@artemiskit/core 0.2.4 → 0.4.0
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/CHANGELOG.md +137 -0
- package/README.md +4 -0
- package/dist/adapters/registry.d.ts.map +1 -1
- package/dist/adapters/types.d.ts +50 -2
- package/dist/adapters/types.d.ts.map +1 -1
- package/dist/agent-evaluation/index.d.ts +3 -0
- package/dist/agent-evaluation/index.d.ts.map +1 -0
- package/dist/agent-evaluation/scorer.d.ts +35 -0
- package/dist/agent-evaluation/scorer.d.ts.map +1 -0
- package/dist/agent-evaluation/types.d.ts +37 -0
- package/dist/agent-evaluation/types.d.ts.map +1 -0
- package/dist/artifacts/manifest.d.ts.map +1 -1
- package/dist/artifacts/types.d.ts +54 -0
- package/dist/artifacts/types.d.ts.map +1 -1
- package/dist/evaluators/index.d.ts +1 -0
- package/dist/evaluators/index.d.ts.map +1 -1
- package/dist/evaluators/json-schema.d.ts +0 -1
- package/dist/evaluators/json-schema.d.ts.map +1 -1
- package/dist/evaluators/llm-grader.d.ts +2 -0
- package/dist/evaluators/llm-grader.d.ts.map +1 -1
- package/dist/evaluators/tool-trace.d.ts +7 -0
- package/dist/evaluators/tool-trace.d.ts.map +1 -0
- package/dist/evaluators/types.d.ts +20 -0
- package/dist/evaluators/types.d.ts.map +1 -1
- package/dist/index.d.ts +2 -0
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +19964 -12482
- package/dist/runner/executor.d.ts.map +1 -1
- package/dist/runner/runner.d.ts.map +1 -1
- package/dist/runner/types.d.ts +4 -0
- package/dist/runner/types.d.ts.map +1 -1
- package/dist/scenario/schema.d.ts +789 -99
- package/dist/scenario/schema.d.ts.map +1 -1
- package/dist/storage/supabase.d.ts +26 -5
- package/dist/storage/supabase.d.ts.map +1 -1
- package/dist/storage/types.d.ts +167 -1
- package/dist/storage/types.d.ts.map +1 -1
- package/dist/tools/fixture-executor.d.ts +10 -0
- package/dist/tools/fixture-executor.d.ts.map +1 -0
- package/dist/tools/index.d.ts +4 -0
- package/dist/tools/index.d.ts.map +1 -0
- package/dist/tools/schema-validator.d.ts +10 -0
- package/dist/tools/schema-validator.d.ts.map +1 -0
- package/dist/tools/types.d.ts +50 -0
- package/dist/tools/types.d.ts.map +1 -0
- package/package.json +2 -1
- package/src/adapters/registry.ts +45 -0
- package/src/adapters/types.test.ts +21 -0
- package/src/adapters/types.ts +56 -0
- package/src/agent-evaluation/index.ts +2 -0
- package/src/agent-evaluation/scorer.test.ts +1194 -0
- package/src/agent-evaluation/scorer.ts +640 -0
- package/src/agent-evaluation/types.test.ts +27 -0
- package/src/agent-evaluation/types.ts +43 -0
- package/src/artifacts/manifest.test.ts +66 -19
- package/src/artifacts/manifest.ts +18 -5
- package/src/artifacts/types.ts +65 -0
- package/src/evaluators/index.ts +3 -0
- package/src/evaluators/json-schema.test.ts +130 -0
- package/src/evaluators/json-schema.ts +38 -63
- package/src/evaluators/llm-grader.test.ts +80 -0
- package/src/evaluators/llm-grader.ts +44 -6
- package/src/evaluators/tool-trace.test.ts +46 -0
- package/src/evaluators/tool-trace.ts +50 -0
- package/src/evaluators/types.ts +20 -0
- package/src/index.ts +6 -0
- package/src/runner/executor.test.ts +340 -0
- package/src/runner/executor.ts +289 -20
- package/src/runner/release-validation.test.ts +169 -0
- package/src/runner/runner.ts +5 -1
- package/src/runner/types.ts +4 -0
- package/src/scenario/schema.ts +66 -1
- package/src/storage/supabase.test.ts +1052 -0
- package/src/storage/supabase.ts +614 -5
- package/src/storage/types.ts +207 -1
- package/src/tools/fixture-executor.test.ts +88 -0
- package/src/tools/fixture-executor.ts +112 -0
- package/src/tools/index.ts +3 -0
- package/src/tools/schema-validator.test.ts +32 -0
- package/src/tools/schema-validator.ts +56 -0
- package/src/tools/types.ts +80 -0
package/CHANGELOG.md
CHANGED
|
@@ -1,5 +1,142 @@
|
|
|
1
1
|
# @artemiskit/core
|
|
2
2
|
|
|
3
|
+
## 0.4.0
|
|
4
|
+
|
|
5
|
+
### Minor Changes
|
|
6
|
+
|
|
7
|
+
- Introduce the v1.1 evaluation-integrity contract for assurance runs. Case results now distinguish
|
|
8
|
+
valid passes and failures from invalid evaluator measurements and target execution errors. Strict
|
|
9
|
+
LLM grading is available for assurance scenarios, results retain bounded evaluator evidence, and
|
|
10
|
+
run/report/CLI/JUnit summaries disclose attempts, valid and invalid measurements, and their rate
|
|
11
|
+
denominator. Supabase storage gains a migration for queryable integrity counts and case evidence.
|
|
12
|
+
|
|
13
|
+
## 0.3.0
|
|
14
|
+
|
|
15
|
+
### Minor Changes
|
|
16
|
+
|
|
17
|
+
- ## v0.3.0 - SDK, Guardian Mode & OWASP Compliance
|
|
18
|
+
|
|
19
|
+
This major release delivers the full programmatic SDK, runtime protection with Guardian Mode, OWASP LLM Top 10 2025 attack vectors, and agentic framework adapters.
|
|
20
|
+
|
|
21
|
+
### Programmatic SDK (`@artemiskit/sdk`)
|
|
22
|
+
|
|
23
|
+
The new SDK package provides a complete programmatic API for LLM evaluation:
|
|
24
|
+
|
|
25
|
+
- **ArtemisKit class** with `run()`, `redteam()`, and `stress()` methods
|
|
26
|
+
- **Jest integration** with custom matchers (`toPassAllCases`, `toHaveSuccessRate`, etc.)
|
|
27
|
+
- **Vitest integration** with identical matchers
|
|
28
|
+
- **Event handling** for real-time progress updates
|
|
29
|
+
- **13 custom matchers** for run, red team, and stress test assertions
|
|
30
|
+
|
|
31
|
+
```typescript
|
|
32
|
+
import { ArtemisKit } from "@artemiskit/sdk";
|
|
33
|
+
import { jestMatchers } from "@artemiskit/sdk/jest";
|
|
34
|
+
|
|
35
|
+
expect.extend(jestMatchers);
|
|
36
|
+
|
|
37
|
+
const kit = new ArtemisKit({ provider: "openai", model: "gpt-4o" });
|
|
38
|
+
const results = await kit.run({ scenario: "./tests.yaml" });
|
|
39
|
+
expect(results).toPassAllCases();
|
|
40
|
+
```
|
|
41
|
+
|
|
42
|
+
### Guardian Mode (Runtime Protection)
|
|
43
|
+
|
|
44
|
+
New Guardian Mode provides runtime protection for AI/LLM applications:
|
|
45
|
+
|
|
46
|
+
- **Three operating modes**: `testing`, `guardian`, `hybrid`
|
|
47
|
+
- **Prompt injection detection** and blocking
|
|
48
|
+
- **PII detection & redaction** (email, SSN, phone, API keys)
|
|
49
|
+
- **Action validation** for agent tool/function calls
|
|
50
|
+
- **Intent classification** with risk assessment
|
|
51
|
+
- **Circuit breaker** for automatic blocking on repeated violations
|
|
52
|
+
- **Rate limiting** and **cost limiting**
|
|
53
|
+
- **Custom policies** via TypeScript or YAML
|
|
54
|
+
|
|
55
|
+
```typescript
|
|
56
|
+
import { createGuardian } from "@artemiskit/sdk/guardian";
|
|
57
|
+
|
|
58
|
+
const guardian = createGuardian({ mode: "guardian", blockOnFailure: true });
|
|
59
|
+
const protectedClient = guardian.protect(myLLMClient);
|
|
60
|
+
```
|
|
61
|
+
|
|
62
|
+
### OWASP LLM Top 10 2025 Attack Vectors
|
|
63
|
+
|
|
64
|
+
New red team mutations aligned with OWASP LLM Top 10 2025:
|
|
65
|
+
|
|
66
|
+
| Mutation | OWASP | Description |
|
|
67
|
+
| -------------------- | ----- | ------------------------------ |
|
|
68
|
+
| `bad-likert-judge` | LLM01 | Exploit evaluation capability |
|
|
69
|
+
| `crescendo` | LLM01 | Multi-turn gradual escalation |
|
|
70
|
+
| `deceptive-delight` | LLM01 | Positive framing bypass |
|
|
71
|
+
| `system-extraction` | LLM07 | System prompt leakage |
