agent-hitch 0.2.7 → 0.2.8
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 +16 -2
- package/README.zh-CN.md +14 -2
- package/dist/bin/hitch.js +25 -3
- package/dist/bin/hitch.js.map +1 -1
- package/dist/scripts/canary-benchmark.js +54 -0
- package/dist/scripts/canary-benchmark.js.map +1 -0
- package/dist/scripts/canary-eval-scheduler-throughput.js +47 -0
- package/dist/scripts/canary-eval-scheduler-throughput.js.map +1 -0
- package/dist/scripts/check-architecture.js +3 -2
- package/dist/scripts/check-architecture.js.map +1 -1
- package/dist/src/adapters/catalog.js +2 -1
- package/dist/src/adapters/catalog.js.map +1 -1
- package/dist/src/adapters/providers/codex-auth.js +29 -0
- package/dist/src/adapters/providers/codex-auth.js.map +1 -0
- package/dist/src/adapters/providers/codex.js +23 -6
- package/dist/src/adapters/providers/codex.js.map +1 -1
- package/dist/src/adapters/providers/model-call.js +26 -0
- package/dist/src/adapters/providers/model-call.js.map +1 -0
- package/dist/src/artifacts/index.js +1 -1
- package/dist/src/artifacts/index.js.map +1 -1
- package/dist/src/artifacts/preparer.js +7 -2
- package/dist/src/artifacts/preparer.js.map +1 -1
- package/dist/src/backends/harbor/agent-budget.js +31 -0
- package/dist/src/backends/harbor/agent-budget.js.map +1 -0
- package/dist/src/backends/harbor/backend.js +6 -4
- package/dist/src/backends/harbor/backend.js.map +1 -1
- package/dist/src/backends/harbor/index.js +1 -0
- package/dist/src/backends/harbor/index.js.map +1 -1
- package/dist/src/backends/harbor/regrade.js +77 -0
- package/dist/src/backends/harbor/regrade.js.map +1 -0
- package/dist/src/backends/index.js +1 -0
- package/dist/src/backends/index.js.map +1 -1
- package/dist/src/benchmarks/index.js +3 -0
- package/dist/src/benchmarks/index.js.map +1 -0
- package/dist/src/benchmarks/loader.js +271 -0
- package/dist/src/benchmarks/loader.js.map +1 -0
- package/dist/src/benchmarks/metrics.js +16 -0
- package/dist/src/benchmarks/metrics.js.map +1 -0
- package/dist/src/benchmarks/toml.js +13 -0
- package/dist/src/benchmarks/toml.js.map +1 -0
- package/dist/src/benchmarks/validation.js +242 -0
- package/dist/src/benchmarks/validation.js.map +1 -0
- package/dist/src/cli/arguments.js +6 -4
- package/dist/src/cli/arguments.js.map +1 -1
- package/dist/src/cli/commands/benchmark.js +22 -0
- package/dist/src/cli/commands/benchmark.js.map +1 -0
- package/dist/src/cli/commands/capabilities.js +18 -0
- package/dist/src/cli/commands/capabilities.js.map +1 -0
- package/dist/src/cli/commands/eval.js +15 -3
- package/dist/src/cli/commands/eval.js.map +1 -1
- package/dist/src/cli/commands/run.js +37 -15
- package/dist/src/cli/commands/run.js.map +1 -1
- package/dist/src/cli/commands/trajectory.js +96 -5
- package/dist/src/cli/commands/trajectory.js.map +1 -1
- package/dist/src/cli/commands/verifier.js +36 -0
- package/dist/src/cli/commands/verifier.js.map +1 -0
- package/dist/src/cli/main.js +6 -0
- package/dist/src/cli/main.js.map +1 -1
- package/dist/src/cli/output.js +14 -2
- package/dist/src/cli/output.js.map +1 -1
- package/dist/src/control-plane/eval-control-work.js +5 -0
- package/dist/src/control-plane/eval-control-work.js.map +1 -1
- package/dist/src/control-plane/eval-scheduler.js +4 -1
- package/dist/src/control-plane/eval-scheduler.js.map +1 -1
- package/dist/src/control-plane/index.js +1 -1
- package/dist/src/control-plane/index.js.map +1 -1
- package/dist/src/control-plane/remote-result-transport.js +3 -0
- package/dist/src/control-plane/remote-result-transport.js.map +1 -1
- package/dist/src/control-plane/remote-work-coordinator.js +1 -0
- package/dist/src/control-plane/remote-work-coordinator.js.map +1 -1
- package/dist/src/control-plane/remote-work-item.js +62 -0
- package/dist/src/control-plane/remote-work-item.js.map +1 -0
- package/dist/src/control-plane/remote-work-recovery.js +44 -5
