAAO,IAAI,QAAQ,EAAE,CAAC;YAC/B,MAAM,MAAM,GAAG,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,OAAO,EAAE,YAAY,EAAE,cAAc,CAAC,CAAC;YACzE,MAAM,OAAO,GAAG,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,OAAO,EAAE,eAAe,EAAE,cAAc,CAAC,CAAC;YAC7E,MAAM,EAAE,GAAG,EAAE,CAAC,UAAU,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,EAAE,CAAC,YAAY,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;YAC5F,MAAM,GAAG,GAAG,EAAE,CAAC,UAAU,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,EAAE,CAAC,YAAY,CAAC,OAAO,EAAE,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;YAC/F,QAAQ,CAAC,GAAG,CAAC,OAAO,EAAE,EAAE,SAAS,EAAE,EAAE,EAAE,YAAY,EAAE,GAAG,EAAE,CAAC,CAAC;QAC9D,CAAC;QACD,OAAO,EAAE,SAAS,EAAE,QAAQ,EAAE,CAAC;IACjC,CAAC;IAAC,MAAM,CAAC;QACP,OAAO,IAAI,CAAC;IACd,CAAC;AACH,CAAC;AAED,SAAS,SAAS,CAAC,GAAqE;IACtF,IAAI,GAAG,CAAC,KAAK;QAAE,OAAO,GAAG,CAAC,KAAK,CAAC;IAChC,IAAI,GAAG,CAAC,IAAI,IAAI,GAAG,CAAC,IAAI,KAAK,GAAG,GAAG,CAAC,MAAM,EAAE;QAAE,OAAO,GAAG,CAAC,IAAI,CAAC;IAC9D,MAAM,SAAS,GAAG,GAAG,CAAC,MAAM,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;IACzD,OAAO,SAAS,CAAC;AACnB,CAAC;AAED,MAAM,OAAO,gBAAgB;IAClB,IAAI,GAAG,UAAU,CAAC;IAE3B,KAAK,CAAC,MAAM,CAAC,OAAoB;QAC/B,MAAM,EAAE,SAAS,EAAE,QAAQ,EAAE,SAAS,EAAE,GAAG,OAAO,CAAC;QAEnD,OAAO,CAAC,GAAG,CAAC,KAAK,CAAC,IAAI,CAAC,gBAAgB,SAAS,EAAE,CAAC,CAAC,CAAC;QACrD,OAAO,CAAC,GAAG,CAAC,KAAK,CAAC,GAAG,CAAC,+CAA+C,CAAC,CAAC,CAAC;QACxE,OAAO,CAAC,GAAG,CAAC,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC;QAEvC,MAAM,IAAI,GAAG,qBAAqB,CAAC,OAAO,CAAC,YAAY,CAAC,CAAC;QAEzD,KAAK,MAAM,GAAG,IAAI,QAAQ,EAAE,CAAC;YAC3B,MAAM,SAAS,GAAG,GAAG,CAAC,SAAS,CAAC,OAAO,CAAC;YACxC,MAAM,MAAM,GAAG,SAAS,EAAE,OAAO,CAAC,SAAS,CAAC;YAC5C,MAAM,OAAO,GAAG,GAAG,CAAC,YAAY,CAAC,OAAO,EAAE,OAAO,CAAC,SAAS,CAAC;YAC5D,MAAM,OAAO,GAAG,MAAM,KAAK,SAAS,CAAC,CAAC,CAAC,GAAG,CAAC,MAAM,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,KAAK,CAAC;YAC/E,MAAM,QAAQ,GAAG,OAAO,KAAK,SAAS,CAAC,CAAC,CAAC,GAAG,CAAC,OAAO,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,KAAK,CAAC;YAClF,MAAM,OAAO,GAAG,MAAM,KAAK,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,MAAM,KAAK,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,KAAK,CAAC,MAAM,CAAC;YACrF,MAAM,SAAS,GAAG,CAAC,GAAG,CAAC,SAAS,CAAC,MAAM,CAAC,WAAW,GAAG,IAAI,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC;YAEvE,8CAA8C;YAC9C,IAAI,YAAY,GAAG,EAAE,CAAC;YACtB,IAAI,IAAI,EAAE,CAAC;gBACT,MAAM,WAAW,GAAG,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC,QAAQ,GAAG,CAAC,IAAI,EAAE,CAAC,CAAC;gBAC1D,MAAM,QAAQ,GAAG,WAAW,EAAE,SAAS,EAAE,OAAO,CAAC,SAAS,CAAC;gBAC3D,IAAI,QAAQ,KAAK,SAAS,IAAI,MAAM,KAAK,SAAS,EAAE,CAAC;oBACnD,MAAM,MAAM,GAAG,MAAM,GAAG,QAAQ,CAAC;oBACjC,IAAI,MAAM,KAAK,CAAC,EAAE,CAAC;wBACjB,MAAM,KAAK,GAAG,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;wBAC7D,YAAY,GAAG,IAAI,KAAK,QAAQ,CAAC,QAAQ,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,GAAG,CAAC;oBACjE,CAAC;gBACH,CAAC;YACH,CAAC;YAED,OAAO,CAAC,GAAG,CAAC,KAAK,KAAK,CAAC,IAAI,CAAC,IAAI,GAAG,CAAC,MAAM,EAAE,CAAC,IAAI,SAAS,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC;YACnE,OAAO,CAAC,GAAG,CAAC,cAAc,OAAO,CAAC,OAAO,CAAC,GAAG,YAAY,gBAAgB,QAAQ,MAAM,SAAS,GAAG,CAAC,CAAC;YAErG,gCAAgC;YAChC,IAAI,SAAS,EAAE,CAAC;gBACd,MAAM,MAAM,GAAG,SAAS,CAAC,iBAAiB,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC;gBACpE,KAAK,MAAM,CAAC,IAAI,MAAM,EAAE,CAAC;oBACvB,OAAO,CAAC,GAAG,CAAC,KAAK,CAAC,GAAG,CAAC,aAAa,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC;oBAC9C,IAAI,CAAC,CAAC,QAAQ,EAAE,CAAC;wBACf,OAAO,CAAC,GAAG,CAAC,KAAK,CAAC,GAAG,CAAC,aAAa,CAAC,CAAC,QAAQ,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,EAAE,CAAC,CAAC,CAAC;oBAClE,CAAC;gBACH,CAAC;YACH,CAAC;QACH,CAAC;QAED,OAAO,CAAC,GAAG,CAAC,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC;QAEvC,MAAM,EAAE,GAAG,SAAS,CAAC,WAAW,CAAC,UAAU,CAAC;QAC5C,MAAM,GAAG,GAAG,SAAS,CAAC,WAAW,CAAC,aAAa,CAAC;QAChD,MAAM,KAAK,GAAG,SAAS,CAAC,WAAW,CAAC,KAAK,CAAC;QAC1C,MAAM,UAAU,GAAG,KAAK,CAAC,SAAS,GAAG,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,KAAK,CAAC,SAAS,GAAG,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,KAAK,CAAC,GAAG,CAAC;QAEnG,OAAO,CAAC,GAAG,CAAC,KAAK,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC,CAAC;QACpC,OAAO,CAAC,GAAG,CAAC,yBAAyB,CAAC,EAAE,CAAC,SAAS,CAAC,IAAI,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC;QAC9E,OAAO,CAAC,GAAG,CAAC,yBAAyB,CAAC,GAAG,CAAC,SAAS,CAAC,IAAI,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC;QAC/E,OAAO,CAAC,GAAG,CAAC,yBAAyB,UAAU,CAAC,GAAG,KAAK,CAAC,SAAS,GAAG,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,GAAG,CAAC,KAAK,CAAC,SAAS,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC;QAE9H,IAAI,IAAI,EAAE,CAAC;YACT,MAAM,QAAQ,GAAG,IAAI,CAAC,SAAS,CAAC,WAAW,CAAC,UAAU,CAAC,SAAS,CAAC,IAAI,CAAC;YACtE,MAAM,QAAQ,GAAG,EAAE,CAAC,SAAS,CAAC,IAAI,CAAC;YACnC,MAAM,MAAM,GAAG,QAAQ,GAAG,QAAQ,CAAC;YACnC,MAAM,WAAW,GAAG,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,KAAK,CAAC,GAAG,CAAC;YAClF,OAAO,CAAC,GAAG,CAAC,yBAAyB,WAAW,CAAC,GAAG,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,GAAG,CAAC,MAAM,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,GAAG,CAAC,SAAS,CAAC,QAAQ,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC;YAEnJ,gCAAgC;YAChC,MAAM,aAAa,GAAG,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC;YACzC,MAAM,aAAa,GAAG,QAAQ,CAAC,MAAM,CAAC;YACtC,IAAI,aAAa,KAAK,aAAa,EAAE,CAAC;gBACpC,OAAO,CAAC,GAAG,CAAC,KAAK,CAAC,GAAG,CAAC,6BAA6B,aAAa,MAAM,aAAa,SAAS,CAAC,CAAC,CAAC;YACjG,CAAC;QACH,CAAC;IACH,CAAC;CACF"}
|
|
@@ -4,47 +4,176 @@ import { WorkspaceManager } from '../engine/workspace.js';
|
|
|
4
4
|
import { runEval } from '../engine/runner.js';
|
|
5
5
|
import { gradeAssertions } from '../engine/grader.js';
|
|
6
6
|
import { computeBenchmark } from '../engine/aggregator.js';
|
|
7
|
-
import { SnapevalError } from '../errors.js';
|
|
7
|
+
import { SnapevalError, FileNotFoundError, ThresholdError } from '../errors.js';
|
|
8
|
+
async function runWithConcurrency(tasks, limit) {
|
|
9
|
+
const results = new Array(tasks.length);
|
|
10
|
+
let index = 0;
|
|
11
|
+
async function worker() {
|
|
12
|
+
while (index < tasks.length) {
|
|
13
|
+
const i = index++;
|
|
14
|
+
results[i] = await tasks[i]();
|
|
15
|
+
}
|
|
16
|
+
}
|
|
17
|
+
await Promise.all(Array.from({ length: Math.min(limit, tasks.length) }, worker));
|
|
18
|
+
return results;
|
|
19
|
+
}
|
|
20
|
+
const MAX_CONCURRENCY = 10;
|
|
21
|
+
/**
|
|
22
|
+
* Average pass rates across multiple grading runs.
|
|
23
|
+
* Uses the last run's assertion_results for display, but averages the
|
|
24
|
+
* pass_rate across all runs so --runs N provides statistical significance.