|
|
72
|
+
| `output-injection` | LLM05 | XSS, SQLi in output |
|
|
73
|
+
| `excessive-agency` | LLM06 | Unauthorized action claims |
|
|
74
|
+
| `hallucination-trap` | LLM09 | Confident fabrication triggers |
|
|
75
|
+
|
|
76
|
+
```bash
|
|
77
|
+
akit redteam scenario.yaml --owasp LLM01,LLM05
|
|
78
|
+
akit redteam scenario.yaml --owasp-full
|
|
79
|
+
```
|
|
80
|
+
|
|
81
|
+
### Agentic Framework Adapters
|
|
82
|
+
|
|
83
|
+
New adapters for testing agentic AI systems:
|
|
84
|
+
|
|
85
|
+
**LangChain Adapter** (`@artemiskit/adapter-langchain`)
|
|
86
|
+
|
|
87
|
+
- Test chains, agents, and runnables
|
|
88
|
+
- Capture intermediate steps and tool usage
|
|
89
|
+
- Support for LCEL, ReAct agents, RAG chains
|
|
90
|
+
|
|
91
|
+
**DeepAgents Adapter** (`@artemiskit/adapter-deepagents`)
|
|
92
|
+
|
|
93
|
+
- Test multi-agent systems and workflows
|
|
94
|
+
- Capture agent traces and inter-agent messages
|
|
95
|
+
- Support for sequential, parallel, and hierarchical workflows
|
|
96
|
+
|
|
97
|
+
```typescript
|
|
98
|
+
import { createLangChainAdapter } from "@artemiskit/adapter-langchain";
|
|
99
|
+
import { createDeepAgentsAdapter } from "@artemiskit/adapter-deepagents";
|
|
100
|
+
|
|
101
|
+
const adapter = createLangChainAdapter(myChain, {
|
|
102
|
+
captureIntermediateSteps: true,
|
|
103
|
+
});
|
|
104
|
+
const result = await adapter.generate({ prompt: "Test query" });
|
|
105
|
+
```
|
|
106
|
+
|
|
107
|
+
### Supabase Storage Enhancements
|
|
108
|
+
|
|
109
|
+
Enhanced cloud storage capabilities:
|
|
110
|
+
|
|
111
|
+
- **Analytics tables** for metrics tracking
|
|
112
|
+
- **Case results table** for granular analysis
|
|
113
|
+
- **Baseline management** for regression detection
|
|
114
|
+
- **Trend analysis** queries
|
|
115
|
+
|
|
116
|
+
### Bug Fixes
|
|
117
|
+
|
|
118
|
+
- **adapter-openai**: Use `max_completion_tokens` for newer OpenAI models (o1, o3, gpt-4.5)
|
|
119
|
+
- **redteam**: Resolve TypeScript and flaky test issues in OWASP mutations
|
|
120
|
+
- **adapters**: Fix TypeScript build errors for agentic adapters
|
|
121
|
+
- **core**: Add `langchain` and `deepagents` to ProviderType union
|
|
122
|
+
|
|
123
|
+
### Examples
|
|
124
|
+
|
|
125
|
+
New comprehensive examples organized by feature:
|
|
126
|
+
|
|
127
|
+
- `examples/guardian/` - Guardian Mode examples (testing, guardian, hybrid modes)
|
|
128
|
+
- `examples/sdk/` - SDK usage examples (Jest, Vitest, events)
|
|
129
|
+
- `examples/adapters/` - Agentic adapter examples
|
|
130
|
+
- `examples/owasp/` - OWASP LLM Top 10 test scenarios
|
|
131
|
+
|
|
132
|
+
### Documentation
|
|
133
|
+
|
|
134
|
+
- Complete SDK documentation with API reference
|
|
135
|
+
- Guardian Mode guide with all three modes explained
|
|
136
|
+
- Agentic adapters documentation (LangChain, DeepAgents)
|
|
137
|
+
- Test matchers reference for Jest/Vitest
|
|
138
|
+
- OWASP LLM Top 10 testing scenarios
|
|
139
|
+
|
|
3
140
|
## 0.2.4
|
|
4
141
|
|
|
5
142
|
### Patch Changes
|
package/README.md
CHANGED
|
@@ -72,11 +72,15 @@ Built-in expectation types:
|
|
|
72
72
|
| Type | Description |
|
|
73
73
|
|------|-------------|
|
|
74
74
|
| `contains` | Check if response contains text |
|
|
75
|
+
| `not_contains` | Check if response does NOT contain text |
|
|
75
76
|
| `exact` | Exact string match |
|
|
76
77
|
| `regex` | Regular expression match |
|
|
77
78
|
| `fuzzy` | Fuzzy string similarity (Levenshtein) |
|
|
79
|
+
| `similarity` | Semantic similarity matching (embedding or LLM-based) |
|
|
78
80
|
| `llm_grader` | LLM-based response grading |
|
|
79
81
|
| `json_schema` | Validate JSON structure |
|
|
82
|
+
| `combined` | AND/OR logic between multiple assertions |
|
|
83
|
+
| `inline` | Safe expression-based custom matchers |
|
|
80
84
|
|
|
81
85
|
## Redaction
|
|
82
86
|
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"registry.d.ts","sourceRoot":"","sources":["../../src/adapters/registry.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,OAAO,KAAK,EAAE,aAAa,EAAE,WAAW,EAAE,YAAY,EAAE,MAAM,SAAS,CAAC;AAExE,KAAK,cAAc,GAAG,CAAC,MAAM,EAAE,aAAa,KAAK,OAAO,CAAC,WAAW,CAAC,CAAC;AAEtE,UAAU,eAAe;IACvB,QAAQ,EAAE,YAAY,CAAC;IACvB,MAAM,EAAE,MAAM,CAAC;CAChB;AAED;;GAEG;AACH,cAAM,eAAe;IACnB,OAAO,CAAC,QAAQ,CAAgD;IAChE,OAAO,CAAC,oBAAoB,CAAwC;IAEpE;;OAEG;IACH,QAAQ,CAAC,QAAQ,EAAE,YAAY,EAAE,OAAO,EAAE,cAAc,GAAG,IAAI;IAK/D;;OAEG;IACH,eAAe,CAAC,QAAQ,EAAE,YAAY,EAAE,MAAM,EAAE,MAAM,GAAG,IAAI;IAI7D;;OAEG;IACH,GAAG,CAAC,QAAQ,EAAE,YAAY,GAAG,cAAc,GAAG,SAAS;IAIvD;;OAEG;IACH,GAAG,CAAC,QAAQ,EAAE,YAAY,GAAG,OAAO;IAIpC;;OAEG;IACH,aAAa,CAAC,QAAQ,EAAE,YAAY,GAAG,OAAO;IAI9C;;OAEG;IACH,oBAAoB,CAAC,QAAQ,EAAE,YAAY,GAAG,MAAM,GAAG,SAAS;IAIhE;;OAEG;IACH,IAAI,IAAI,YAAY,EAAE;IAItB;;OAEG;IACH,eAAe,IAAI,eAAe,EAAE;CAMrC;AAED,eAAO,MAAM,eAAe,iBAAwB,CAAC;AAgBrD;;;GAGG;AACH,wBAAsB,uBAAuB,IAAI,OAAO,CAAC,IAAI,CAAC,
|
|
1
|
+
{"version":3,"file":"registry.d.ts","sourceRoot":"","sources":["../../src/adapters/registry.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,OAAO,KAAK,EAAE,aAAa,EAAE,WAAW,EAAE,YAAY,EAAE,MAAM,SAAS,CAAC;AAExE,KAAK,cAAc,GAAG,CAAC,MAAM,EAAE,aAAa,KAAK,OAAO,CAAC,WAAW,CAAC,CAAC;AAEtE,UAAU,eAAe;IACvB,QAAQ,EAAE,YAAY,CAAC;IACvB,MAAM,EAAE,MAAM,CAAC;CAChB;AAED;;GAEG;AACH,cAAM,eAAe;IACnB,OAAO,CAAC,QAAQ,CAAgD;IAChE,OAAO,CAAC,oBAAoB,CAAwC;IAEpE;;OAEG;IACH,QAAQ,CAAC,QAAQ,EAAE,YAAY,EAAE,OAAO,EAAE,cAAc,GAAG,IAAI;IAK/D;;OAEG;IACH,eAAe,CAAC,QAAQ,EAAE,YAAY,EAAE,MAAM,EAAE,MAAM,GAAG,IAAI;IAI7D;;OAEG;IACH,GAAG,CAAC,QAAQ,EAAE,YAAY,GAAG,cAAc,GAAG,SAAS;IAIvD;;OAEG;IACH,GAAG,CAAC,QAAQ,EAAE,YAAY,GAAG,OAAO;IAIpC;;OAEG;IACH,aAAa,CAAC,QAAQ,EAAE,YAAY,GAAG,OAAO;IAI9C;;OAEG;IACH,oBAAoB,CAAC,QAAQ,EAAE,YAAY,GAAG,MAAM,GAAG,SAAS;IAIhE;;OAEG;IACH,IAAI,IAAI,YAAY,EAAE;IAItB;;OAEG;IACH,eAAe,IAAI,eAAe,EAAE;CAMrC;AAED,eAAO,MAAM,eAAe,iBAAwB,CAAC;AAgBrD;;;GAGG;AACH,wBAAsB,uBAAuB,IAAI,OAAO,CAAC,IAAI,CAAC,CAsF7D"}
|
package/dist/adapters/types.d.ts
CHANGED
|
@@ -8,6 +8,8 @@ export interface ChatMessage {
|
|
|
8
8
|
role: 'system' | 'user' | 'assistant' | 'function' | 'tool';
|
|
9
9
|
content: string;
|
|
10
10
|
name?: string;
|
|
11
|
+
/** The provider tool-call ID this tool result answers. */
|
|
12
|
+
toolCallId?: string;
|
|
11
13
|
function_call?: {
|
|
12
14
|
name: string;
|
|
13
15
|
arguments: string;
|
|
@@ -39,6 +41,16 @@ export interface GenerateOptions {
|
|
|
39
41
|
type: 'text' | 'json_object';
|
|
40
42
|
};
|
|
41
43
|
metadata?: Record<string, unknown>;
|
|
44
|
+
providerOptions?: {
|
|
45
|
+
ling?: LingRequestOptions;
|
|
46
|
+
};
|
|
47
|
+
}
|
|
48
|
+
export interface LingRequestOptions {
|
|
49
|
+
thinking?: {
|
|
50
|
+
type: 'enabled' | 'disabled';
|
|
51
|
+
};
|
|
52
|
+
enableSearch?: boolean;
|
|
53
|
+
searchOptions?: Record<string, unknown>;
|
|
42
54
|
}
|
|
43
55
|
/**
|
|
44
56
|
* Function/Tool definitions for function calling
|
|
@@ -102,7 +114,7 @@ export interface ModelClient {
|
|
|
102
114
|
/**
|
|
103
115
|
* Provider types - all supported providers
|
|
104
116
|
*/
|
|
105
|
-
export type ProviderType = 'openai' | 'azure-openai' | 'vercel-ai' | 'anthropic' | 'google' | 'mistral' | 'cohere' | 'huggingface' | 'ollama' | 'custom';
|
|
117
|
+