- package/dist/src/control-plane/remote-work-recovery.js.map +1 -1
- package/dist/src/control-plane/remote-worker-protocol.js +2 -23
- package/dist/src/control-plane/remote-worker-protocol.js.map +1 -1
- package/dist/src/control-plane/rerun-scheduler.js +51 -14
- package/dist/src/control-plane/rerun-scheduler.js.map +1 -1
- package/dist/src/control-plane/rerun-submission.js +10 -4
- package/dist/src/control-plane/rerun-submission.js.map +1 -1
- package/dist/src/control-plane/work-admission.js +2 -0
- package/dist/src/control-plane/work-admission.js.map +1 -1
- package/dist/src/control-plane/work-dispatcher.js +9 -1
- package/dist/src/control-plane/work-dispatcher.js.map +1 -1
- package/dist/src/controller-runtime/hash.js +7 -2
- package/dist/src/controller-runtime/hash.js.map +1 -1
- package/dist/src/daemon/auth.js +5 -0
- package/dist/src/daemon/auth.js.map +1 -1
- package/dist/src/domain/benchmarks.js +2 -0
- package/dist/src/domain/benchmarks.js.map +1 -0
- package/dist/src/domain/index.js +4 -0
- package/dist/src/domain/index.js.map +1 -1
- package/dist/src/domain/runs.js.map +1 -1
- package/dist/src/domain/validation.js +11 -4
- package/dist/src/domain/validation.js.map +1 -1
- package/dist/src/domain/verifier-evidence-validation.js +214 -0
- package/dist/src/domain/verifier-evidence-validation.js.map +1 -0
- package/dist/src/domain/verifier-evidence.js +3 -0
- package/dist/src/domain/verifier-evidence.js.map +1 -0
- package/dist/src/domain/verifier-score-contract.js +232 -0
- package/dist/src/domain/verifier-score-contract.js.map +1 -0
- package/dist/src/evals/benchmark-adapter-manifest.js +219 -0
- package/dist/src/evals/benchmark-adapter-manifest.js.map +1 -0
- package/dist/src/evals/benchmark-candidate.js +36 -0
- package/dist/src/evals/benchmark-candidate.js.map +1 -0
- package/dist/src/evals/benchmark-run.js +166 -0
- package/dist/src/evals/benchmark-run.js.map +1 -0
- package/dist/src/evals/collect-only-rerun.js +2 -1
- package/dist/src/evals/collect-only-rerun.js.map +1 -1
- package/dist/src/evals/directory.js +1 -1
- package/dist/src/evals/directory.js.map +1 -1
- package/dist/src/evals/duration-estimator.js +126 -0
- package/dist/src/evals/duration-estimator.js.map +1 -0
- package/dist/src/evals/eval-lifecycle-events.js +13 -0
- package/dist/src/evals/eval-lifecycle-events.js.map +1 -1
- package/dist/src/evals/eval-logical-plan.js +51 -0
- package/dist/src/evals/eval-logical-plan.js.map +1 -0
- package/dist/src/evals/eval-result-builder.js +81 -0
- package/dist/src/evals/eval-result-builder.js.map +1 -0
- package/dist/src/evals/evolution-baseline.js +191 -0
- package/dist/src/evals/evolution-baseline.js.map +1 -0
- package/dist/src/evals/execution-plan.js +39 -3
- package/dist/src/evals/execution-plan.js.map +1 -1
- package/dist/src/evals/failure-classifier.js +61 -0
- package/dist/src/evals/failure-classifier.js.map +1 -0
- package/dist/src/evals/harbor-artifact-builder.js +6 -2
- package/dist/src/evals/harbor-artifact-builder.js.map +1 -1
- package/dist/src/evals/harbor-bridge-error.js +41 -1
- package/dist/src/evals/harbor-bridge-error.js.map +1 -1
- package/dist/src/evals/harbor-node-runtime.js +146 -0
- package/dist/src/evals/harbor-node-runtime.js.map +1 -0
- package/dist/src/evals/index.js +15 -0
- package/dist/src/evals/index.js.map +1 -1
- package/dist/src/evals/infrastructure-retry.js +109 -22
- package/dist/src/evals/infrastructure-retry.js.map +1 -1
- package/dist/src/evals/native-phase-evidence.js +440 -0
- package/dist/src/evals/native-phase-evidence.js.map +1 -0
- package/dist/src/evals/physical-retry-work.js +31 -0
- package/dist/src/evals/physical-retry-work.js.map +1 -0
- package/dist/src/evals/planned-execution-support.js +22 -10
- package/dist/src/evals/planned-execution-support.js.map +1 -1
- package/dist/src/evals/planned-execution.js +120 -112