|
|
25
|
+
*/
|
|
26
|
+
function averageGradings(gradings) {
|
|
27
|
+
const valid = gradings.filter((g) => g !== null);
|
|
28
|
+
if (valid.length === 0)
|
|
29
|
+
return undefined;
|
|
30
|
+
if (valid.length === 1)
|
|
31
|
+
return valid[0];
|
|
32
|
+
const avgPassRate = valid.reduce((sum, g) => sum + g.summary.pass_rate, 0) / valid.length;
|
|
33
|
+
const avgPassed = valid.reduce((sum, g) => sum + g.summary.passed, 0) / valid.length;
|
|
34
|
+
const avgFailed = valid.reduce((sum, g) => sum + g.summary.failed, 0) / valid.length;
|
|
35
|
+
const last = valid[valid.length - 1];
|
|
36
|
+
return {
|
|
37
|
+
assertion_results: last.assertion_results,
|
|
38
|
+
summary: {
|
|
39
|
+
passed: Math.round(avgPassed),
|
|
40
|
+
failed: Math.round(avgFailed),
|
|
41
|
+
total: last.summary.total,
|
|
42
|
+
pass_rate: avgPassRate,
|
|
43
|
+
},
|
|
44
|
+
};
|
|
45
|
+
}
|
|
46
|
+
function validateEvalsFile(evalsFile, evalsPath) {
|
|
47
|
+
if (!evalsFile.skill_name || typeof evalsFile.skill_name !== 'string') {
|
|
48
|
+
throw new SnapevalError(`Invalid evals.json at ${evalsPath}: missing or invalid "skill_name" field.`);
|
|
49
|
+
}
|
|
50
|
+
if (!Array.isArray(evalsFile.evals)) {
|
|
51
|
+
throw new SnapevalError(`Invalid evals.json at ${evalsPath}: "evals" must be an array.`);
|
|
52
|
+
}
|
|
53
|
+
for (const [i, evalCase] of evalsFile.evals.entries()) {
|
|
54
|
+
const prefix = `Invalid evals.json at ${evalsPath}: evals[${i}]`;
|
|
55
|
+
if (typeof evalCase.id !== 'number') {
|
|
56
|
+
throw new SnapevalError(`${prefix} missing or invalid "id" (must be a number).`);
|
|
57
|
+
}
|
|
58
|
+
if (typeof evalCase.prompt !== 'string') {
|
|
59
|
+
throw new SnapevalError(`${prefix} (id:${evalCase.id}) missing "prompt" field.`);
|
|
60
|
+
}
|
|
61
|
+
if (typeof evalCase.expected_output !== 'string') {
|
|
62
|
+
throw new SnapevalError(`${prefix} (id:${evalCase.id}) missing "expected_output" field.`);
|
|
63
|
+
}
|
|
64
|
+
if (evalCase.assertions !== undefined && !Array.isArray(evalCase.assertions)) {
|
|
65
|
+
throw new SnapevalError(`${prefix} (id:${evalCase.id}) "assertions" must be an array of strings.`);
|
|
66
|
+
}
|
|
67
|
+
}
|
|
68
|
+
}
|
|
8
69
|
export async function evalCommand(skillPath, harness, inference, options) {
|
|
9
70
|
const evalsPath = path.join(skillPath, 'evals', 'evals.json');
|
|
10
71
|
if (!fs.existsSync(evalsPath)) {
|
|
11
|
-
throw new
|
|
72
|
+
throw new FileNotFoundError(evalsPath, 'Create evals/evals.json with test scenarios first');
|
|
73
|
+
}
|
|
74
|
+
let evalsFile;
|
|
75
|
+
try {
|
|
76
|
+
evalsFile = JSON.parse(fs.readFileSync(evalsPath, 'utf-8'));
|
|
77
|
+
}
|
|
78
|
+
catch {
|
|
79
|
+
throw new SnapevalError(`Invalid JSON in ${evalsPath}. Check for syntax errors (missing commas, trailing commas, etc).`);
|
|
80
|
+
}
|
|
81
|
+
validateEvalsFile(evalsFile, evalsPath);
|
|
82
|
+
// Filter to specific eval IDs if --only is provided
|
|
83
|
+
if (options.only && options.only.length > 0) {
|
|
84
|
+
const ids = new Set(options.only);
|
|
85
|
+
const filtered = evalsFile.evals.filter((e) => ids.has(e.id));
|
|
86
|
+
if (filtered.length === 0) {
|
|
87
|
+
throw new SnapevalError(`No eval cases match --only ${options.only.join(',')}. Available IDs: ${evalsFile.evals.map((e) => e.id).join(', ')}`);
|
|
88
|
+
}
|
|
89
|
+
evalsFile = { ...evalsFile, evals: filtered };
|
|
90
|
+
}
|
|
91
|
+
if (options.threshold !== undefined && (options.threshold < 0 || options.threshold > 1)) {
|
|
92
|
+
throw new SnapevalError(`Threshold must be between 0 and 1 (e.g., 0.8 for 80%). Got: ${options.threshold}`);
|
|
12
93
|
}
|
|
13
|
-
const evalsFile = JSON.parse(fs.readFileSync(evalsPath, 'utf-8'));