export type ProviderType = 'openai' | 'azure-openai' | 'vercel-ai' | 'anthropic' | 'langchain' | 'deepagents' | 'ling' | 'google' | 'mistral' | 'cohere' | 'huggingface' | 'ollama' | 'langchain' | 'deepagents' | 'custom';
|
|
106
118
|
/**
|
|
107
119
|
* Base adapter configuration
|
|
108
120
|
*/
|
|
@@ -151,8 +163,44 @@ export interface VercelAIAdapterConfig extends BaseAdapterConfig {
|
|
|
151
163
|
export interface AnthropicAdapterConfig extends BaseAdapterConfig {
|
|
152
164
|
provider: 'anthropic';
|
|
153
165
|
}
|
|
166
|
+
export interface LingAdapterConfig extends BaseAdapterConfig {
|
|
167
|
+
provider: 'ling';
|
|
168
|
+
thinking?: LingRequestOptions['thinking'];
|
|
169
|
+
enableSearch?: boolean;
|
|
170
|
+
searchOptions?: Record<string, unknown>;
|
|
171
|
+
}
|
|
172
|
+
/**
|
|
173
|
+
* LangChain adapter configuration
|
|
174
|
+
*/
|
|
175
|
+
export interface LangChainAdapterConfig extends BaseAdapterConfig {
|
|
176
|
+
provider: 'langchain';
|
|
177
|
+
/** Type of LangChain runnable */
|
|
178
|
+
runnableType?: 'chain' | 'agent' | 'llm' | 'runnable';
|
|
179
|
+
/** Name identifier for the chain/agent */
|
|
180
|
+
name?: string;
|
|
181
|
+
/** Capture intermediate steps from agents */
|
|
182
|
+
captureIntermediateSteps?: boolean;
|
|
183
|
+
/** Custom input key */
|
|
184
|
+
inputKey?: string;
|
|
185
|
+
/** Custom output key */
|
|
186
|
+
outputKey?: string;
|
|
187
|
+
}
|
|
188
|
+
/**
|
|
189
|
+
* DeepAgents adapter configuration
|
|
190
|
+
*/
|
|
191
|
+
export interface DeepAgentsAdapterConfig extends BaseAdapterConfig {
|
|
192
|
+
provider: 'deepagents';
|
|
193
|
+
/** Name identifier for the agent system */
|
|
194
|
+
name?: string;
|
|
195
|
+
/** Capture agent execution traces */
|
|
196
|
+
captureTraces?: boolean;
|
|
197
|
+
/** Capture inter-agent messages */
|
|
198
|
+
captureMessages?: boolean;
|
|
199
|
+
/** Maximum execution time in milliseconds */
|
|
200
|
+
executionTimeout?: number;
|
|
201
|
+
}
|
|
154
202
|
/**
|
|
155
203
|
* Union type for all adapter configs
|
|
156
204
|
*/
|
|
157
|
-
export type AdapterConfig = OpenAIAdapterConfig | AzureOpenAIAdapterConfig | VercelAIAdapterConfig | AnthropicAdapterConfig | BaseAdapterConfig;
|
|
205
|
+
export type AdapterConfig = OpenAIAdapterConfig | AzureOpenAIAdapterConfig | VercelAIAdapterConfig | AnthropicAdapterConfig | LingAdapterConfig | LangChainAdapterConfig | DeepAgentsAdapterConfig | BaseAdapterConfig;
|
|
158
206
|
//# sourceMappingURL=types.d.ts.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../../src/adapters/types.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH;;GAEG;AACH,MAAM,WAAW,WAAW;IAC1B,IAAI,EAAE,QAAQ,GAAG,MAAM,GAAG,WAAW,GAAG,UAAU,GAAG,MAAM,CAAC;IAC5D,OAAO,EAAE,MAAM,CAAC;IAChB,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,aAAa,CAAC,EAAE;QACd,IAAI,EAAE,MAAM,CAAC;QACb,SAAS,EAAE,MAAM,CAAC;KACnB,CAAC;IACF,UAAU,CAAC,EAAE,QAAQ,EAAE,CAAC;CACzB;AAED,MAAM,WAAW,QAAQ;IACvB,EAAE,EAAE,MAAM,CAAC;IACX,IAAI,EAAE,UAAU,CAAC;IACjB,QAAQ,EAAE;QACR,IAAI,EAAE,MAAM,CAAC;QACb,SAAS,EAAE,MAAM,CAAC;KACnB,CAAC;CACH;AAED;;GAEG;AACH,MAAM,WAAW,eAAe;IAC9B,MAAM,EAAE,MAAM,GAAG,WAAW,EAAE,CAAC;IAC/B,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,IAAI,CAAC,EAAE,MAAM,EAAE,CAAC;IAChB,SAAS,CAAC,EAAE,kBAAkB,EAAE,CAAC;IACjC,KAAK,CAAC,EAAE,cAAc,EAAE,CAAC;IACzB,cAAc,CAAC,EAAE;QAAE,IAAI,EAAE,MAAM,GAAG,aAAa,CAAA;KAAE,CAAC;IAClD,QAAQ,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;
|
|
1
|
+
{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../../src/adapters/types.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH;;GAEG;AACH,MAAM,WAAW,WAAW;IAC1B,IAAI,EAAE,QAAQ,GAAG,MAAM,GAAG,WAAW,GAAG,UAAU,GAAG,MAAM,CAAC;IAC5D,OAAO,EAAE,MAAM,CAAC;IAChB,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,0DAA0D;IAC1D,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,aAAa,CAAC,EAAE;QACd,IAAI,EAAE,MAAM,CAAC;QACb,SAAS,EAAE,MAAM,CAAC;KACnB,CAAC;IACF,UAAU,CAAC,EAAE,QAAQ,EAAE,CAAC;CACzB;AAED,MAAM,WAAW,QAAQ;IACvB,EAAE,EAAE,MAAM,CAAC;IACX,IAAI,EAAE,UAAU,CAAC;IACjB,QAAQ,EAAE;QACR,IAAI,EAAE,MAAM,CAAC;QACb,SAAS,EAAE,MAAM,CAAC;KACnB,CAAC;CACH;AAED;;GAEG;AACH,MAAM,WAAW,eAAe;IAC9B,MAAM,EAAE,MAAM,GAAG,WAAW,EAAE,CAAC;IAC/B,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,IAAI,CAAC,EAAE,MAAM,EAAE,CAAC;IAChB,SAAS,CAAC,EAAE,kBAAkB,EAAE,CAAC;IACjC,KAAK,CAAC,EAAE,cAAc,EAAE,CAAC;IACzB,cAAc,CAAC,EAAE;QAAE,IAAI,EAAE,MAAM,GAAG,aAAa,CAAA;KAAE,CAAC;IAClD,QAAQ,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IACnC,eAAe,CAAC,EAAE;QAAE,IAAI,CAAC,EAAE,kBAAkB,CAAA;KAAE,CAAC;CACjD;AAED,MAAM,WAAW,kBAAkB;IACjC,QAAQ,CAAC,EAAE;QAAE,IAAI,EAAE,SAAS,GAAG,UAAU,CAAA;KAAE,CAAC;IAC5C,YAAY,CAAC,EAAE,OAAO,CAAC;IACvB,aAAa,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CACzC;AAED;;GAEG;AACH,MAAM,WAAW,kBAAkB;IACjC,IAAI,EAAE,MAAM,CAAC;IACb,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,UAAU,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CACrC;AAED,MAAM,WAAW,cAAc;IAC7B,IAAI,EAAE,UAAU,CAAC;IACjB,QAAQ,EAAE,kBAAkB,CAAC;CAC9B;AAED;;GAEG;AACH,MAAM,WAAW,UAAU;IACzB,MAAM,EAAE,MAAM,CAAC;IACf,UAAU,EAAE,MAAM,CAAC;IACnB,KAAK,EAAE,MAAM,CAAC;CACf;AAED;;GAEG;AACH,MAAM,WAAW,cAAc;IAC7B,EAAE,EAAE,MAAM,CAAC;IACX,KAAK,EAAE,MAAM,CAAC;IACd,IAAI,EAAE,MAAM,CAAC;IACb,MAAM,EAAE,UAAU,CAAC;IACnB,SAAS,EAAE,MAAM,CAAC;IAClB,YAAY,CAAC,EAAE,MAAM,GAAG,QAAQ,GAAG,eAAe,GAAG,YAAY,GAAG,gBAAgB,CAAC;IACrF,YAAY,CAAC,EAAE;QACb,IAAI,EAAE,MAAM,CAAC;QACb,SAAS,EAAE,MAAM,CAAC;KACnB,CAAC;IACF,SAAS,CAAC,EAAE,QAAQ,EAAE,CAAC;IACvB,GAAG,CAAC,EAAE,OAAO,CAAC;CACf;AAED;;GAEG;AACH,MAAM,WAAW,iBAAiB;IAChC,SAAS,EAAE,OAAO,CAAC;IACnB,eAAe,EAAE,OAAO,CAAC;IACzB,OAAO,EAAE,OAAO,CAAC;IACjB,UAAU,EAAE,MAAM,CAAC;IACnB,MAAM,CAAC,EAAE,OAAO,CAAC;IACjB,QAAQ,CAAC,EAAE,OAAO,CAAC;CACpB;AAED;;GAEG;AACH,MAAM,WAAW,WAAW;IAC1B,QAAQ,CAAC,QAAQ,EAAE,MAAM,CAAC;IAE1B,QAAQ,CAAC,OAAO,EAAE,eAAe,GAAG,OAAO,CAAC,cAAc,CAAC,CAAC;IAE5D,MAAM,CAAC,CAAC,OAAO,EAAE,eAAe,EAAE,OAAO,EAAE,CAAC,KAAK,EAAE,MAAM,KAAK,IAAI,GAAG,aAAa,CAAC,MAAM,CAAC,CAAC;IAE3F,KAAK,CAAC,CAAC,IAAI,EAAE,MAAM,EAAE,KAAK,CAAC,EAAE,MAAM,GAAG,OAAO,CAAC,MAAM,EAAE,CAAC,CAAC;IAExD,YAAY,IAAI,OAAO,CAAC,iBAAiB,CAAC,CAAC;IAE3C,KAAK,CAAC,IAAI,OAAO,CAAC,IAAI,CAAC,CAAC;CACzB;AAED;;GAEG;AACH,MAAM,MAAM,YAAY,GACpB,QAAQ,GACR,cAAc,GACd,WAAW,GACX,WAAW,GACX,WAAW,GACX,YAAY,GACZ,MAAM,GACN,QAAQ,GACR,SAAS,GACT,QAAQ,GACR,aAAa,GACb,QAAQ,GACR,WAAW,GACX,YAAY,GACZ,QAAQ,CAAC;AAEb;;GAEG;AACH,MAAM,WAAW,iBAAiB;IAChC,QAAQ,EAAE,YAAY,CAAC;IACvB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,UAAU,CAAC,EAAE,MAAM,CAAC;CACrB;AAED;;GAEG;AACH,MAAM,WAAW,mBAAoB,SAAQ,iBAAiB;IAC5D,QAAQ,EAAE,QAAQ,CAAC;IACnB,YAAY,CAAC,EAAE,MAAM,CAAC;CACvB;AAED;;GAEG;AACH,MAAM,WAAW,wBAAyB,SAAQ,iBAAiB;IACjE,QAAQ,EAAE,cAAc,CAAC;IACzB,YAAY,EAAE,MAAM,CAAC;IACrB,cAAc,EAAE,MAAM,CAAC;IACvB,UAAU,EAAE,MAAM,CAAC;IACnB,6DAA6D;IAC7D,uBAAuB,CAAC,EAAE,MAAM,CAAC;IACjC;;;OAGG;IACH,WAAW,CAAC,EAAE,MAAM,CAAC;CACtB;AAED;;GAEG;AACH,MAAM,WAAW,qBAAsB,SAAQ,iBAAiB;IAC9D,QAAQ,EAAE,WAAW,CAAC;IACtB,kBAAkB,EAAE,QAAQ,GAAG,OAAO,GAAG,WAAW,GAAG,QAAQ,GAAG,SAAS,CAAC;IAC5E,cAAc,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CAC1C;AAED;;GAEG;AACH,MAAM,WAAW,sBAAuB,SAAQ,iBAAiB;IAC/D,QAAQ,EAAE,WAAW,CAAC;CACvB;AAED,MAAM,WAAW,iBAAkB,SAAQ,iBAAiB;IAC1D,QAAQ,EAAE,MAAM,CAAC;IACjB,QAAQ,CAAC,EAAE,kBAAkB,CAAC,UAAU,CAAC,CAAC;IAC1C,YAAY,CAAC,EAAE,OAAO,CAAC;IACvB,aAAa,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CACzC;AAED;;GAEG;AACH,MAAM,WAAW,sBAAuB,SAAQ,iBAAiB;IAC/D,QAAQ,EAAE,WAAW,CAAC;IACtB,iCAAiC;IACjC,YAAY,CAAC,EAAE,OAAO,GAAG,OAAO,GAAG,KAAK,GAAG,UAAU,CAAC;IACtD,0CAA0C;IAC1C,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,6CAA6C;IAC7C,wBAAwB,CAAC,EAAE,OAAO,CAAC;IACnC,uBAAuB;IACvB,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,wBAAwB;IACxB,SAAS,CAAC,EAAE,MAAM,CAAC;CACpB;AAED;;GAEG;AACH,MAAM,WAAW,uBAAwB,SAAQ,iBAAiB;IAChE,QAAQ,EAAE,YAAY,CAAC;IACvB,2CAA2C;IAC3C,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,qCAAqC;IACrC,aAAa,CAAC,EAAE,OAAO,CAAC;IACxB,mCAAmC;IACnC,eAAe,CAAC,EAAE,OAAO,CAAC;IAC1B,6CAA6C;IAC7C,gBAAgB,CAAC,EAAE,MAAM,CAAC;CAC3B;AAED;;GAEG;AACH,MAAM,MAAM,aAAa,GACrB,mBAAmB,GACnB,wBAAwB,GACxB,qBAAqB,GACrB,sBAAsB,GACtB,iBAAiB,GACjB,sBAAsB,GACtB,uBAAuB,GACvB,iBAAiB,CAAC"}