- package/dist/src/evals/planned-execution.js.map +1 -1
- package/dist/src/evals/planned-progress-publisher.js +62 -0
- package/dist/src/evals/planned-progress-publisher.js.map +1 -0
- package/dist/src/evals/planned-retry-execution.js +174 -0
- package/dist/src/evals/planned-retry-execution.js.map +1 -0
- package/dist/src/evals/planned-retry-lifecycle.js +4 -2
- package/dist/src/evals/planned-retry-lifecycle.js.map +1 -1
- package/dist/src/evals/preparation-rerun.js +225 -0
- package/dist/src/evals/preparation-rerun.js.map +1 -0
- package/dist/src/evals/progress.js +48 -8
- package/dist/src/evals/progress.js.map +1 -1
- package/dist/src/evals/recovery.js +54 -4
- package/dist/src/evals/recovery.js.map +1 -1
- package/dist/src/evals/regrade-evidence.js +77 -0
- package/dist/src/evals/regrade-evidence.js.map +1 -0
- package/dist/src/evals/remote-infrastructure-retry.js +84 -18
- package/dist/src/evals/remote-infrastructure-retry.js.map +1 -1
- package/dist/src/evals/request.js +21 -3
- package/dist/src/evals/request.js.map +1 -1
- package/dist/src/evals/rerun-inputs.js +19 -0
- package/dist/src/evals/rerun-inputs.js.map +1 -1
- package/dist/src/evals/rerun-types.js +2 -6
- package/dist/src/evals/rerun-types.js.map +1 -1
- package/dist/src/evals/rerun.js +27 -20
- package/dist/src/evals/rerun.js.map +1 -1
- package/dist/src/evals/result-helpers.js +13 -0
- package/dist/src/evals/result-helpers.js.map +1 -1
- package/dist/src/evals/retry-backoff.js +20 -0
- package/dist/src/evals/retry-backoff.js.map +1 -0
- package/dist/src/evals/retry-state.js +195 -0
- package/dist/src/evals/retry-state.js.map +1 -0
- package/dist/src/evals/scheduler-metrics.js +95 -0
- package/dist/src/evals/scheduler-metrics.js.map +1 -0
- package/dist/src/evals/scheduler-trace-replay.js +107 -0
- package/dist/src/evals/scheduler-trace-replay.js.map +1 -0
- package/dist/src/evals/service.js +87 -138
- package/dist/src/evals/service.js.map +1 -1
- package/dist/src/evals/trial-import.js +47 -50
- package/dist/src/evals/trial-import.js.map +1 -1
- package/dist/src/evals/trial-publication-recovery.js +45 -5
- package/dist/src/evals/trial-publication-recovery.js.map +1 -1
- package/dist/src/evals/trial-reference-validation.js +49 -0
- package/dist/src/evals/trial-reference-validation.js.map +1 -0
- package/dist/src/evals/verifier-artifacts.js +317 -0
- package/dist/src/evals/verifier-artifacts.js.map +1 -0
- package/dist/src/evals/verifier-eligibility.js +25 -0
- package/dist/src/evals/verifier-eligibility.js.map +1 -0
- package/dist/src/evals/verifier-only-rerun.js +210 -0
- package/dist/src/evals/verifier-only-rerun.js.map +1 -0
- package/dist/src/evals/verifier-runtime.js +61 -0
- package/dist/src/evals/verifier-runtime.js.map +1 -0
- package/dist/src/evals/verifier-score-artifacts.js +119 -0
- package/dist/src/evals/verifier-score-artifacts.js.map +1 -0
- package/dist/src/foundation/contained-file.js +85 -0
- package/dist/src/foundation/contained-file.js.map +1 -0
- package/dist/src/foundation/credential-redaction.js +26 -1
- package/dist/src/foundation/credential-redaction.js.map +1 -1
- package/dist/src/foundation/fs.js +2 -1
- package/dist/src/foundation/fs.js.map +1 -1
- package/dist/src/foundation/index.js +2 -1
- package/dist/src/foundation/index.js.map +1 -1
- package/dist/src/runs/adapter-process.js +12 -0
- package/dist/src/runs/adapter-process.js.map +1 -0
- package/dist/src/runs/executor.js +24 -30
- package/dist/src/runs/executor.js.map +1 -1
- package/dist/src/runs/finalizer.js +17 -0
- package/dist/src/runs/finalizer.js.map +1 -1
- package/dist/src/runs/index.js +5 -0
- package/dist/src/runs/index.js.map +1 -1
- package/dist/src/runs/phase-bundle.js +86 -0
- package/dist/src/runs/phase-bundle.js.map +1 -0
- package/dist/src/runs/phase-cancellation.js +107 -0
- package/dist/src/runs/phase-cancellation.js.map +1 -0