|
|
14
94
|
const ws = new WorkspaceManager(skillPath, options.workspace);
|
|
15
95
|
const iterationDir = ws.createIteration();
|
|
96
|
+
// Track which SKILL.md was used for this iteration
|
|
97
|
+
const skillMdPath = path.join(skillPath, 'SKILL.md');
|
|
98
|
+
if (fs.existsSync(skillMdPath)) {
|
|
99
|
+
fs.copyFileSync(skillMdPath, path.join(iterationDir, 'SKILL.md.snapshot'));
|
|
100
|
+
}
|
|
16
101
|
const runs = options.runs ?? 1;
|
|
102
|
+
const concurrency = Math.min(Math.max(options.concurrency ?? 1, 1), MAX_CONCURRENCY);
|
|
17
103
|
const baselineVariant = options.oldSkill ? 'old_skill' : 'without_skill';
|
|
18
104
|
const scriptsDir = path.join(skillPath, 'evals', 'scripts');
|
|
19
|
-
|
|
20
|
-
|
|
105
|
+
// Pre-create eval directories sequentially (filesystem setup)
|
|
106
|
+
const evalDirs = evalsFile.evals.map((evalCase) => {
|
|
21
107
|
const slug = WorkspaceManager.getEvalSlug(evalCase).replace('eval-', '');
|
|
22
|
-
|
|
108
|
+
return { evalCase, slug, evalDir: ws.createEvalDir(iterationDir, slug, baselineVariant) };
|
|
109
|
+
});
|
|
110
|
+
const tasks = evalDirs.map(({ evalCase, slug, evalDir }) => async () => {
|
|
111
|
+
const assertions = evalCase.assertions ?? [];
|
|
112
|
+
const allGradings = [];
|
|
23
113
|
let lastRun = null;
|
|
24
114
|
for (let i = 0; i < runs; i++) {
|
|
25
115
|
lastRun = await runEval(evalCase, skillPath, evalDir, harness, options.oldSkill);
|
|
116
|
+
// Grade every run, not just the last
|
|
117
|
+
const [wsGrading, wosGrading] = await Promise.all([
|
|
118
|
+
gradeAssertions(assertions, lastRun.withSkill.output, path.join(evalDir, 'with_skill'), inference, fs.existsSync(scriptsDir) ? scriptsDir : undefined),
|
|
119
|
+
gradeAssertions(assertions, lastRun.withoutSkill.output, path.join(evalDir, baselineVariant), inference, fs.existsSync(scriptsDir) ? scriptsDir : undefined),
|
|
120
|
+
]);
|
|
121
|
+
allGradings.push({ withSkill: wsGrading, withoutSkill: wosGrading });
|
|
26
122
|
}
|
|
27
|
-
if (!lastRun)
|
|
28
|
-
|
|
29
|
-
|
|
30
|
-
|
|
31
|
-
const
|
|
32
|
-
|
|
123
|
+
if (!lastRun) {
|
|
124
|
+
throw new SnapevalError(`No runs completed for eval ${evalCase.id}`);
|
|
125
|
+
}
|
|
126
|
+
// Average pass rates across all runs for statistical significance
|
|
127
|
+
const withSkillGrading = averageGradings(allGradings.map(g => g.withSkill));
|
|
128
|
+
const withoutSkillGrading = averageGradings(allGradings.map(g => g.withoutSkill));
|
|
129
|
+
// When runs > 1, overwrite grading.json with averaged results so
|
|
130
|
+
// artifacts match the benchmark (not just the last run's raw data)
|
|
131
|
+
if (runs > 1) {
|
|
132
|
+
if (withSkillGrading) {
|
|
133
|
+
fs.writeFileSync(path.join(evalDir, 'with_skill', 'grading.json'), JSON.stringify(withSkillGrading, null, 2));
|
|
134
|
+
}
|
|
135
|
+
if (withoutSkillGrading) {
|
|
136
|
+
fs.writeFileSync(path.join(evalDir, baselineVariant, 'grading.json'), JSON.stringify(withoutSkillGrading, null, 2));
|
|
137
|
+
}
|
|
138
|
+
}
|
|
139
|
+
return {
|
|
33
140
|
evalId: evalCase.id,
|
|
34
141
|
slug,
|
|
142
|
+
label: evalCase.label,
|
|
35
143
|
prompt: evalCase.prompt,
|
|
36
144
|
withSkill: {
|
|
37
145
|
output: lastRun.withSkill.output,
|
|
38
|
-
grading: withSkillGrading
|
|
146
|
+
grading: withSkillGrading,
|
|
39
147
|
},
|
|
40
148
|
withoutSkill: {
|
|
41
149
|
output: lastRun.withoutSkill.output,
|
|
42
|
-
grading: withoutSkillGrading
|
|
150
|
+
grading: withoutSkillGrading,
|
|
43
151
|
},
|
|
44
|
-
}
|
|
45
|
-
}
|
|
152
|
+
};
|
|
153
|
+
});
|
|
154
|
+
const evalRuns = await runWithConcurrency(tasks, concurrency);
|
|
46
155
|
const benchmark = computeBenchmark(evalRuns);
