@tangle-network/agent-eval 0.121.0 → 0.122.1
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/dist/analyst/index.d.ts +1 -0
- package/dist/analyst/index.js +1 -1
- package/dist/belief-state/index.d.ts +43 -0
- package/dist/belief-state/index.js +48 -326
- package/dist/belief-state/index.js.map +1 -1
- package/dist/benchmarks/index.d.ts +5 -1
- package/dist/benchmarks/index.js +2 -2
- package/dist/campaign/index.d.ts +16 -3
- package/dist/campaign/index.js +4 -2
- package/dist/chunk-7VYEGHWF.js +1425 -0
- package/dist/chunk-7VYEGHWF.js.map +1 -0
- package/dist/{chunk-JM2SKQMS.js → chunk-P62I523A.js} +3 -1
- package/dist/{chunk-JM2SKQMS.js.map → chunk-P62I523A.js.map} +1 -1
- package/dist/{chunk-JN2FCO5W.js → chunk-TDHX3BVA.js} +157 -4
- package/dist/chunk-TDHX3BVA.js.map +1 -0
- package/dist/{chunk-32BZXMSO.js → chunk-VZ6VKOJT.js} +23 -160
- package/dist/chunk-VZ6VKOJT.js.map +1 -0
- package/dist/contract/index.d.ts +118 -14
- package/dist/contract/index.js +548 -497
- package/dist/contract/index.js.map +1 -1
- package/dist/index.d.ts +7 -19
- package/dist/index.js +7 -5
- package/dist/index.js.map +1 -1
- package/dist/openapi.json +1 -1
- package/dist/rl.d.ts +5 -1
- package/package.json +2 -2
- package/dist/chunk-32BZXMSO.js.map +0 -1
- package/dist/chunk-HKUCJ437.js +0 -787
- package/dist/chunk-HKUCJ437.js.map +0 -1
- package/dist/chunk-JN2FCO5W.js.map +0 -1
package/dist/openapi.json
CHANGED
|
@@ -2,7 +2,7 @@
|
|
|
2
2
|
"openapi": "3.1.0",
|
|
3
3
|
"info": {
|
|
4
4
|
"title": "@tangle-network/agent-eval — wire protocol",
|
|
5
|
-
"version": "0.
|
|
5
|
+
"version": "0.122.1",
|
|
6
6
|
"description": "HTTP and stdio RPC interface to agent-eval. The TypeScript runtime is the source of truth; this spec is the contract that cross-language clients (Python, Rust, Go) generate from.\n\nWire-protocol version: 1.0.0. Bumps on breaking changes to request/response schemas.",
|
|
7
7
|
"contact": {
|
|
8
8
|
"name": "Tangle Network",
|
package/dist/rl.d.ts
CHANGED
|
@@ -3700,11 +3700,15 @@ interface GenerationCandidate {
|
|
|
3700
3700
|
* reflective proposer grounds its next edit on. Keep `notes` GENERALIZABLE
|
|
3701
3701
|
* (which checks/lines/dimensions failed and how), NOT case-specific ground
|
|
3702
3702
|
* truth: leaking expected answers into the prompt is memorization, and the
|
|
3703
|
-
* held-out gate would reject it anyway.
|
|
3703
|
+
* held-out gate would reject it anyway. `emitted` is a bounded excerpt of
|
|
3704
|
+
* the candidate's raw output for the scenario (worst rep when reps > 1) —
|
|
3705
|
+
* the "what it actually did" evidence; optional so trajectory capture is
|
|
3706
|
+
* never required of a dispatch. */
|
|
3704
3707
|
scenarios: Array<{
|
|
3705
3708
|
scenarioId: string;
|
|
3706
3709
|
composite: number;
|
|
3707
3710
|
notes?: string;
|
|
3711
|
+
emitted?: string;
|
|
3708
3712
|
}>;
|
|
3709
3713
|
/** Proposer-supplied short label for the change. Present when the proposer
|
|
3710
3714
|
* returned a `ProposedCandidate`; absent for bare-surface mutators. */
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@tangle-network/agent-eval",
|
|
3
|
-
"version": "0.
|
|
3
|
+
"version": "0.122.1",
|
|
4
4
|
"description": "Evaluate and improve AI agents from runs, traces, judges, and feedback. Compare candidates, cluster failures, measure lift, and gate releases.",
|
|
5
5
|
"homepage": "https://github.com/tangle-network/agent-eval#readme",
|
|
6
6
|
"repository": {
|
|
@@ -152,7 +152,7 @@
|
|
|
152
152
|
"@ax-llm/ax": "^23.0.0",
|
|
153
153
|
"@hono/node-server": "^2.0.0",
|
|
154
154
|
"@tangle-network/agent-core": "^0.4.11",
|
|
155
|
-
"@tangle-network/agent-interface": "^0.
|
|
155
|
+
"@tangle-network/agent-interface": "^0.30.0",
|
|
156
156
|
"@tangle-network/tcloud": "^0.4.14",
|
|
157
157
|
"hono": "^4.12.30",
|
|
158
158
|
"zod": "^4.3.6"
|