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/agent-evaluation/index.ts"],"names":[],"mappings":"AAAA,cAAc,SAAS,CAAC;AACxB,cAAc,UAAU,CAAC"}
|
|
@@ -0,0 +1,35 @@
|
|
|
1
|
+
import { type AgentOutcome, type AgentTask } from './types';
|
|
2
|
+
export type AgentTerminationStatus = 'completed' | 'agent_error' | 'tool_error' | 'infrastructure_error' | 'timed_out';
|
|
3
|
+
export interface AgentTerminationEvidence {
|
|
4
|
+
status: AgentTerminationStatus;
|
|
5
|
+
}
|
|
6
|
+
export interface AgentAcceptanceCheck {
|
|
7
|
+
command: string;
|
|
8
|
+
status: 'passed' | 'failed' | 'executor_error';
|
|
9
|
+
exitCode: number | null;
|
|
10
|
+
durationMs: number;
|
|
11
|
+
}
|
|
12
|
+
export interface AgentArtifactCheck {
|
|
13
|
+
id: string;
|
|
14
|
+
status: 'passed' | 'failed' | 'executor_error';
|
|
15
|
+
durationMs: number;
|
|
16
|
+
}
|
|
17
|
+
export interface AgentEvaluationEvidence {
|
|
18
|
+
termination: AgentTerminationEvidence;
|
|
19
|
+
acceptanceChecks: AgentAcceptanceCheck[];
|
|
20
|
+
artifactChecks?: AgentArtifactCheck[];
|
|
21
|
+
}
|
|
22
|
+
export type AgentEvaluationVerdict = 'passed' | 'passed_with_recovery' | 'task_failed' | 'infrastructure_failed';
|
|
23
|
+
export interface AgentEvaluationIssue {
|
|
24
|
+
code: string;
|
|
25
|
+
message: string;
|
|
26
|
+
}
|
|
27
|
+
export interface AgentEvaluationScore {
|
|
28
|
+
taskId: string;
|
|
29
|
+
verdict: AgentEvaluationVerdict;
|
|
30
|
+
passed: boolean;
|
|
31
|
+
recoveredActionCount: number;
|
|
32
|
+
issues: AgentEvaluationIssue[];
|
|
33
|
+
}
|
|
34
|
+
export declare function scoreAgentOutcome(task: AgentTask, outcome: AgentOutcome, evidence: AgentEvaluationEvidence): AgentEvaluationScore;
|
|
35
|
+
//# sourceMappingURL=scorer.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"scorer.d.ts","sourceRoot":"","sources":["../../src/agent-evaluation/scorer.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,KAAK,YAAY,EAAE,KAAK,SAAS,EAAwB,MAAM,SAAS,CAAC;AAElF,MAAM,MAAM,sBAAsB,GAC9B,WAAW,GACX,aAAa,GACb,YAAY,GACZ,sBAAsB,GACtB,WAAW,CAAC;AAEhB,MAAM,WAAW,wBAAwB;IACvC,MAAM,EAAE,sBAAsB,CAAC;CAChC;AAED,MAAM,WAAW,oBAAoB;IACnC,OAAO,EAAE,MAAM,CAAC;IAChB,MAAM,EAAE,QAAQ,GAAG,QAAQ,GAAG,gBAAgB,CAAC;IAC/C,QAAQ,EAAE,MAAM,GAAG,IAAI,CAAC;IACxB,UAAU,EAAE,MAAM,CAAC;CACpB;AAED,MAAM,WAAW,kBAAkB;IACjC,EAAE,EAAE,MAAM,CAAC;IACX,MAAM,EAAE,QAAQ,GAAG,QAAQ,GAAG,gBAAgB,CAAC;IAC/C,UAAU,EAAE,MAAM,CAAC;CACpB;AAED,MAAM,WAAW,uBAAuB;IACtC,WAAW,EAAE,wBAAwB,CAAC;IACtC,gBAAgB,EAAE,oBAAoB,EAAE,CAAC;IACzC,cAAc,CAAC,EAAE,kBAAkB,EAAE,CAAC;CACvC;AAED,MAAM,MAAM,sBAAsB,GAC9B,QAAQ,GACR,sBAAsB,GACtB,aAAa,GACb,uBAAuB,CAAC;AAE5B,MAAM,WAAW,oBAAoB;IACnC,IAAI,EAAE,MAAM,CAAC;IACb,OAAO,EAAE,MAAM,CAAC;CACjB;AAED,MAAM,WAAW,oBAAoB;IACnC,MAAM,EAAE,MAAM,CAAC;IACf,OAAO,EAAE,sBAAsB,CAAC;IAChC,MAAM,EAAE,OAAO,CAAC;IAChB,oBAAoB,EAAE,MAAM,CAAC;IAC7B,MAAM,EAAE,oBAAoB,EAAE,CAAC;CAChC;AAsQD,wBAAgB,iBAAiB,CAC/B,IAAI,EAAE,SAAS,EACf,OAAO,EAAE,YAAY,EACrB,QAAQ,EAAE,uBAAuB,GAChC,oBAAoB,CAoUtB"}
|
|
@@ -0,0 +1,37 @@
|
|
|
1
|
+
export interface AgentTask {
|
|
2
|
+
id: string;
|
|
3
|
+
fixturePath: string;
|
|
4
|
+
allowedPaths: string[];
|
|
5
|
+
allowedTools: string[];
|
|
6
|
+
maxActions: number;
|
|
7
|
+
timeoutMs: number;
|
|
8
|
+
acceptanceCommands: string[];
|
|
9
|
+
requiredArtifactChecks?: string[];
|
|
10
|
+
}
|
|
11
|
+
export interface AgentAction {
|
|
12
|
+
type: 'tool' | 'command' | 'file';
|
|
13
|
+
name: string;
|
|
14
|
+
status: 'success' | 'error' | 'rejected';
|
|
15
|
+
durationMs: number;
|
|
16
|
+
summary?: string;
|
|
17
|
+
}
|
|
18
|
+
export interface AgentTrace {
|
|
19
|
+
taskId: string;
|
|
20
|
+
actions: AgentAction[];
|
|
21
|
+
changedPaths: string[];
|
|
22
|
+
startedAt: string;
|
|
23
|
+
completedAt: string;
|
|
24
|
+
}
|
|
25
|
+
export interface AgentOutcome {
|
|
26
|
+
taskId: string;
|
|
27
|
+
completed: boolean;
|
|
28
|
+
acceptancePassed: boolean;
|
|
29
|
+
trace: AgentTrace;
|
|
30
|
+
finalDiff?: string;
|
|
31
|
+
error?: string;
|
|
32
|
+
}
|
|
33
|
+
export interface AgentHarness {
|
|
34
|
+
run(task: AgentTask): Promise<AgentOutcome>;
|
|
35
|
+
}
|
|
36
|
+
export declare function actionBudgetExceeded(task: AgentTask, trace: AgentTrace): boolean;
|
|
37
|
+
//# sourceMappingURL=types.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../../src/agent-evaluation/types.ts"],"names":[],"mappings":"AAAA,MAAM,WAAW,SAAS;IACxB,EAAE,EAAE,MAAM,CAAC;IACX,WAAW,EAAE,MAAM,CAAC;IACpB,YAAY,EAAE,MAAM,EAAE,CAAC;IACvB,YAAY,EAAE,MAAM,EAAE,CAAC;IACvB,UAAU,EAAE,MAAM,CAAC;IACnB,SAAS,EAAE,MAAM,CAAC;IAClB,kBAAkB,EAAE,MAAM,EAAE,CAAC;IAC7B,sBAAsB,CAAC,EAAE,MAAM,EAAE,CAAC;CACnC;AAED,MAAM,WAAW,WAAW;IAC1B,IAAI,EAAE,MAAM,GAAG,SAAS,GAAG,MAAM,CAAC;IAClC,IAAI,EAAE,MAAM,CAAC;IACb,MAAM,EAAE,SAAS,GAAG,OAAO,GAAG,UAAU,CAAC;IACzC,UAAU,EAAE,MAAM,CAAC;IACnB,OAAO,CAAC,EAAE,MAAM,CAAC;CAClB;AAED,MAAM,WAAW,UAAU;IACzB,MAAM,EAAE,MAAM,CAAC;IACf,OAAO,EAAE,WAAW,EAAE,CAAC;IACvB,YAAY,EAAE,MAAM,EAAE,CAAC;IACvB,SAAS,EAAE,MAAM,CAAC;IAClB,WAAW,EAAE,MAAM,CAAC;CACrB;AAED,MAAM,WAAW,YAAY;IAC3B,MAAM,EAAE,MAAM,CAAC;IACf,SAAS,EAAE,OAAO,CAAC;IACnB,gBAAgB,EAAE,OAAO,CAAC;IAC1B,KAAK,EAAE,UAAU,CAAC;IAClB,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,KAAK,CAAC,EAAE,MAAM,CAAC;CAChB;AAED,MAAM,WAAW,YAAY;IAC3B,GAAG,CAAC,IAAI,EAAE,SAAS,GAAG,OAAO,CAAC,YAAY,CAAC,CAAC;CAC7C;AAED,wBAAgB,oBAAoB,CAAC,IAAI,EAAE,SAAS,EAAE,KAAK,EAAE,UAAU,GAAG,OAAO,CAEhF"}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"manifest.d.ts","sourceRoot":"","sources":["../../src/artifacts/manifest.ts"],"names":[],"mappings":"AAAA;;GAEG;AAMH,OAAO,KAAK,EACV,UAAU,EAEV,qBAAqB,EACrB,cAAc,EACd,SAAS,EACT,WAAW,EAEZ,MAAM,SAAS,CAAC;
|
|
1
|
+
{"version":3,"file":"manifest.d.ts","sourceRoot":"","sources":["../../src/artifacts/manifest.ts"],"names":[],"mappings":"AAAA;;GAEG;AAMH,OAAO,KAAK,EACV,UAAU,EAEV,qBAAqB,EACrB,cAAc,EACd,SAAS,EACT,WAAW,EAEZ,MAAM,SAAS,CAAC;AAGjB;;GAEG;AACH,wBAAgB,iBAAiB,CAAC,OAAO,EAAE;IACzC,OAAO,EAAE,MAAM,CAAC;IAChB,MAAM,EAAE,SAAS,CAAC;IAClB,cAAc,CAAC,EAAE,cAAc,CAAC;IAChC,KAAK,EAAE,UAAU,EAAE,CAAC;IACpB,SAAS,EAAE,IAAI,CAAC;IAChB,OAAO,EAAE,IAAI,CAAC;IACd,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,SAAS,CAAC,EAAE,qBAAqB,CAAC;CACnC,GAAG,WAAW,CAuCd"}
|
|
@@ -31,6 +31,24 @@ export interface ManifestRedactionInfo {
|
|
|
31
31
|
totalRedactions: number;
|
|
32
32
|
};
|
|
33
33
|
}
|
|
34
|
+
/**
|
|
35
|
+
* Terminal status of a case measurement.