|
|
47
|
-
|
|
156
|
+
// Add iteration metadata for cross-iteration comparison
|
|
157
|
+
const benchmarkWithMeta = {
|
|
158
|
+
...benchmark,
|
|
159
|
+
metadata: {
|
|
160
|
+
eval_count: evalRuns.length,
|
|
161
|
+
eval_ids: evalRuns.map((r) => r.evalId),
|
|
162
|
+
skill_name: evalsFile.skill_name,
|
|
163
|
+
runs_per_eval: runs,
|
|
164
|
+
timestamp: new Date().toISOString(),
|
|
165
|
+
},
|
|
166
|
+
};
|
|
167
|
+
fs.writeFileSync(path.join(iterationDir, 'benchmark.json'), JSON.stringify(benchmarkWithMeta, (_key, value) => typeof value === 'number' ? Math.round(value * 10000) / 10000 : value, 2));
|
|
168
|
+
// Check threshold if set (for CI gating)
|
|
169
|
+
if (options.threshold !== undefined) {
|
|
170
|
+
const passRate = benchmark.run_summary.with_skill.pass_rate.mean;
|
|
171
|
+
if (passRate < options.threshold) {
|
|
172
|
+
// Still return results so the reporter can display them before the error
|
|
173
|
+
const results = { skillName: evalsFile.skill_name, evalRuns, benchmark, iterationDir };
|
|
174
|
+
throw Object.assign(new ThresholdError(passRate, options.threshold), { results });
|
|
175
|
+
}
|
|
176
|
+
}
|
|
48
177
|
return {
|
|
49
178
|
skillName: evalsFile.skill_name,
|
|
50
179
|
evalRuns,
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"eval.js","sourceRoot":"","sources":["../../../src/commands/eval.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,MAAM,SAAS,CAAC;AAC9B,OAAO,KAAK,IAAI,MAAM,WAAW,CAAC;
|
|
1
|
+
{"version":3,"file":"eval.js","sourceRoot":"","sources":["../../../src/commands/eval.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,MAAM,SAAS,CAAC;AAC9B,OAAO,KAAK,IAAI,MAAM,WAAW,CAAC;AASlC,OAAO,EAAE,gBAAgB,EAAE,MAAM,wBAAwB,CAAC;AAC1D,OAAO,EAAE,OAAO,EAAE,MAAM,qBAAqB,CAAC;AAC9C,OAAO,EAAE,eAAe,EAAE,MAAM,qBAAqB,CAAC;AACtD,OAAO,EAAE,gBAAgB,EAAE,MAAM,yBAAyB,CAAC;AAC3D,OAAO,EAAE,aAAa,EAAE,iBAAiB,EAAE,cAAc,EAAE,MAAM,cAAc,CAAC;AAEhF,KAAK,UAAU,kBAAkB,CAC/B,KAA2B,EAC3B,KAAa;IAEb,MAAM,OAAO,GAAQ,IAAI,KAAK,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC;IAC7C,IAAI,KAAK,GAAG,CAAC,CAAC;IACd,KAAK,UAAU,MAAM;QACnB,OAAO,KAAK,GAAG,KAAK,CAAC,MAAM,EAAE,CAAC;YAC5B,MAAM,CAAC,GAAG,KAAK,EAAE,CAAC;YAClB,OAAO,CAAC,CAAC,CAAC,GAAG,MAAM,KAAK,CAAC,CAAC,CAAC,EAAE,CAAC;QAChC,CAAC;IACH,CAAC;IACD,MAAM,OAAO,CAAC,GAAG,CAAC,KAAK,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,IAAI,CAAC,GAAG,CAAC,KAAK,EAAE,KAAK,CAAC,MAAM,CAAC,EAAE,EAAE,MAAM,CAAC,CAAC,CAAC;IACjF,OAAO,OAAO,CAAC;AACjB,CAAC;AAED,MAAM,eAAe,GAAG,EAAE,CAAC;AAE3B;;;;GAIG;AACH,SAAS,eAAe,CAAC,QAAkC;IACzD,MAAM,KAAK,GAAG,QAAQ,CAAC,MAAM,CAAC,CAAC,CAAC,EAAsB,EAAE,CAAC,CAAC,KAAK,IAAI,CAAC,CAAC;IACrE,IAAI,KAAK,CAAC,MAAM,KAAK,CAAC;QAAE,OAAO,SAAS,CAAC;IACzC,IAAI,KAAK,CAAC,MAAM,KAAK,CAAC;QAAE,OAAO,KAAK,CAAC,CAAC,CAAC,CAAC;IAExC,MAAM,WAAW,GAAG,KAAK,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,CAAC,EAAE,EAAE,CAAC,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,SAAS,EAAE,CAAC,CAAC,GAAG,KAAK,CAAC,MAAM,CAAC;IAC1F,MAAM,SAAS,GAAG,KAAK,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,CAAC,EAAE,EAAE,CAAC,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,MAAM,EAAE,CAAC,CAAC,GAAG,KAAK,CAAC,MAAM,CAAC;IACrF,MAAM,SAAS,GAAG,KAAK,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,CAAC,EAAE,EAAE,CAAC,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,MAAM,EAAE,CAAC,CAAC,GAAG,KAAK,CAAC,MAAM,CAAC;IACrF,MAAM,IAAI,GAAG,KAAK,CAAC,KAAK,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;IAErC,OAAO;QACL,iBAAiB,EAAE,IAAI,CAAC,iBAAiB;QACzC,OAAO,EAAE;YACP,MAAM,EAAE,IAAI,CAAC,KAAK,CAAC,SAAS,CAAC;YAC7B,MAAM,EAAE,IAAI,CAAC,KAAK,CAAC,SAAS,CAAC;YAC7B,KAAK,EAAE,IAAI,CAAC,OAAO,CAAC,KAAK;YACzB,SAAS,EAAE,WAAW;SACvB;KACF,CAAC;AACJ,CAAC;AAED,SAAS,iBAAiB,CAAC,SAAoB,EAAE,SAAiB;IAChE,IAAI,CAAC,SAAS,CAAC,UAAU,IAAI,OAAO,SAAS,CAAC,UAAU,KAAK,QAAQ,EAAE,CAAC;QACtE,MAAM,IAAI,aAAa,CAAC,yBAAyB,SAAS,0CAA0C,CAAC,CAAC;IACxG,CAAC;IACD,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,SAAS,CAAC,KAAK,CAAC,EAAE,CAAC;QACpC,MAAM,IAAI,aAAa,CAAC,yBAAyB,SAAS,6BAA6B,CAAC,CAAC;IAC3F,CAAC;IACD,KAAK,MAAM,CAAC,CAAC,EAAE,QAAQ,CAAC,IAAI,SAAS,CAAC,KAAK,CAAC,OAAO,EAAE,EAAE,CAAC;QACtD,MAAM,MAAM,GAAG,yBAAyB,SAAS,WAAW,CAAC,GAAG,CAAC;QACjE,IAAI,OAAO,QAAQ,CAAC,EAAE,KAAK,QAAQ,EAAE,CAAC;YACpC,MAAM,IAAI,aAAa,CAAC,GAAG,MAAM,8CAA8C,CAAC,CAAC;QACnF,CAAC;QACD,IAAI,OAAO,QAAQ,CAAC,MAAM,KAAK,QAAQ,EAAE,CAAC;YACxC,MAAM,IAAI,aAAa,CAAC,GAAG,MAAM,QAAQ,QAAQ,CAAC,EAAE,2BAA2B,CAAC,CAAC;QACnF,CAAC;QACD,IAAI,OAAO,QAAQ,CAAC,eAAe,KAAK,QAAQ,EAAE,CAAC;YACjD,MAAM,IAAI,aAAa,CAAC,GAAG,MAAM,QAAQ,QAAQ,CAAC,EAAE,oCAAoC,CAAC,CAAC;QAC5F,CAAC;QACD,IAAI,QAAQ,CAAC,UAAU,KAAK,SAAS,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,QAAQ,CAAC,UAAU,CAAC,EAAE,CAAC;YAC7E,MAAM,IAAI,aAAa,CAAC,GAAG,MAAM,QAAQ,QAAQ,CAAC,EAAE,6CAA6C,CAAC,CAAC;QACrG,CAAC;IACH,CAAC;AACH,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,WAAW,CAC/B,SAAiB,EACjB,OAAgB,EAChB,SAA2B,EAC3B,OAA4H;IAE5H,MAAM,SAAS,GAAG,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,OAAO,EAAE,YAAY,CAAC,CAAC;IAC9D,IAAI,CAAC,EAAE,CAAC,UAAU,CAAC,SAAS,CAAC,EAAE,CAAC;QAC9B,MAAM,IAAI,iBAAiB,CAAC,SAAS,EAAE,mDAAmD,CAAC,CAAC;IAC9F,CAAC;IAED,IAAI,SAAoB,CAAC;IACzB,IAAI,CAAC;QACH,SAAS,GAAG,IAAI,CAAC,KAAK,CAAC,EAAE,CAAC,YAAY,CAAC,SAAS,EAAE,OAAO,CAAC,CAAC,CAAC;IAC9D,CAAC;IAAC,MAAM,CAAC;QACP,MAAM,IAAI,aAAa,CAAC,mBAAmB,SAAS,mEAAmE,CAAC,CAAC;IAC3H,CAAC;IACD,iBAAiB,CAAC,SAAS,EAAE,SAAS,CAAC,CAAC;IAExC,oDAAoD;IACpD,IAAI,OAAO,CAAC,IAAI,IAAI,OAAO,CAAC,IAAI,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;QAC5C,MAAM,GAAG,GAAG,IAAI,GAAG,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC;QAClC,MAAM,QAAQ,GAAG,SAAS,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC;QAC9D,IAAI,QAAQ,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YAC1B,MAAM,IAAI,aAAa,CAAC,8BAA8B,OAAO,CAAC,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC,oBAAoB,SAAS,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;QACjJ,CAAC;QACD,SAAS,GAAG,EAAE,GAAG,SAAS,EAAE,KAAK,EAAE,QAAQ,EAAE,CAAC;IAChD,CAAC;IAED,IAAI,OAAO,CAAC,SAAS,KAAK,SAAS,IAAI,CAAC,OAAO,CAAC,SAAS,GAAG,CAAC,IAAI,OAAO,CAAC,SAAS,GAAG,CAAC,CAAC,EAAE,CAAC;QACxF,MAAM,IAAI,aAAa,CAAC,+DAA+D,OAAO,CAAC,SAAS,EAAE,CAAC,CAAC;IAC9G,CAAC;IAED,MAAM,EAAE,GAAG,IAAI,gBAAgB,CAAC,SAAS,EAAE,OAAO,CAAC,SAAS,CAAC,CAAC;IAC9D,MAAM,YAAY,GAAG,EAAE,CAAC,eAAe,EAAE,CAAC;IAE1C,mDAAmD;IACnD,MAAM,WAAW,GAAG,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,UAAU,CAAC,CAAC;IACrD,IAAI,EAAE,CAAC,UAAU,CAAC,WAAW,CAAC,EAAE,CAAC;QAC/B,EAAE,CAAC,YAAY,CAAC,WAAW,EAAE,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,mBAAmB,CAAC,CAAC,CAAC;IAC7E,CAAC;IACD,MAAM,IAAI,GAAG,OAAO,CAAC,IAAI,IAAI,CAAC,CAAC;IAC/B,MAAM,WAAW,GAAG,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,OAAO,CAAC,WAAW,IAAI,CAAC,EAAE,CAAC,CAAC,EAAE,eAAe,CAAC,CAAC;IACrF,MAAM,eAAe,GAAG,OAAO,CAAC,QAAQ,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC,CAAC,eAAe,CAAC;IACzE,MAAM,UAAU,GAAG,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,OAAO,EAAE,SAAS,CAAC,CAAC;IAE5D,8DAA8D;IAC9D,MAAM,QAAQ,GAAG,SAAS,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,QAAQ,EAAE,EAAE;QAChD,MAAM,IAAI,GAAG,gBAAgB,CAAC,WAAW,CAAC,QAAQ,CAAC,CAAC,OAAO,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC;QACzE,OAAO,EAAE,QAAQ,EAAE,IAAI,EAAE,OAAO,EAAE,EAAE,CAAC,aAAa,CAAC,YAAY,EAAE,IAAI,EAAE,eAAe,CAAC,EAAE,CAAC;IAC5F,CAAC,CAAC,CAAC;IAEH,MAAM,KAAK,GAAG,QAAQ,CAAC,GAAG,CAAC,CAAC,EAAE,QAAQ,EAAE,IAAI,EAAE,OAAO,EAAE,EAAE,EAAE,CAAC,KAAK,IAA4B,EAAE;QAC7F,MAAM,UAAU,GAAG,QAAQ,CAAC,UAAU