- package/dist/src/runs/phase-group.js +113 -0
- package/dist/src/runs/phase-group.js.map +1 -0
- package/dist/src/runs/query.js +1 -1
- package/dist/src/runs/query.js.map +1 -1
- package/dist/src/runs/records.js +3 -0
- package/dist/src/runs/records.js.map +1 -1
- package/dist/src/runs/request.js +4 -2
- package/dist/src/runs/request.js.map +1 -1
- package/dist/src/runs/verifier-evidence-redaction.js +61 -0
- package/dist/src/runs/verifier-evidence-redaction.js.map +1 -0
- package/dist/src/runs/verifier-evidence.js +459 -0
- package/dist/src/runs/verifier-evidence.js.map +1 -0
- package/dist/src/runs/verifier-structured-evidence.js +83 -0
- package/dist/src/runs/verifier-structured-evidence.js.map +1 -0
- package/dist/src/trajectories/analysis.js +288 -0
- package/dist/src/trajectories/analysis.js.map +1 -0
- package/dist/src/trajectories/chunk-projection.js +148 -0
- package/dist/src/trajectories/chunk-projection.js.map +1 -0
- package/dist/src/trajectories/content-projection.js +389 -0
- package/dist/src/trajectories/content-projection.js.map +1 -0
- package/dist/src/trajectories/dsh-chunk-contract.js +176 -0
- package/dist/src/trajectories/dsh-chunk-contract.js.map +1 -0
- package/dist/src/trajectories/dsh-contract.js +284 -0
- package/dist/src/trajectories/dsh-contract.js.map +1 -0
- package/dist/src/trajectories/events-chunk-drill.js +95 -0
- package/dist/src/trajectories/events-chunk-drill.js.map +1 -0
- package/dist/src/trajectories/events-page.js +367 -0
- package/dist/src/trajectories/events-page.js.map +1 -0
- package/dist/src/trajectories/index.js +3 -0
- package/dist/src/trajectories/index.js.map +1 -1
- package/dist/src/trajectories/request-attempt.js +47 -0
- package/dist/src/trajectories/request-attempt.js.map +1 -0
- package/dist/src/trajectories/stream-reader.js +311 -0
- package/dist/src/trajectories/stream-reader.js.map +1 -0
- package/dist/src/trajectories/surface-fold.js +126 -0
- package/dist/src/trajectories/surface-fold.js.map +1 -0
- package/dist/src/workers/remote-harbor-worker.js +3 -0
- package/dist/src/workers/remote-harbor-worker.js.map +1 -1
- package/docs/schemas/benchmark-hook-request-v1.schema.json +65 -0
- package/docs/schemas/benchmark-hook-response-v1.schema.json +75 -0
- package/docs/schemas/benchmark-hook-v1.schema.json +33 -0
- package/docs/schemas/benchmark-lock-v1.schema.json +351 -0
- package/docs/schemas/benchmark-metric-v1.schema.json +37 -0
- package/docs/schemas/benchmark-package-v1.schema.json +167 -0
- package/docs/schemas/benchmark-phase-group.schema.json +41 -0
- package/docs/schemas/benchmark-profile-v1.schema.json +141 -0
- package/docs/schemas/benchmark-task-v1.schema.json +384 -0
- package/docs/schemas/benchmark-tool-result-v1.schema.json +30 -0
- package/docs/schemas/error.schema.json +22 -0
- package/docs/schemas/eval-progress.schema.json +1 -15
- package/docs/schemas/eval-rerun-result.schema.json +6 -1
- package/docs/schemas/eval-rerun-submission.schema.json +3 -1
- package/docs/schemas/eval-result.schema.json +1 -15
- package/docs/schemas/eval-trial-publication.schema.json +1 -25
- package/docs/schemas/eval-trial-reference.schema.json +129 -0
- package/docs/schemas/regrade-assessment-reference.schema.json +11 -0
- package/docs/schemas/run-context.schema.json +15 -0
- package/docs/schemas/run-manifest.schema.json +4 -0
- package/docs/schemas/run-request.schema.json +4 -0
- package/docs/schemas/trajectory-analysis.schema.json +258 -0
- package/docs/schemas/trajectory-events-page.schema.json +40 -0
- package/docs/schemas/verifier-assessment.schema.json +77 -0
- package/docs/schemas/verifier-evidence.schema.json +276 -0
- package/integrations/harbor/hitch_benchmark.py +252 -0
- package/integrations/harbor/hitch_candidate_recycle.py +262 -0
- package/integrations/harbor/hitch_harbor_agent.py +541 -43