|
|
36
|
+
*
|
|
37
|
+
* `invalid` means a target response was available but could not be evaluated
|
|
38
|
+
* reliably. `error` means execution did not produce a usable target response.
|
|
39
|
+
*/
|
|
40
|
+
export type CaseEvaluationStatus = 'passed' | 'failed' | 'invalid' | 'error';
|
|
41
|
+
/** Reviewed, bounded evaluator evidence retained in a run artifact. */
|
|
42
|
+
export interface CaseEvaluationEvidence {
|
|
43
|
+
evaluator: string;
|
|
44
|
+
score?: number;
|
|
45
|
+
threshold?: number;
|
|
46
|
+
model?: string;
|
|
47
|
+
validation?: {
|
|
48
|
+
status: 'valid' | 'invalid';
|
|
49
|
+
code?: string;
|
|
50
|
+
};
|
|
51
|
+
}
|
|
34
52
|
/**
|
|
35
53
|
* Individual test case result
|
|
36
54
|
*/
|
|
@@ -38,6 +56,10 @@ export interface CaseResult {
|
|
|
38
56
|
id: string;
|
|
39
57
|
name?: string;
|
|
40
58
|
ok: boolean;
|
|
59
|
+
/** Present in manifest v1.1+. Missing values use the documented legacy mapping. */
|
|
60
|
+
status?: CaseEvaluationStatus;
|
|
61
|
+
/** Number of execution attempts represented by this terminal result. */
|
|
62
|
+
attempts?: number;
|
|
41
63
|
score: number;
|
|
42
64
|
matcherType: string;
|
|
43
65
|
reason?: string;
|
|
@@ -52,8 +74,14 @@ export interface CaseResult {
|
|
|
52
74
|
expected: object;
|
|
53
75
|
tags: string[];
|
|
54
76
|
error?: string;
|
|
77
|
+
/** Sanitized evaluator evidence; arbitrary evaluator details are never stored here. */
|
|
78
|
+
evidence?: CaseEvaluationEvidence;
|
|
55
79
|
/** Redaction information for this case */
|
|
56
80
|
redaction?: CaseRedactionInfo;
|
|
81
|
+
/** Ordered tool activity captured for an enabled tool loop. */
|
|
82
|
+
toolTrace?: import('../tools').ToolTraceEntry[];
|
|
83
|
+
/** Terminal status for an enabled tool loop. */
|
|
84
|
+
toolLoop?: import('../tools').ToolLoopSummary;
|
|
57
85
|
}
|
|
58
86
|
/**
|
|
59
87
|
* Cost estimation details
|
|
@@ -78,7 +106,15 @@ export interface CostEstimateInfo {
|
|
|
78
106
|
*/
|
|
79
107
|
export interface RunMetrics {
|
|
80
108
|
success_rate: number;
|
|
109
|
+
/** Number of execution attempts, including retries. */
|
|
110
|
+
total_attempts?: number;
|
|
81
111
|
total_cases: number;
|
|
112
|
+
/** Case results with a valid evaluator outcome (passed or failed). */
|
|
113
|
+
valid_evaluations?: number;
|
|
114
|
+
/** Case results excluded from outcome rates (invalid or error). */
|
|
115
|
+
invalid_evaluations?: number;
|
|
116
|
+
/** Denominator used for success_rate; zero produces a success_rate of zero. */
|
|
117
|
+
outcome_rate_denominator?: number;
|
|
82
118
|
passed_cases: number;
|
|
83
119
|
failed_cases: number;
|
|
84
120
|
median_latency_ms: number;
|
|
@@ -135,6 +171,14 @@ export interface ResolvedConfig {
|
|
|
135
171
|
api_version?: string;
|
|
136
172
|
/** Underlying provider for Vercel AI SDK */
|
|
137
173
|
underlying_provider?: string;
|
|
174
|
+
/** Name identifier for the chain/agent */
|
|
175
|
+
name?: string;
|
|
176
|
+
/** Type of LangChain runnable */
|
|
177
|
+
runnable_type?: string;
|
|
178
|
+
/** Capture agent execution traces */
|
|
179
|
+
capture_traces?: boolean;
|
|
180
|
+
/** Capture inter-agent messages */
|
|
181
|
+
capture_messages?: boolean;
|
|
138
182
|
/** Request timeout in ms */
|
|
139
183
|
timeout?: number;
|
|
140
184
|
/** Max retries */
|
|
@@ -153,6 +197,10 @@ export interface ResolvedConfig {
|
|
|
153
197
|
deployment_name?: ConfigSource;
|
|
154
198
|
api_version?: ConfigSource;
|
|
155
199
|
underlying_provider?: ConfigSource;
|
|
200
|
+
name?: ConfigSource;
|
|
201
|
+
runnable_type?: ConfigSource;
|
|
202
|
+
capture_traces?: ConfigSource;
|
|
203
|
+
capture_messages?: ConfigSource;
|
|
156
204
|
timeout?: ConfigSource;
|
|
157
205
|
max_retries?: ConfigSource;
|
|
158
206
|
temperature?: ConfigSource;
|
|
@@ -195,6 +243,12 @@ export interface RunManifest {
|
|
|
195
243
|
/** Redaction information for this run */
|
|
196
244
|
redaction?: ManifestRedactionInfo;
|
|
197
245
|
}
|
|
246
|
+
/**
|
|
247
|
+
* Read a case status from both v1.1 artifacts and historical v1.0 artifacts.
|
|
248
|
+
* Historical records cannot distinguish evaluator failures from ordinary failed
|
|
249
|
+
* criteria unless they set the legacy `error` field.
|
|
250
|
+
*/
|
|
251
|
+
export declare function getCaseEvaluationStatus(caseResult: CaseResult): CaseEvaluationStatus;
|
|
198
252
|
/**
|
|
199
253
|
* Red team test result status
|
|
200
254
|
*/
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../../src/artifacts/types.ts"],"names":[],"mappings":"AAAA;;GAEG;AAMH;;GAEG;AACH,MAAM,WAAW,iBAAiB;IAChC,8CAA8C;IAC9C,QAAQ,EAAE,OAAO,CAAC;IAClB,kCAAkC;IAClC,cAAc,EAAE,OAAO,CAAC;IACxB,oCAAoC;IACpC,gBAAgB,EAAE,OAAO,CAAC;IAC1B,wCAAwC;IACxC,cAAc,EAAE,MAAM,CAAC;CACxB;AAED;;GAEG;AACH,MAAM,WAAW,qBAAqB;IACpC,oCAAoC;IACpC,OAAO,EAAE,OAAO,CAAC;IACjB,yDAAyD;IACzD,YAAY,EAAE,MAAM,EAAE,CAAC;IACvB,8BAA8B;IAC9B,WAAW,EAAE,MAAM,CAAC;IACpB,4BAA4B;IAC5B,OAAO,EAAE;QACP,eAAe,EAAE,MAAM,CAAC;QACxB,iBAAiB,EAAE,MAAM,CAAC;QAC1B,eAAe,EAAE,MAAM,CAAC;KACzB,CAAC;CACH;AAMD;;GAEG;AACH,MAAM,WAAW,UAAU;IACzB,EAAE,EAAE,MAAM,CAAC;IACX,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,EAAE,EAAE,OAAO,CAAC;IACZ,KAAK,EAAE,MAAM,CAAC;IACd,WAAW,EAAE,MAAM,CAAC;IACpB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,SAAS,EAAE,MAAM,CAAC;IAClB,MAAM,EAAE;QACN,MAAM,EAAE,MAAM,CAAC;QACf,UAAU,EAAE,MAAM,CAAC;QACnB,KAAK,EAAE,MAAM,CAAC;KACf,CAAC;IACF,MAAM,EAAE,MAAM,GAAG,MAAM,CAAC;IACxB,QAAQ,EAAE,MAAM,CAAC;IACjB,QAAQ,EAAE,MAAM,CAAC;IACjB,IAAI,EAAE,MAAM,EAAE,CAAC;IACf,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,0CAA0C;IAC1C,SAAS,CAAC,EAAE,iBAAiB,CAAC;CAC/
|
|
1
|
+