,IAAI,EAAE,CAAC;QAC7C,MAAM,WAAW,GAA8E,EAAE,CAAC;QAClG,IAAI,OAAO,GAA+C,IAAI,CAAC;QAE/D,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,IAAI,EAAE,CAAC,EAAE,EAAE,CAAC;YAC9B,OAAO,GAAG,MAAM,OAAO,CAAC,QAAQ,EAAE,SAAS,EAAE,OAAO,EAAE,OAAO,EAAE,OAAO,CAAC,QAAQ,CAAC,CAAC;YAEjF,qCAAqC;YACrC,MAAM,CAAC,SAAS,EAAE,UAAU,CAAC,GAAG,MAAM,OAAO,CAAC,GAAG,CAAC;gBAChD,eAAe,CACb,UAAU,EACV,OAAO,CAAC,SAAS,CAAC,MAAM,EACxB,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,YAAY,CAAC,EAChC,SAAS,EACT,EAAE,CAAC,UAAU,CAAC,UAAU,CAAC,CAAC,CAAC,CAAC,UAAU,CAAC,CAAC,CAAC,SAAS,CACnD;gBACD,eAAe,CACb,UAAU,EACV,OAAO,CAAC,YAAY,CAAC,MAAM,EAC3B,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,eAAe,CAAC,EACnC,SAAS,EACT,EAAE,CAAC,UAAU,CAAC,UAAU,CAAC,CAAC,CAAC,CAAC,UAAU,CAAC,CAAC,CAAC,SAAS,CACnD;aACF,CAAC,CAAC;YACH,WAAW,CAAC,IAAI,CAAC,EAAE,SAAS,EAAE,SAAS,EAAE,YAAY,EAAE,UAAU,EAAE,CAAC,CAAC;QACvE,CAAC;QAED,IAAI,CAAC,OAAO,EAAE,CAAC;YACb,MAAM,IAAI,aAAa,CAAC,8BAA8B,QAAQ,CAAC,EAAE,EAAE,CAAC,CAAC;QACvE,CAAC;QAED,kEAAkE;QAClE,MAAM,gBAAgB,GAAG,eAAe,CAAC,WAAW,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC;QAC5E,MAAM,mBAAmB,GAAG,eAAe,CAAC,WAAW,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,YAAY,CAAC,CAAC,CAAC;QAElF,iEAAiE;QACjE,mEAAmE;QACnE,IAAI,IAAI,GAAG,CAAC,EAAE,CAAC;YACb,IAAI,gBAAgB,EAAE,CAAC;gBACrB,EAAE,CAAC,aAAa,CACd,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,YAAY,EAAE,cAAc,CAAC,EAChD,IAAI,CAAC,SAAS,CAAC,gBAAgB,EAAE,IAAI,EAAE,CAAC,CAAC,CAC1C,CAAC;YACJ,CAAC;YACD,IAAI,mBAAmB,EAAE,CAAC;gBACxB,EAAE,CAAC,aAAa,CACd,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,eAAe,EAAE,cAAc,CAAC,EACnD,IAAI,CAAC,SAAS,CAAC,mBAAmB,EAAE,IAAI,EAAE,CAAC,CAAC,CAC7C,CAAC;YACJ,CAAC;QACH,CAAC;QAED,OAAO;YACL,MAAM,EAAE,QAAQ,CAAC,EAAE;YACnB,IAAI;YACJ,KAAK,EAAE,QAAQ,CAAC,KAAK;YACrB,MAAM,EAAE,QAAQ,CAAC,MAAM;YACvB,SAAS,EAAE;gBACT,MAAM,EAAE,OAAO,CAAC,SAAS,CAAC,MAAM;gBAChC,OAAO,EAAE,gBAAgB;aAC1B;YACD,YAAY,EAAE;gBACZ,MAAM,EAAE,OAAO,CAAC,YAAY,CAAC,MAAM;gBACnC,OAAO,EAAE,mBAAmB;aAC7B;SACF,CAAC;IACJ,CAAC,CAAC,CAAC;IAEH,MAAM,QAAQ,GAAG,MAAM,kBAAkB,CAAC,KAAK,EAAE,WAAW,CAAC,CAAC;IAC9D,MAAM,SAAS,GAAG,gBAAgB,CAAC,QAAQ,CAAC,CAAC;IAE7C,wDAAwD;IACxD,MAAM,iBAAiB,GAAG;QACxB,GAAG,SAAS;QACZ,QAAQ,EAAE;YACR,UAAU,EAAE,QAAQ,CAAC,MAAM;YAC3B,QAAQ,EAAE,QAAQ,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC;YACvC,UAAU,EAAE,SAAS,CAAC,UAAU;YAChC,aAAa,EAAE,IAAI;YACnB,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;SACpC;KACF,CAAC;IAEF,EAAE,CAAC,aAAa,CACd,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,gBAAgB,CAAC,EACzC,IAAI,CAAC,SAAS,CAAC,iBAAiB,EAAE,CAAC,IAAI,EAAE,KAAK,EAAE,EAAE,CAChD,OAAO,KAAK,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,KAAK,GAAG,KAAK,CAAC,GAAG,KAAK,CAAC,CAAC,CAAC,KAAK,EAAE,CAAC,CAAC,CAC5E,CAAC;IAEF,yCAAyC;IACzC,IAAI,OAAO,CAAC,SAAS,KAAK,SAAS,EAAE,CAAC;QACpC,MAAM,QAAQ,GAAG,SAAS,CAAC,WAAW,CAAC,UAAU,CAAC,SAAS,CAAC,IAAI,CAAC;QACjE,IAAI,QAAQ,GAAG,OAAO,CAAC,SAAS,EAAE,CAAC;YACjC,yEAAyE;YACzE,MAAM,OAAO,GAAG,EAAE,SAAS,EAAE,SAAS,CAAC,UAAU,EAAE,QAAQ,EAAE,SAAS,EAAE,YAAY,EAAE,CAAC;YACvF,MAAM,MAAM,CAAC,MAAM,CAAC,IAAI,cAAc,CAAC,QAAQ,EAAE,OAAO,CAAC,SAAS,CAAC,EAAE,EAAE,OAAO,EAAE,CAAC,CAAC;QACpF,CAAC;IACH,CAAC;IAED,OAAO;QACL,SAAS,EAAE,SAAS,CAAC,UAAU;QAC/B,QAAQ;QACR,SAAS;QACT,YAAY;KACb,CAAC;AACJ,CAAC"}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"review.js","sourceRoot":"","sources":["../../../src/commands/review.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,QAAQ,EAAE,MAAM,oBAAoB,CAAC;AAC9C,OAAO,KAAK,EAAE,MAAM,SAAS,CAAC;AAC9B,OAAO,KAAK,IAAI,MAAM,WAAW,CAAC;AAClC,OAAO,KAAK,OAAO,MAAM,cAAc,CAAC;AAExC,OAAO,EAAE,WAAW,EAAE,MAAM,WAAW,CAAC;AACxC,OAAO,EAAE,gBAAgB,EAAE,MAAM,gCAAgC,CAAC;AAElE,MAAM,CAAC,KAAK,UAAU,aAAa,CACjC,SAAiB,EACjB,OAAgB,EAChB,SAA2B,EAC3B,
|
|
1
|
+