- package/integrations/harbor/hitch_harbor_environment.py +138 -0
- package/integrations/harbor/hitch_harbor_verifier.py +60 -2
- package/integrations/harbor/hitch_phase_supervisor.py +452 -0
- package/integrations/harbor/hitch_tool_client.mjs +91 -0
- package/integrations/model-call/cli.js +57 -0
- package/package.json +17 -6
|
@@ -0,0 +1,141 @@
|
|
|
1
|
+
{
|
|
2
|
+
"$schema": "https://json-schema.org/draft/2020-12/schema",
|
|
3
|
+
"$id": "https://hitch.dev/schemas/benchmark-profile-v1.schema.json",
|
|
4
|
+
"type": "object",
|
|
5
|
+
"properties": {
|
|
6
|
+
"schema_version": {
|
|
7
|
+
"const": "1"
|
|
8
|
+
},
|
|
9
|
+
"id": {
|
|
10
|
+
"type": "string",
|
|
11
|
+
"minLength": 1
|
|
12
|
+
},
|
|
13
|
+
"track": {
|
|
14
|
+
"enum": [
|
|
15
|
+
"custom",
|
|
16
|
+
"public-subset"
|
|
17
|
+
]
|
|
18
|
+
},
|
|
19
|
+
"input_mode": {
|
|
20
|
+
"const": "instruction"
|
|
21
|
+
},
|
|
22
|
+
"tool_policy": {
|
|
23
|
+
"type": "object",
|
|
24
|
+
"properties": {
|
|
25
|
+
"id": {
|
|
26
|
+
"type": "string",
|
|
27
|
+
"minLength": 1
|
|
28
|
+
},
|
|
29
|
+
"allowed": {
|
|
30
|
+
"type": "array",
|
|
31
|
+
"items": {
|
|
32
|
+
"type": "string",
|
|
33
|
+
"minLength": 1
|
|
34
|
+
},
|
|
35
|
+
"uniqueItems": true
|
|
36
|
+
},
|
|
37
|
+
"network": {
|
|
38
|
+
"const": "open"
|
|
39
|
+
},
|
|
40
|
+
"enforcement": {
|
|
41
|
+
"const": "required"
|
|
42
|
+
}
|
|
43
|
+
},
|
|
44
|
+
"required": [
|
|
45
|
+
"id",
|
|
46
|
+
"allowed",
|
|
47
|
+
"network",
|
|
48
|
+
"enforcement"
|
|
49
|
+
],
|
|
50
|
+
"additionalProperties": false
|
|
51
|
+
},
|
|
52
|
+
"budget": {
|
|
53
|
+
"type": "object",
|
|
54
|
+
"properties": {
|
|
55
|
+
"agent_timeout": {
|
|
56
|
+
"type": "object",
|
|
57
|
+
"properties": {
|
|
58
|
+
"source": {
|
|
59
|
+
"const": "task"
|
|
60
|
+
}
|
|
61
|
+
},
|
|
62
|
+
"required": [
|
|
63
|
+
"source"
|
|
64
|
+
],
|
|
65
|
+
"additionalProperties": false
|
|
66
|
+
},
|
|
67
|
+
"setup_timeout_ms": {
|
|
68
|
+
"type": "integer",
|
|
69
|
+
"minimum": 1
|
|
70
|
+
},
|
|
71
|
+
"collection_timeout_ms": {
|
|
72
|
+
"type": "integer",
|
|
73
|
+
"minimum": 1
|
|
74
|
+
},
|
|
75
|
+
"cleanup_grace_ms": {
|
|
76
|
+
"type": "integer",
|
|
77
|
+
"minimum": 1
|
|
78
|
+
}
|
|
79
|
+
},
|
|
80
|
+
"required": [
|
|
81
|
+
"agent_timeout",
|
|
82
|
+
"setup_timeout_ms",
|
|
83
|
+
"collection_timeout_ms",
|
|
84
|
+
"cleanup_grace_ms"
|
|
85
|
+
],
|
|
86
|
+
"additionalProperties": false
|
|
87
|
+
},
|
|
88
|
+
"sampling": {
|
|
89
|
+
"type": "object",
|
|
90
|
+
"properties": {
|
|
91
|
+
"attempts_per_task": {
|
|
92
|
+
"type": "integer",
|
|
93
|
+
"minimum": 1
|
|
94
|
+
},
|
|
95
|
+
"seed": {
|
|
96
|
+
"type": "integer"
|
|
97
|
+
}
|
|
98
|
+
},
|
|
99
|
+
"required": [
|
|
100
|
+
"attempts_per_task",
|
|
101
|
+
"seed"
|
|
102
|
+
],
|
|
103
|
+
"additionalProperties": false
|
|
104
|
+
},
|
|
105
|
+
"grading": {
|
|
106
|
+
"type": "object",
|
|
107
|
+
"properties": {
|
|
108
|
+
"on_agent_budget_exhausted": {
|
|
109
|
+
"const": "grade_final_state"
|
|
110
|
+
},
|
|
111
|
+
"on_missing_submission": {
|
|
112
|
+
"const": "error"
|
|
113
|
+
},
|
|
114
|
+
"infrastructure_retries": {
|
|
115
|
+
"type": "integer",
|
|
116
|
+
"minimum": 0
|
|
117
|
+
}
|
|
118
|
+
},
|
|
119
|
+
"required": [
|
|
120
|
+
"on_agent_budget_exhausted",
|
|
121
|
+
"on_missing_submission",
|
|
122
|
+
"infrastructure_retries"
|
|
123
|
+
],
|
|
124
|
+
"additionalProperties": false
|
|
125
|
+
},
|
|
126
|
+
"extensions": {
|
|
127
|
+
"type": "object"
|
|
128
|
+
}
|
|
129
|
+
},
|
|
130
|
+
"required": [
|
|
131
|
+
"schema_version",
|
|
132
|
+
"id",
|
|
133
|
+
"track",
|
|
134
|
+
"input_mode",
|
|
135
|
+
"tool_policy",
|
|
136
|
+
"budget",
|
|
137
|
+
"sampling",
|
|
138
|
+
"grading"
|
|
139
|
+
],
|
|
140
|
+
"additionalProperties": false
|
|