{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../../src/artifacts/types.ts"],"names":[],"mappings":"AAAA;;GAEG;AAMH;;GAEG;AACH,MAAM,WAAW,iBAAiB;IAChC,8CAA8C;IAC9C,QAAQ,EAAE,OAAO,CAAC;IAClB,kCAAkC;IAClC,cAAc,EAAE,OAAO,CAAC;IACxB,oCAAoC;IACpC,gBAAgB,EAAE,OAAO,CAAC;IAC1B,wCAAwC;IACxC,cAAc,EAAE,MAAM,CAAC;CACxB;AAED;;GAEG;AACH,MAAM,WAAW,qBAAqB;IACpC,oCAAoC;IACpC,OAAO,EAAE,OAAO,CAAC;IACjB,yDAAyD;IACzD,YAAY,EAAE,MAAM,EAAE,CAAC;IACvB,8BAA8B;IAC9B,WAAW,EAAE,MAAM,CAAC;IACpB,4BAA4B;IAC5B,OAAO,EAAE;QACP,eAAe,EAAE,MAAM,CAAC;QACxB,iBAAiB,EAAE,MAAM,CAAC;QAC1B,eAAe,EAAE,MAAM,CAAC;KACzB,CAAC;CACH;AAMD;;;;;GAKG;AACH,MAAM,MAAM,oBAAoB,GAAG,QAAQ,GAAG,QAAQ,GAAG,SAAS,GAAG,OAAO,CAAC;AAE7E,uEAAuE;AACvE,MAAM,WAAW,sBAAsB;IACrC,SAAS,EAAE,MAAM,CAAC;IAClB,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,UAAU,CAAC,EAAE;QACX,MAAM,EAAE,OAAO,GAAG,SAAS,CAAC;QAC5B,IAAI,CAAC,EAAE,MAAM,CAAC;KACf,CAAC;CACH;AAED;;GAEG;AACH,MAAM,WAAW,UAAU;IACzB,EAAE,EAAE,MAAM,CAAC;IACX,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,EAAE,EAAE,OAAO,CAAC;IACZ,mFAAmF;IACnF,MAAM,CAAC,EAAE,oBAAoB,CAAC;IAC9B,wEAAwE;IACxE,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,KAAK,EAAE,MAAM,CAAC;IACd,WAAW,EAAE,MAAM,CAAC;IACpB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,SAAS,EAAE,MAAM,CAAC;IAClB,MAAM,EAAE;QACN,MAAM,EAAE,MAAM,CAAC;QACf,UAAU,EAAE,MAAM,CAAC;QACnB,KAAK,EAAE,MAAM,CAAC;KACf,CAAC;IACF,MAAM,EAAE,MAAM,GAAG,MAAM,CAAC;IACxB,QAAQ,EAAE,MAAM,CAAC;IACjB,QAAQ,EAAE,MAAM,CAAC;IACjB,IAAI,EAAE,MAAM,EAAE,CAAC;IACf,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,uFAAuF;IACvF,QAAQ,CAAC,EAAE,sBAAsB,CAAC;IAClC,0CAA0C;IAC1C,SAAS,CAAC,EAAE,iBAAiB,CAAC;IAC9B,+DAA+D;IAC/D,SAAS,CAAC,EAAE,OAAO,UAAU,EAAE,cAAc,EAAE,CAAC;IAChD,gDAAgD;IAChD,QAAQ,CAAC,EAAE,OAAO,UAAU,EAAE,eAAe,CAAC;CAC/C;AAED;;GAEG;AACH,MAAM,WAAW,gBAAgB;IAC/B,kCAAkC;IAClC,SAAS,EAAE,MAAM,CAAC;IAClB,mCAAmC;IACnC,eAAe,EAAE,MAAM,CAAC;IACxB,wCAAwC;IACxC,mBAAmB,EAAE,MAAM,CAAC;IAC5B,sCAAsC;IACtC,KAAK,EAAE,MAAM,CAAC;IACd,mCAAmC;IACnC,OAAO,EAAE;QACP,aAAa,EAAE,MAAM,CAAC;QACtB,iBAAiB,EAAE,MAAM,CAAC;KAC3B,CAAC;CACH;AAED;;GAEG;AACH,MAAM,WAAW,UAAU;IACzB,YAAY,EAAE,MAAM,CAAC;IACrB,uDAAuD;IACvD,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,WAAW,EAAE,MAAM,CAAC;IACpB,sEAAsE;IACtE,iBAAiB,CAAC,EAAE,MAAM,CAAC;IAC3B,mEAAmE;IACnE,mBAAmB,CAAC,EAAE,MAAM,CAAC;IAC7B,+EAA+E;IAC/E,wBAAwB,CAAC,EAAE,MAAM,CAAC;IAClC,YAAY,EAAE,MAAM,CAAC;IACrB,YAAY,EAAE,MAAM,CAAC;IACrB,iBAAiB,EAAE,MAAM,CAAC;IAC1B,cAAc,EAAE,MAAM,CAAC;IACvB,YAAY,EAAE,MAAM,CAAC;IACrB,mBAAmB,EAAE,MAAM,CAAC;IAC5B,uBAAuB,EAAE,MAAM,CAAC;IAChC,iCAAiC;IACjC,IAAI,CAAC,EAAE,gBAAgB,CAAC;CACzB;AAED;;GAEG;AACH,MAAM,WAAW,OAAO;IACtB,MAAM,EAAE,MAAM,CAAC;IACf,MAAM,EAAE,MAAM,CAAC;IACf,KAAK,EAAE,OAAO,CAAC;IACf,MAAM,CAAC,EAAE,MAAM,CAAC;CACjB;AAED;;GAEG;AACH,MAAM,WAAW,cAAc;IAC7B,MAAM,EAAE,MAAM,CAAC;IACf,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,EAAE,CAAC,EAAE;QACH,QAAQ,EAAE,MAAM,CAAC;QACjB,QAAQ,EAAE,MAAM,CAAC;QACjB,SAAS,CAAC,EAAE,MAAM,CAAC;KACpB,CAAC;CACH;AAED;;GAEG;AACH,MAAM,MAAM,YAAY,GAAG,KAAK,GAAG,UAAU,GAAG,QAAQ,GAAG,KAAK,GAAG,SAAS,CAAC;AAE7E;;;GAGG;AACH,MAAM,WAAW,cAAc;IAC7B,oBAAoB;IACpB,QAAQ,EAAE,MAAM,CAAC;IACjB,yCAAyC;IACzC,KAAK,CAAC,EAAE,MAAM,CAAC;IAGf,6BAA6B;IAC7B,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,0CAA0C;IAC1C,QAAQ,CAAC,EAAE,MAAM,CAAC;IAGlB,0BAA0B;IAC1B,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB,4BAA4B;IAC5B,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,wBAAwB;IACxB,WAAW,CAAC,EAAE,MAAM,CAAC;IAGrB,4CAA4C;IAC5C,mBAAmB,CAAC,EAAE,MAAM,CAAC;IAG7B,0CAA0C;IAC1C,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,iCAAiC;IACjC,aAAa,CAAC,EAAE,MAAM,CAAC;IAGvB,qCAAqC;IACrC,cAAc,CAAC,EAAE,OAAO,CAAC;IACzB,mCAAmC;IACnC,gBAAgB,CAAC,EAAE,OAAO,CAAC;IAG3B,4BAA4B;IAC5B,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,kBAAkB;IAClB,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,0BAA0B;IAC1B,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,iBAAiB;IACjB,UAAU,CAAC,EAAE,MAAM,CAAC;IAEpB,mDAAmD;IACnD,MAAM,EAAE;QACN,QAAQ,CAAC,EAAE,YAAY,CAAC;QACxB,KAAK,CAAC,EAAE,YAAY,CAAC;QACrB,YAAY,CAAC,EAAE,YAAY,CAAC;QAC5B,QAAQ,CAAC,EAAE,YAAY,CAAC;QACxB,aAAa,CAAC,EAAE,YAAY,CAAC;QAC7B,eAAe,CAAC,EAAE,YAAY,CAAC;QAC/B,WAAW,CAAC,EAAE,YAAY,CAAC;QAC3B,mBAAmB,CAAC,EAAE,YAAY,CAAC;QACnC,IAAI,CAAC,EAAE,YAAY,CAAC;QACpB,aAAa,CAAC,EAAE,YAAY,CAAC;QAC7B,cAAc,CAAC,EAAE,YAAY,CAAC;QAC9B,gBAAgB,CAAC,EAAE,YAAY,CAAC;QAChC,OAAO,CAAC,EAAE,YAAY,CAAC;QACvB,WAAW,CAAC,EAAE,YAAY,CAAC;QAC3B,WAAW,CAAC,EAAE,YAAY,CAAC;QAC3B,UAAU,CAAC,EAAE,YAAY,CAAC;KAC3B,CAAC;CACH;AAED;;GAEG;AACH,MAAM,WAAW,SAAS;IACxB,QAAQ,EAAE,MAAM,CAAC;IACjB,QAAQ,EAAE,MAAM,CAAC;IACjB,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,IAAI,CAAC,EAAE,MAAM,CAAC;CACf;AAED;;GAEG;AACH,MAAM,WAAW,WAAW;IAC1B,OAAO,EAAE,MAAM,CAAC;IAChB,MAAM,EAAE,MAAM,CAAC;IACf,OAAO,EAAE,MAAM,CAAC;IAChB,UAAU,EAAE,MAAM,CAAC;IACnB,QAAQ,EAAE,MAAM,CAAC;IACjB,WAAW,EAAE,MAAM,CAAC;IACpB,MAAM,EAAE,SAAS,CAAC;IAClB,4EAA4E;IAC5E,eAAe,CAAC,EAAE,cAAc,CAAC;IACjC,OAAO,EAAE,UAAU,CAAC;IACpB,GAAG,EAAE,OAAO,CAAC;IACb,UAAU,EAAE,cAAc,CAAC;IAC3B,KAAK,EAAE,UAAU,EAAE,CAAC;IACpB,WAAW,EAAE;QACX,YAAY,EAAE,MAAM,CAAC;QACrB,QAAQ,EAAE,MAAM,CAAC;QACjB,IAAI,EAAE,MAAM,CAAC;KACd,CAAC;IACF,yCAAyC;IACzC,SAAS,CAAC,EAAE,qBAAqB,CAAC;CACnC;AAED;;;;GAIG;AACH,wBAAgB,uBAAuB,CAAC,UAAU,EAAE,UAAU,GAAG,oBAAoB,CAIpF;AAMD;;GAEG;AACH,MAAM,MAAM,aAAa,GAAG,MAAM,GAAG,QAAQ,GAAG,SAAS,GAAG,OAAO,CAAC;AAEpE;;GAEG;AACH,MAAM,MAAM,eAAe,GAAG,MAAM,GAAG,KAAK,GAAG,QAAQ,GAAG,MAAM,GAAG,UAAU,CAAC;AAE9E;;GAEG;AACH,MAAM,WAAW,iBAAiB;IAChC,4BAA4B;IAC5B,MAAM,EAAE,MAAM,CAAC;IACf,wBAAwB;IACxB,QAAQ,EAAE,MAAM,CAAC;IACjB,uCAAuC;IACvC,MAAM,EAAE,MAAM,CAAC;IACf,qBAAqB;IACrB,QAAQ,EAAE,MAAM,CAAC;IACjB,oBAAoB;IACpB,MAAM,EAAE,aAAa,CAAC;IACtB,yBAAyB;IACzB,QAAQ,EAAE,eAAe,CAAC;IAC1B,6BAA6B;IAC7B,OAAO,EAAE,MAAM,EAAE,CAAC;IAClB,8BAA8B;IAC9B,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,0CAA0C;IAC1C,SAAS,CAAC,EAAE,iBAAiB,CAAC;CAC/B;AAED;;GAEG;AACH,MAAM,WAAW,cAAc;IAC7B,gCAAgC;IAChC,WAAW,EAAE,MAAM,CAAC;IACpB,yCAAyC;IACzC,cAAc,EAAE,MAAM,CAAC;IACvB,+CAA+C;IAC/C,iBAAiB,EAAE,MAAM,CAAC;IAC1B,8CAA8C;IAC9C,gBAAgB,EAAE,MAAM,CAAC;IACzB,yBAAyB;IACzB,eAAe,EAAE,MAAM,CAAC;IACxB,sCAAsC;IACtC,QAAQ,EAAE,MAAM,CAAC;IACjB,iDAAiD;IACjD,YAAY,EAAE,MAAM,CAAC;IACrB,4BAA4B;IAC5B,WAAW,EAAE;QACX,GAAG,EAAE,MAAM,CAAC;QACZ,MAAM,EAAE,MAAM,CAAC;QACf,IAAI,EAAE,MAAM,CAAC;QACb,QAAQ,EAAE,MAAM,CAAC;KAClB,CAAC;CACH;AAED;;GAEG;AACH,MAAM,WAAW,aAAa;IAC5B,QAAQ,EAAE,MAAM,CAAC;IACjB,QAAQ,EAAE,MAAM,CAAC;IACjB,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,SAAS,EAAE,MAAM,EAAE,CAAC;IACpB,cAAc,EAAE,MAAM,CAAC;CACxB;AAED;;GAEG;AACH,MAAM,WAAW,eAAe;IAC9B,OAAO,EAAE,MAAM,CAAC;IAChB,IAAI,EAAE,SAAS,CAAC;IAChB,MAAM,EAAE,MAAM,CAAC;IACf,OAAO,EAAE,MAAM,CAAC;IAChB,UAAU,EAAE,MAAM,CAAC;IACnB,QAAQ,EAAE,MAAM,CAAC;IACjB,WAAW,EAAE,MAAM,CAAC;IACpB,MAAM,EAAE,aAAa,CAAC;IACtB,4EAA4E;IAC5E,eAAe,CAAC,EAAE,cAAc,CAAC;IACjC,OAAO,EAAE,cAAc,CAAC;IACxB,GAAG,EAAE,OAAO,CAAC;IACb,UAAU,EAAE,cAAc,CAAC;IAC3B,OAAO,EAAE,iBAAiB,EAAE,CAAC;IAC7B,WAAW,EAAE;QACX,YAAY,EAAE,MAAM,CAAC;QACrB,QAAQ,EAAE,MAAM,CAAC;QACjB,IAAI,EAAE,MAAM,CAAC;KACd,CAAC;IACF,yCAAyC;IACzC,SAAS,CAAC,EAAE,qBAAqB,CAAC;CACnC;AAMD;;GAEG;AACH,MAAM,WAAW,mBAAmB;IAClC,oCAAoC;IACpC,OAAO,EAAE,OAAO,CAAC;IACjB,8BAA8B;IAC9B,SAAS,EAAE,MAAM,CAAC;IAClB,8BAA8B;IAC9B,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,+BAA+B;IAC/B,SAAS,EAAE,MAAM,CAAC;IAClB,mCAAmC;IACnC,MAAM,CAAC,EAAE;QACP,0BAA0B;QAC1B,MAAM,EAAE,MAAM,CAAC;QACf,+BAA+B;QAC/B,UAAU,EAAE,MAAM,CAAC;QACnB,mBAAmB;QACnB,KAAK,EAAE,MAAM,CAAC;KACf,CAAC;CACH;AAED;;GAEG;AACH,MAAM,WAAW,aAAa;IAC5B,0BAA0B;IAC1B,cAAc,EAAE,MAAM,CAAC;IACvB,0BAA0B;IAC1B,mBAAmB,EAAE,MAAM,CAAC;IAC5B,sBAAsB;IACtB,eAAe,EAAE,MAAM,CAAC;IACxB,yBAAyB;IACzB,YAAY,EAAE,MAAM,CAAC;IACrB,0BAA0B;IAC1B,mBAAmB,EAAE,MAAM,CAAC;IAC5B,4BAA4B;IAC5B,cAAc,EAAE,MAAM,CAAC;IACvB,4BAA4B;IAC5B,cAAc,EAAE,MAAM,CAAC;IACvB,4BAA4B;IAC5B,cAAc,EAAE,MAAM,CAAC;IACvB,8BAA8B;IAC9B,cAAc,EAAE,MAAM,CAAC;IACvB,8BAA8B;IAC9B,cAAc,EAAE,MAAM,CAAC;IACvB,8BAA8B;IAC9B,cAAc,EAAE,MAAM,CAAC;IACvB,8BAA8B;IAC9B,cAAc,EAAE,MAAM,CAAC;IACvB,6EAA6E;IAC7E,MAAM,CAAC,EAAE;QACP,oDAAoD;QACpD,mBAAmB,EAAE,MAAM,CAAC;QAC5B,yDAAyD;QACzD,uBAAuB,EAAE,MAAM,CAAC;QAChC,yCAAyC;QACzC,YAAY,EAAE,MAAM,CAAC;QACrB,iCAAiC;QACjC,sBAAsB,EAAE,MAAM,CAAC;KAChC,CAAC;IACF,+EAA+E;IAC/E,IAAI,CAAC,EAAE;QACL,kCAAkC;QAClC,mBAAmB,EAAE,MAAM,CAAC;QAC5B,mCAAmC;QACnC,SAAS,EAAE;YACT,mCAAmC;YACnC,eAAe,EAAE,MAAM,CAAC;YACxB,wCAAwC;YACxC,mBAAmB,EAAE,MAAM,CAAC;SAC7B,CAAC;QACF,sCAAsC;QACtC,KAAK,EAAE,MAAM,CAAC;QACd,mCAAmC;QACnC,OAAO,EAAE;YACP,aAAa,EAAE,MAAM,CAAC;YACtB,iBAAiB,EAAE,MAAM,CAAC;SAC3B,CAAC;KACH,CAAC;CACH;AAED;;GAEG;AACH,MAAM,WAAW,YAAY;IAC3B,QAAQ,EAAE,MAAM,CAAC;IACjB,QAAQ,EAAE,MAAM,CAAC;IACjB,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,WAAW,EAAE,MAAM,CAAC;IACpB,gBAAgB,EAAE,MAAM,CAAC;IACzB,eAAe,EAAE,MAAM,CAAC;IACxB,YAAY,CAAC,EAAE,MAAM,CAAC;CACvB;AAED;;GAEG;AACH,MAAM,WAAW,cAAc;IAC7B,OAAO,EAAE,MAAM,CAAC;IAChB,IAAI,EAAE,QAAQ,CAAC;IACf,MAAM,EAAE,MAAM,CAAC;IACf,OAAO,EAAE,MAAM,CAAC;IAChB,UAAU,EAAE,MAAM,CAAC;IACnB,QAAQ,EAAE,MAAM,CAAC;IACjB,WAAW,EAAE,MAAM,CAAC;IACpB,MAAM,EAAE,YAAY,CAAC;IACrB,4EAA4E;IAC5E,eAAe,CAAC,EAAE,cAAc,CAAC;IACjC,OAAO,EAAE,aAAa,CAAC;IACvB,GAAG,EAAE,OAAO,CAAC;IACb,UAAU,EAAE,cAAc,CAAC;IAC3B,mEAAmE;IACnE,cAAc,EAAE,mBAAmB,EAAE,CAAC;IACtC,WAAW,EAAE;QACX,YAAY,EAAE,MAAM,CAAC;QACrB,QAAQ,EAAE,MAAM,CAAC;QACjB,IAAI,EAAE,MAAM,CAAC;KACd,CAAC;IACF,yCAAyC;IACzC,SAAS,CAAC,EAAE,qBAAqB,CAAC;CACnC;AAMD;;GAEG;AACH,MAAM,MAAM,WAAW,GAAG,WAAW,GAAG,eAAe,GAAG,cAAc,CAAC;AAEzE;;GAEG;AACH,wBAAgB,aAAa,CAAC,QAAQ,EAAE,WAAW,GAAG,QAAQ,IAAI,WAAW,CAE5E;AAED;;GAEG;AACH,wBAAgB,iBAAiB,CAAC,QAAQ,EAAE,WAAW,GAAG,QAAQ,IAAI,eAAe,CAEpF;AAED;;GAEG;AACH,wBAAgB,gBAAgB,CAAC,QAAQ,EAAE,WAAW,GAAG,QAAQ,IAAI,cAAc,CAElF"}
|
|
@@ -25,4 +25,5 @@ export { JsonSchemaEvaluator } from './json-schema';
|
|
|
25
25
|
export { LLMGraderEvaluator } from './llm-grader';
|
|
26
26
|
export { SimilarityEvaluator } from './similarity';
|
|
27
27
|
export { InlineEvaluator, SUPPORTED_EXPRESSIONS } from './inline';
|
|
28
|
+
export { ToolTraceEvaluator } from './tool-trace';
|
|
28
29
|
//# sourceMappingURL=index.d.ts.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/evaluators/index.ts"],"names":[],"mappings":"AAAA;;GAEG;
|
|
1
|
+
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/evaluators/index.ts"],"names":[],"mappings":"AAAA;;GAEG;AAaH,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,SAAS,CAAC;AAezC;;GAEG;AACH,wBAAgB,YAAY,CAAC,IAAI,EAAE,MAAM,GAAG,SAAS,CAOpD;AAED;;GAEG;AACH,wBAAgB,iBAAiB,CAAC,IAAI,EAAE,MAAM,EAAE,SAAS,EAAE,SAAS,GAAG,IAAI,CAE1E;AAED;;GAEG;AACH,wBAAgB,cAAc,IAAI,MAAM,EAAE,CAEzC;AAED,cAAc,SAAS,CAAC;AACxB,OAAO,EAAE,cAAc,EAAE,MAAM,SAAS,CAAC;AACzC,OAAO,EAAE,cAAc,EAAE,MAAM,SAAS,CAAC;AACzC,OAAO,EAAE,cAAc,EAAE,MAAM,SAAS,CAAC;AACzC,OAAO,EAAE,iBAAiB,EAAE,MAAM,YAAY,CAAC;AAC/C,OAAO,EAAE,oBAAoB,EAAE,MAAM,gBAAgB,CAAC;AACtD,OAAO,EAAE,iBAAiB,EAAE,MAAM,YAAY,CAAC;AAC/C,OAAO,EAAE,mBAAmB,EAAE,MAAM,eAAe,CAAC;AACpD,OAAO,EAAE,kBAAkB,EAAE,MAAM,cAAc,CAAC;AAClD,OAAO,EAAE,mBAAmB,EAAE,MAAM,cAAc,CAAC;AACnD,OAAO,EAAE,eAAe,EAAE,qBAAqB,EAAE,MAAM,UAAU,CAAC;AAClE,OAAO,EAAE,kBAAkB,EAAE,MAAM,cAAc,CAAC"}
|
|
@@ -6,6 +6,5 @@ import type { Evaluator, EvaluatorResult } from './types';
|
|
|
6
6
|
export declare class JsonSchemaEvaluator implements Evaluator {
|
|
7
7
|
readonly type = "json_schema";
|
|
8
8
|
evaluate(response: string, expected: Expected): Promise<EvaluatorResult>;
|
|
9
|
-
private jsonSchemaToZod;
|
|
10
9
|
}
|
|
11
10
|
//# sourceMappingURL=json-schema.d.ts.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"json-schema.d.ts","sourceRoot":"","sources":["../../src/evaluators/json-schema.ts"],"names":[],"mappings":"AAAA;;GAEG;AAGH,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,oBAAoB,CAAC;AACnD,OAAO,KAAK,EAAE,SAAS,EAAE,eAAe,EAAE,MAAM,SAAS,CAAC;
|
|
1
|
+
{"version":3,"file":"json-schema.d.ts","sourceRoot":"","sources":["../../src/evaluators/json-schema.ts"],"names":[],"mappings":"AAAA;;GAEG;AAGH,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,oBAAoB,CAAC;AACnD,OAAO,KAAK,EAAE,SAAS,EAAE,eAAe,EAAE,MAAM,SAAS,CAAC;AAa1D,qBAAa,mBAAoB,YAAW,SAAS;IACnD,QAAQ,CAAC,IAAI,iBAAiB;IAExB,QAAQ,CAAC,QAAQ,EAAE,MAAM,EAAE,QAAQ,EAAE,QAAQ,GAAG,OAAO,CAAC,eAAe,CAAC;CAkD/E"}