{"version":3,"file":"review.js","sourceRoot":"","sources":["../../../src/commands/review.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,QAAQ,EAAE,MAAM,oBAAoB,CAAC;AAC9C,OAAO,KAAK,EAAE,MAAM,SAAS,CAAC;AAC9B,OAAO,KAAK,IAAI,MAAM,WAAW,CAAC;AAClC,OAAO,KAAK,OAAO,MAAM,cAAc,CAAC;AAExC,OAAO,EAAE,WAAW,EAAE,MAAM,WAAW,CAAC;AACxC,OAAO,EAAE,gBAAgB,EAAE,MAAM,gCAAgC,CAAC;AAElE,MAAM,CAAC,KAAK,UAAU,aAAa,CACjC,SAAiB,EACjB,OAAgB,EAChB,SAA2B,EAC3B,OAAyG;IAEzG,MAAM,OAAO,GAAG,MAAM,WAAW,CAAC,SAAS,EAAE,OAAO,EAAE,SAAS,EAAE,OAAO,CAAC,CAAC;IAE1E,MAAM,QAAQ,GAAG,IAAI,gBAAgB,EAAE,CAAC;IACxC,MAAM,QAAQ,CAAC,MAAM,CAAC,OAAO,CAAC,CAAC;IAE/B,yBAAyB;IACzB,MAAM,QAAQ,GAAiB,EAAE,CAAC;IAClC,KAAK,MAAM,GAAG,IAAI,OAAO,CAAC,QAAQ,EAAE,CAAC;QACnC,QAAQ,CAAC,QAAQ,GAAG,CAAC,IAAI,EAAE,CAAC,GAAG,EAAE,CAAC;IACpC,CAAC;IACD,EAAE,CAAC,aAAa,CACd,IAAI,CAAC,IAAI,CAAC,OAAO,CAAC,YAAY,EAAE,eAAe,CAAC,EAChD,IAAI,CAAC,SAAS,CAAC,QAAQ,EAAE,IAAI,EAAE,CAAC,CAAC,CAClC,CAAC;IAEF,oEAAoE;IACpE,IAAI,CAAC,OAAO,CAAC,MAAM,EAAE,CAAC;QACpB,MAAM,UAAU,GAAG,IAAI,CAAC,IAAI,CAAC,OAAO,CAAC,YAAY,EAAE,gBAAgB,CAAC,CAAC;QACrE,aAAa,CAAC,UAAU,CAAC,CAAC;IAC5B,CAAC;AACH,CAAC;AAED,SAAS,aAAa,CAAC,QAAgB;IACrC,MAAM,GAAG,GACP,OAAO,CAAC,QAAQ,KAAK,QAAQ,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC;QACxC,OAAO,CAAC,QAAQ,KAAK,OAAO,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,UAAU,CAAC;IACpD,MAAM,IAAI,GACR,OAAO,CAAC,QAAQ,KAAK,OAAO,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,OAAO,EAAE,EAAE,EAAE,QAAQ,CAAC,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC;IAC5E,QAAQ,CAAC,GAAG,EAAE,IAAI,EAAE,CAAC,GAAG,EAAE,EAAE;QAC1B,IAAI,GAAG;YAAE,OAAO,CAAC,IAAI,CAAC,2BAA2B,GAAG,CAAC,OAAO,EAAE,CAAC,CAAC;IAClE,CAAC,CAAC,CAAC;AACL,CAAC"}
|
package/dist/src/config.js
CHANGED
|
@@ -1,10 +1,11 @@
|
|
|
1
1
|
import * as fs from 'node:fs';
|
|
2
2
|
import * as path from 'node:path';
|
|
3
3
|
export const DEFAULT_CONFIG = {
|
|
4
|
-
harness: 'copilot-
|
|
4
|
+
harness: 'copilot-sdk',
|
|
5
5
|
inference: 'auto',
|
|
6
6
|
workspace: '../{skill_name}-workspace',
|
|
7
7
|
runs: 1,
|
|
8
|
+
concurrency: 1,
|
|
8
9
|
};
|
|
9
10
|
function loadConfigFile(dirPath) {
|
|
10
11
|
const configPath = path.join(dirPath, 'snapeval.config.json');
|
package/dist/src/config.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"config.js","sourceRoot":"","sources":["../../src/config.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,MAAM,SAAS,CAAC;AAC9B,OAAO,KAAK,IAAI,MAAM,WAAW,CAAC;AAGlC,MAAM,CAAC,MAAM,cAAc,GAAmB;IAC5C,OAAO,EAAE,aAAa;IACtB,SAAS,EAAE,MAAM;IACjB,SAAS,EAAE,2BAA2B;IACtC,IAAI,EAAE,CAAC;
|
|
1
|
+
{"version":3,"file":"config.js","sourceRoot":"","sources":["../../src/config.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,MAAM,SAAS,CAAC;AAC9B,OAAO,KAAK,IAAI,MAAM,WAAW,CAAC;AAGlC,MAAM,CAAC,MAAM,cAAc,GAAmB;IAC5C,OAAO,EAAE,aAAa;IACtB,SAAS,EAAE,MAAM;IACjB,SAAS,EAAE,2BAA2B;IACtC,IAAI,EAAE,CAAC;IACP,WAAW,EAAE,CAAC;CACf,CAAC;AAEF,SAAS,cAAc,CAAC,OAAe;IACrC,MAAM,UAAU,GAAG,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,sBAAsB,CAAC,CAAC;IAC9D,IAAI,CAAC,EAAE,CAAC,UAAU,CAAC,UAAU,CAAC;QAAE,OAAO,IAAI,CAAC;IAC5C,MAAM,GAAG,GAAG,EAAE,CAAC,YAAY,CAAC,UAAU,EAAE,OAAO,CAAC,CAAC;IACjD,OAAO,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC;AACzB,CAAC;AAED,MAAM,UAAU,aAAa,CAC3B,QAAiC,EACjC,WAAmB,EACnB,QAAiB;IAEjB,MAAM,cAAc,GAAG,QAAQ,CAAC,CAAC,CAAC,cAAc,CAAC,QAAQ,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC;IAClE,MAAM,aAAa,GAAG,cAAc,CAAC,WAAW,CAAC,CAAC;IAClD,OAAO;QACL,GAAG,cAAc;QACjB,GAAG,CAAC,aAAa,IAAI,EAAE,CAAC;QACxB,GAAG,CAAC,cAAc,IAAI,EAAE,CAAC;QACzB,GAAG,cAAc,CAAC,QAAQ,CAAC;KAC5B,CAAC;AACJ,CAAC;AAED,SAAS,cAAc,CAAC,GAA4B;IAClD,OAAO,MAAM,CAAC,WAAW,CAAC,MAAM,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,KAAK,SAAS,CAAC,CAAC,CAAC;AACpF,CAAC"}
|