141
|
+
}
|
|
@@ -0,0 +1,384 @@
|
|
|
1
|
+
{
|
|
2
|
+
"$schema": "https://json-schema.org/draft/2020-12/schema",
|
|
3
|
+
"$id": "https://hitch.dev/schemas/benchmark-task-v1.schema.json",
|
|
4
|
+
"type": "object",
|
|
5
|
+
"properties": {
|
|
6
|
+
"schema_version": {
|
|
7
|
+
"const": "1"
|
|
8
|
+
},
|
|
9
|
+
"source_task_id": {
|
|
10
|
+
"type": "string",
|
|
11
|
+
"minLength": 1
|
|
12
|
+
},
|
|
13
|
+
"driver": {
|
|
14
|
+
"oneOf": [
|
|
15
|
+
{
|
|
16
|
+
"type": "object",
|
|
17
|
+
"properties": {
|
|
18
|
+
"kind": {
|
|
19
|
+
"const": "tool-server"
|
|
20
|
+
},
|
|
21
|
+
"protocol_version": {
|
|
22
|
+
"const": "1"
|
|
23
|
+
},
|
|
24
|
+
"config": {
|
|
25
|
+
"type": "object",
|
|
26
|
+
"properties": {
|
|
27
|
+
"transport": {
|
|
28
|
+
"const": "http-json-cli"
|
|
29
|
+
},
|
|
30
|
+
"endpoint": {
|
|
31
|
+
"type": "string",
|
|
32
|
+
"minLength": 1
|
|
33
|
+
},
|
|
34
|
+
"schema": {
|
|
35
|
+
"type": "string",
|
|
36
|
+
"minLength": 1
|
|
37
|
+
},
|
|
38
|
+
"service": {
|
|
39
|
+
"type": "string",
|
|
40
|
+
"minLength": 1
|
|
41
|
+
},
|
|
42
|
+
"native_phases": {
|
|
43
|
+
"type": "object",
|
|
44
|
+
"additionalProperties": false,
|
|
45
|
+
"required": [
|
|
46
|
+
"protocol",
|
|
47
|
+
"argv",
|
|
48
|
+
"audit_path",
|
|
49
|
+
"shutdown_timeout_ms"
|
|
50
|
+
],
|
|
51
|
+
"properties": {
|
|
52
|
+
"protocol": {
|
|
53
|
+
"enum": ["hitch-native-phase-control@1", "hitch-native-phase-control@2"]
|
|
54
|
+
},
|
|
55
|
+
"argv": {
|
|
56
|
+
"type": "array",
|
|
57
|
+
"minItems": 1,
|
|
58
|
+
"maxItems": 32,
|
|
59
|
+
"items": {
|
|
60
|
+
"type": "string",
|
|
61
|
+
"minLength": 1,
|
|
62
|
+
"maxLength": 8192
|
|
63
|
+
}
|
|
64
|
+
},
|
|
65
|
+
"audit_path": {
|
|
66
|
+
"type": "string",
|
|
67
|
+
"pattern": "^/"
|
|
68
|
+
},
|
|
69
|
+
"shutdown_timeout_ms": {
|
|
70
|
+
"type": "integer",
|
|
71
|
+
"minimum": 1,
|
|
72
|
+
"maximum": 600000
|
|
73
|
+
},
|
|
74
|
+
"finalization_timeout_ms": {
|
|
75
|
+
"type": "integer",
|
|
76
|
+
"minimum": 1
|
|
77
|
+
}
|
|
78
|
+
},
|
|
79
|
+
"if": { "properties": { "protocol": { "const": "hitch-native-phase-control@2" } } },
|
|
80
|
+
"then": { "required": ["finalization_timeout_ms"] },
|
|
81
|
+
"else": { "not": { "required": ["finalization_timeout_ms"] } }
|
|
82
|
+
}
|
|
83
|
+
},
|
|
84
|
+
"required": [
|
|
85
|
+
"transport",
|
|
86
|
+
"endpoint",
|
|
87
|
+
"schema",
|
|
88
|
+
"service"
|
|
89
|
+
],
|
|
90
|
+
"additionalProperties": false
|
|
91
|
+
}
|
|
92
|
+
},
|
|
93
|
+
"required": [
|
|
94
|
+
"kind",
|
|
95
|
+
"protocol_version",
|
|
96
|
+
"config"
|
|
97
|
+
],
|
|
98
|
+
"additionalProperties": false
|
|
99
|
+
},
|
|
100
|
+
{
|
|
101
|
+
"type": "object",
|
|
102
|
+
"properties": {
|
|
103
|
+
"kind": {
|
|
104
|
+
"const": "terminal"
|
|
105
|
+
},
|
|
106
|
+
"protocol_version": {
|
|
107
|
+
"const": "1"
|
|
108
|
+
},
|
|
109
|
+
"config": {
|
|
110
|
+
"type": "object",
|
|
111
|
+
"maxProperties": 0
|
|
112
|
+
}
|
|
113
|
+
},
|
|
114
|
+
"required": [
|
|
115
|
+
"kind",
|
|
116
|
+
"protocol_version",
|
|
117
|
+
"config"
|
|
118
|
+
],
|
|
119
|
+
"additionalProperties": false
|
|
120
|
+
},
|
|
121
|
+
{
|
|
122
|
+
"type": "object",
|
|
123
|
+
"properties": {
|
|
124
|
+
"kind": {
|
|
125
|
+
"const": "model-call"
|
|
126
|
+
},
|
|
127
|
+
"protocol_version": {
|
|
128
|
+
"const": "1"
|
|
129
|
+
},
|
|
130
|
+
"config": {
|
|
131
|
+
"type": "object",
|
|
132
|
+
"properties": {
|
|
133
|
+
"input": {
|
|
134
|
+
"type": "string",
|
|
135
|
+
"minLength": 1
|
|
136
|
+
}
|
|
137
|
+
},
|
|
138
|
+
"required": [
|
|
139
|
+
"input"
|
|
140
|
+
],
|
|
141
|
+
"additionalProperties": false
|
|
142
|
+
}
|
|
143
|
+
},
|
|
144
|
+
"required": [
|
|
145
|
+
"kind",
|
|
146
|
+
"protocol_version",
|
|
147
|
+
"config"
|
|
148
|
+
],
|
|