|
|
@@ -7,5 +7,7 @@ export declare class LLMGraderEvaluator implements Evaluator {
|
|
|
7
7
|
readonly type = "llm_grader";
|
|
8
8
|
evaluate(response: string, expected: Expected, context?: EvaluatorContext): Promise<EvaluatorResult>;
|
|
9
9
|
private parseGraderResponse;
|
|
10
|
+
private parseStrictGraderResponse;
|
|
11
|
+
private parseLegacyGraderResponse;
|
|
10
12
|
}
|
|
11
13
|
//# sourceMappingURL=llm-grader.d.ts.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"llm-grader.d.ts","sourceRoot":"","sources":["../../src/evaluators/llm-grader.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,oBAAoB,CAAC;AACnD,OAAO,KAAK,EAAE,SAAS,EAAE,gBAAgB,EAAE,eAAe,EAAE,MAAM,SAAS,CAAC;AAwB5E,qBAAa,kBAAmB,YAAW,SAAS;IAClD,QAAQ,CAAC,IAAI,gBAAgB;IAEvB,QAAQ,CACZ,QAAQ,EAAE,MAAM,EAChB,QAAQ,EAAE,QAAQ,EAClB,OAAO,CAAC,EAAE,gBAAgB,GACzB,OAAO,CAAC,eAAe,CAAC;
|
|
1
|
+
{"version":3,"file":"llm-grader.d.ts","sourceRoot":"","sources":["../../src/evaluators/llm-grader.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,oBAAoB,CAAC;AACnD,OAAO,KAAK,EAAE,SAAS,EAAE,gBAAgB,EAAE,eAAe,EAAE,MAAM,SAAS,CAAC;AAwB5E,qBAAa,kBAAmB,YAAW,SAAS;IAClD,QAAQ,CAAC,IAAI,gBAAgB;IAEvB,QAAQ,CACZ,QAAQ,EAAE,MAAM,EAChB,QAAQ,EAAE,QAAQ,EAClB,OAAO,CAAC,EAAE,gBAAgB,GACzB,OAAO,CAAC,eAAe,CAAC;IAoD3B,OAAO,CAAC,mBAAmB;IAQ3B,OAAO,CAAC,yBAAyB;IA0BjC,OAAO,CAAC,yBAAyB;CA+ClC"}
|
|
@@ -0,0 +1,7 @@
|
|
|
1
|
+
import type { Expected } from '../scenario/schema';
|
|
2
|
+
import type { Evaluator, EvaluatorContext, EvaluatorResult } from './types';
|
|
3
|
+
export declare class ToolTraceEvaluator implements Evaluator {
|
|
4
|
+
readonly type = "tool_trace";
|
|
5
|
+
evaluate(_response: string, expected: Expected, context?: EvaluatorContext): Promise<EvaluatorResult>;
|
|
6
|
+
}
|
|
7
|
+
//# sourceMappingURL=tool-trace.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"tool-trace.d.ts","sourceRoot":"","sources":["../../src/evaluators/tool-trace.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,oBAAoB,CAAC;AACnD,OAAO,KAAK,EAAE,SAAS,EAAE,gBAAgB,EAAE,eAAe,EAAE,MAAM,SAAS,CAAC;AAI5E,qBAAa,kBAAmB,YAAW,SAAS;IAClD,QAAQ,CAAC,IAAI,gBAAgB;IAEvB,QAAQ,CACZ,SAAS,EAAE,MAAM,EACjB,QAAQ,EAAE,QAAQ,EAClB,OAAO,CAAC,EAAE,gBAAgB,GACzB,OAAO,CAAC,eAAe,CAAC;CAuB5B"}
|
|
@@ -3,12 +3,14 @@
|
|
|
3
3
|
*/
|
|
4
4
|
import type { ModelClient } from '../adapters/types';
|
|
5
5
|
import type { Expected, TestCase } from '../scenario/schema';
|
|
6
|
+
import type { ToolTraceEntry } from '../tools';
|
|
6
7
|
/**
|
|
7
8
|
* Context provided to evaluators
|
|
8
9
|
*/
|
|
9
10
|
export interface EvaluatorContext {
|
|
10
11
|
client?: ModelClient;
|
|
11
12
|
testCase?: TestCase;
|
|
13
|
+
toolTrace?: ToolTraceEntry[];
|
|
12
14
|
}
|
|
13
15
|
/**
|
|
14
16
|
* Result from an evaluation
|
|
@@ -17,6 +19,24 @@ export interface EvaluatorResult {
|
|
|
17
19
|
passed: boolean;
|
|
18
20
|
score: number;
|
|
19
21
|
reason?: string;
|
|
22
|
+
/**
|
|
23
|
+
* Whether this evaluator produced a valid measurement. Omitted by legacy
|
|
24
|
+
* evaluators; the executor derives it from `passed` for compatibility.
|
|
25
|
+
*/
|
|
26
|
+
status?: 'passed' | 'failed' | 'invalid';
|
|
27
|
+
/** Bounded evaluator metadata that may be retained in a run artifact. */
|
|
28
|
+
evidence?: {
|
|
29
|
+
threshold?: number;
|
|
30
|
+
model?: string;
|
|
31
|
+
validation?: {
|
|
32
|
+
status: 'valid' | 'invalid';
|
|
33
|
+
code?: string;
|
|
34
|
+
};
|
|
35
|
+
};
|
|
36
|
+
/**
|
|
37
|
+
* Legacy evaluator-private metadata. The standard executor never serializes
|
|
38
|
+
* this field; use `evidence` for reviewed, bounded artifact data.
|
|
39
|
+
*/
|
|
20
40
|
details?: Record<string, unknown>;
|
|
21
41
|
}
|
|
22
42
|
/**
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../../src/evaluators/types.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,mBAAmB,CAAC;AACrD,OAAO,KAAK,EAAE,QAAQ,EAAE,QAAQ,EAAE,MAAM,oBAAoB,CAAC;
|
|
1
|
+
{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../../src/evaluators/types.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,mBAAmB,CAAC;AACrD,OAAO,KAAK,EAAE,QAAQ,EAAE,QAAQ,EAAE,MAAM,oBAAoB,CAAC;AAC7D,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,UAAU,CAAC;AAE/C;;GAEG;AACH,MAAM,WAAW,gBAAgB;IAC/B,MAAM,CAAC,EAAE,WAAW,CAAC;IACrB,QAAQ,CAAC,EAAE,QAAQ,CAAC;IACpB,SAAS,CAAC,EAAE,cAAc,EAAE,CAAC;CAC9B;AAED;;GAEG;AACH,MAAM,WAAW,eAAe;IAC9B,MAAM,EAAE,OAAO,CAAC;IAChB,KAAK,EAAE,MAAM,CAAC;IACd,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB;;;OAGG;IACH,MAAM,CAAC,EAAE,QAAQ,GAAG,QAAQ,GAAG,SAAS,CAAC;IACzC,yEAAyE;IACzE,QAAQ,CAAC,EAAE;QACT,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,KAAK,CAAC,EAAE,MAAM,CAAC;QACf,UAAU,CAAC,EAAE;YACX,MAAM,EAAE,OAAO,GAAG,SAAS,CAAC;YAC5B,IAAI,CAAC,EAAE,MAAM,CAAC;SACf,CAAC;KACH,CAAC;IACF;;;OAGG;IACH,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CACnC;AAED;;GAEG;AACH,MAAM,WAAW,SAAS;IACxB,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IAEtB,QAAQ,CACN,QAAQ,EAAE,MAAM,EAChB,QAAQ,EAAE,QAAQ,EAClB,OAAO,CAAC,EAAE,gBAAgB,GACzB,OAAO,CAAC,eAAe,CAAC,CAAC;CAC7B"}
|
package/dist/index.d.ts
CHANGED
package/dist/index.d.ts.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA;;;GAGG;AAGH,cAAc,YAAY,CAAC;AAG3B,cAAc,YAAY,CAAC;AAG3B,cAAc,cAAc,CAAC;AAG7B,cAAc,UAAU,CAAC;AAGzB,cAAc,WAAW,CAAC;AAG1B,cAAc,aAAa,CAAC;AAG5B,cAAc,cAAc,CAAC;AAG7B,cAAc,SAAS,CAAC;AAGxB,cAAc,aAAa,CAAC;AAG5B,cAAc,QAAQ,CAAC;AAGvB,cAAc,aAAa,CAAC"}
|
|
1
|
+
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA;;;GAGG;AAGH,cAAc,YAAY,CAAC;AAG3B,cAAc,YAAY,CAAC;AAG3B,cAAc,cAAc,CAAC;AAG7B,cAAc,UAAU,CAAC;AAGzB,cAAc,WAAW,CAAC;AAG1B,cAAc,aAAa,CAAC;AAG5B,cAAc,cAAc,CAAC;AAG7B,cAAc,SAAS,CAAC;AAGxB,cAAc,aAAa,CAAC;AAG5B,cAAc,QAAQ,CAAC;AAGvB,cAAc,SAAS,CAAC;AAGxB,cAAc,oBAAoB,CAAC;AAGnC,cAAc,aAAa,CAAC"}
|