149
|
+
"additionalProperties": false
|
|
150
|
+
}
|
|
151
|
+
]
|
|
152
|
+
},
|
|
153
|
+
"requirements": {
|
|
154
|
+
"type": "array",
|
|
155
|
+
"items": {
|
|
156
|
+
"type": "string",
|
|
157
|
+
"minLength": 1
|
|
158
|
+
},
|
|
159
|
+
"uniqueItems": true
|
|
160
|
+
},
|
|
161
|
+
"lifecycle": {
|
|
162
|
+
"type": "object",
|
|
163
|
+
"properties": {
|
|
164
|
+
"prepare": {
|
|
165
|
+
"$ref": "benchmark-hook-v1.schema.json"
|
|
166
|
+
},
|
|
167
|
+
"quiesce": {
|
|
168
|
+
"$ref": "benchmark-hook-v1.schema.json"
|
|
169
|
+
},
|
|
170
|
+
"snapshot": {
|
|
171
|
+
"$ref": "benchmark-hook-v1.schema.json"
|
|
172
|
+
},
|
|
173
|
+
"cleanup": {
|
|
174
|
+
"$ref": "benchmark-hook-v1.schema.json"
|
|
175
|
+
}
|
|
176
|
+
},
|
|
177
|
+
"additionalProperties": false
|
|
178
|
+
},
|
|
179
|
+
"submission": {
|
|
180
|
+
"type": "object",
|
|
181
|
+
"properties": {
|
|
182
|
+
"kind": {
|
|
183
|
+
"enum": [
|
|
184
|
+
"artifacts",
|
|
185
|
+
"environment"
|
|
186
|
+
]
|
|
187
|
+
},
|
|
188
|
+
"paths": {
|
|
189
|
+
"type": "array",
|
|
190
|
+
"items": {
|
|
191
|
+
"type": "string",
|
|
192
|
+
"minLength": 1
|
|
193
|
+
},
|
|
194
|
+
"minItems": 0,
|
|
195
|
+
"uniqueItems": true
|
|
196
|
+
},
|
|
197
|
+
"max_bytes": {
|
|
198
|
+
"type": "integer",
|
|
199
|
+
"minimum": 1
|
|
200
|
+
},
|
|
201
|
+
"final_response": {
|
|
202
|
+
"const": "/hitch-evidence/final-response.json"
|
|
203
|
+
}
|
|
204
|
+
},
|
|
205
|
+
"required": [
|
|
206
|
+
"kind",
|
|
207
|
+
"paths",
|
|
208
|
+
"max_bytes"
|
|
209
|
+
],
|
|
210
|
+
"additionalProperties": false
|
|
211
|
+
},
|
|
212
|
+
"grading": {
|
|
213
|
+
"type": "object",
|
|
214
|
+
"properties": {
|
|
215
|
+
"kind": {
|
|
216
|
+
"enum": [
|
|
217
|
+
"command",
|
|
218
|
+
"harbor"
|
|
219
|
+
]
|
|
220
|
+
},
|
|
221
|
+
"entrypoint": {
|
|
222
|
+
"const": [
|
|
223
|
+
"bash",
|
|
224
|
+
"/tests/test.sh"
|
|
225
|
+
]
|
|
226
|
+
},
|
|
227
|
+
"metric_map": {
|
|
228
|
+
"type": "object",
|
|
229
|
+
"minProperties": 1,
|
|
230
|
+
"additionalProperties": {
|
|
231
|
+
"type": "string",
|
|
232
|
+
"minLength": 1
|
|
233
|
+
}
|
|
234
|
+
}
|
|
235
|
+
},
|
|
236
|
+
"required": [
|
|
237
|
+
"kind",
|
|
238
|
+
"entrypoint",
|
|
239
|
+
"metric_map"
|
|
240
|
+
],
|
|
241
|
+
"additionalProperties": false
|
|
242
|
+
},
|
|
243
|
+
"extensions": {
|
|
244
|
+
"type": "object"
|
|
245
|
+
}
|
|
246
|
+
},
|
|
247
|
+
"required": [
|
|
248
|
+
"schema_version",
|
|
249
|
+
"source_task_id",
|
|
250
|
+
"driver",
|
|
251
|
+
"requirements",
|
|
252
|
+
"lifecycle",
|
|
253
|
+
"submission",
|
|
254
|
+
"grading"
|
|
255
|
+
],
|
|
256
|
+
"additionalProperties": false,
|
|
257
|
+
"allOf": [
|
|
258
|
+
{
|
|
259
|
+
"if": {
|
|
260
|
+
"properties": {
|
|
261
|
+
"driver": {
|
|
262
|
+
"properties": {
|
|
263
|
+
"kind": {
|
|
264
|
+
"const": "tool-server"
|
|
265
|
+
}
|
|
266
|
+
}
|
|
267
|
+
}
|
|
268
|
+
}
|
|
269
|
+
},
|
|
270
|
+
"then": {
|
|
271
|
+
"properties": {
|
|
272
|
+
"lifecycle": {
|
|
273
|
+
"required": [
|
|
274
|
+
"prepare",
|
|
275
|
+
"quiesce",
|
|
276
|
+
"snapshot",
|
|
277
|
+
"cleanup"
|
|
278
|
+
]
|
|
279
|
+
},
|
|
280
|
+
"submission": {
|
|
281
|
+
"properties": {
|
|
282
|
+
"kind": {
|
|
283
|
+
"const": "artifacts"
|
|
284
|
+
},
|
|
285
|
+
"paths": {
|
|
286
|
+
"minItems": 1
|
|
287
|
+
}
|
|
288
|
+
}
|
|
289
|
+
},
|
|
290
|
+
"grading": {
|
|
291
|
+
"properties": {
|
|
292
|
+
"kind": {
|
|
293
|
+
"const": "command"
|
|
294
|
+
}
|
|
295
|
+
}
|
|
296
|
+
}
|
|
297
|
+
}
|
|
298
|
+
},
|
|
299
|
+
"else": {
|
|
300
|
+
"properties": {
|
|
301
|
+
"lifecycle": {
|
|
302
|
+
"maxProperties": 0
|
|
303
|
+
}
|
|
304
|
+
}
|
|
305
|
+
}
|
|
306
|
+
},
|
|
307
|
+
{
|
|
308
|
+
"if": {
|
|
309
|
+
"properties": {
|
|
310
|
+
"driver": {
|
|
311
|
+
"properties": {
|
|
312
|
+
"kind": {
|
|
313
|
+
"const": "model-call"
|
|
314
|
+
}
|
|
315
|
+
}
|
|
316
|
+
}
|
|
317
|
+
}
|
|
318
|
+
},
|
|
319
|
+
"then": {
|
|
320
|
+
"properties": {
|
|
321
|
+
"submission": {
|
|
322
|
+
"required": [
|
|
323
|
+
"final_response"
|
|
324
|
+
],
|
|
325
|
+
"properties": {
|
|
326
|
+
"kind": {
|
|
327
|
+
"const": "artifacts"
|
|
328
|
+
}
|
|
329
|
+
}
|
|
330
|
+
},
|
|
331
|
+
"grading": {
|
|
332
|
+
"properties": {
|
|
333
|
+
"kind": {
|
|
334
|
+
"const": "command"
|
|
335
|
+
}
|
|
336
|
+
}
|
|
337
|
+
}
|
|
338
|
+
}
|
|
339
|
+
}
|
|
340
|
+
},
|
|
341
|
+
{
|
|
342
|
+
"if": {
|
|
343
|
+
"properties": {
|
|
344
|
+
"submission": {
|
|
345
|
+
"properties": {
|
|
346
|
+
"kind": {
|
|
347
|
+
"const": "artifacts"
|
|
348
|
+
}
|
|
349
|
+
}
|
|
350
|
+
}
|
|
351
|
+
}
|
|
352
|
+
},
|
|
353
|
+
"then": {
|
|
354
|
+
"properties": {
|
|
355
|
+
"submission": {
|
|
356
|
+
"properties": {
|
|
357
|
+
"paths": {
|
|
358
|
+
"minItems": 1
|
|
359
|
+
}
|
|
360
|
+
}
|
|
361
|
+
}
|
|
362
|
+
}
|
|
363
|
+
},
|
|
364
|
+
"else": {
|
|
365
|
+
"properties": {
|
|
366
|
+
"driver": {
|
|
367
|
+
"properties": {
|
|
368
|
+
"kind": {
|
|
369
|
+
"const": "terminal"
|
|
370
|
+
}
|
|
371
|
+
}
|
|
372
|
+
},
|
|
373
|
+
"submission": {
|
|
374
|
+
"properties": {
|
|
375
|
+
"paths": {
|
|
376
|
+
"maxItems": 0
|
|
377
|
+
}
|
|
378
|
+
}
|
|
379
|
+
}
|
|
380
|
+
}
|
|
381
|
+
}
|
|
382
|
+
}
|
|
383
|
+
]
|
|
384
|
+
}
|
|
@@ -0,0 +1,30 @@
|
|
|
1
|
+
{
|
|
2
|
+
"$schema": "https://json-schema.org/draft/2020-12/schema",
|
|
3
|
+
"$id": "https://github.com/rsi-gear/agent-hitch/schemas/benchmark-tool-result-v1.schema.json",
|
|
4
|
+
"title": "Native image tool result envelope",
|
|
5
|
+
"type": "object",
|
|
6
|
+
"required": ["protocol", "content"],
|
|
7
|
+
"properties": {
|
|
8
|
+
"protocol": {"const": "hitch-tool-result@1"},
|
|
9
|
+
"content": {
|
|
10
|
+
"type": "array",
|
|
11
|
+
"maxItems": 64,
|
|
12
|
+
"contains": {"type": "object", "required": ["type"], "properties": {"type": {"const": "image"}}},
|
|
13
|
+
"minContains": 0,
|
|
14
|
+
"maxContains": 8,
|
|
15
|
+
"items": {
|
|
16
|
+
"oneOf": [
|
|
17
|
+
{"type": "object", "required": ["type", "text"], "properties": {
|
|
18
|
+
"type": {"const": "text"}, "text": {"type": "string"}
|
|
19
|
+
}},
|
|
20
|
+
{"type": "object", "required": ["type", "mimeType", "data"], "properties": {
|
|
21
|
+
"type": {"const": "image"},
|
|
22
|
+
"mimeType": {"enum": ["image/png", "image/jpeg", "image/webp"]},
|
|
23
|
+
"data": {"type": "string", "minLength": 4, "maxLength": 5592408, "contentEncoding": "base64"}
|
|
24
|
+
}}
|
|
25
|
+
]
|
|
26
|
+
}
|
|
27
|
+
}
|
|
28
|
+
},
|
|
29
|
+
"$comment": "Wire payload is limited to 16 MiB. Runtime additionally verifies canonical base64, MIME signatures and a 4 MiB decoded image limit. The CLI replaces each image data field with a generated absolute path, byte count and SHA256; this schema describes the server's wire response, not that materialized CLI output."
|
|
30
|
+
}
|
|
@@ -0,0 +1,22 @@
|
|
|
1
|
+
{
|
|
2
|
+
"$schema": "https://json-schema.org/draft/2020-12/schema",
|
|
3
|
+
"$id": "https://agent-hitch.local/schemas/error.schema.json",
|
|
4
|
+
"title": "Hitch CLI error envelope",
|
|
5
|
+
"type": "object",
|
|
6
|
+
"additionalProperties": false,
|
|
7
|
+
"required": ["schema_version", "kind", "error"],
|
|
8
|
+
"properties": {
|
|
9
|
+
"schema_version": { "const": "1" },
|
|
10
|
+
"kind": { "const": "error" },
|
|
11
|
+
"error": {
|
|
12
|
+
"type": "object",
|
|
13
|
+
"additionalProperties": false,
|
|
14
|
+
"required": ["code", "message", "exit_code"],
|
|
15
|
+
"properties": {
|
|
16
|
+
"code": { "type": "string", "pattern": "^[a-z][a-z0-9_-]*$", "maxLength": 128 },
|
|
17
|
+
"message": { "type": "string" },
|
|
18
|
+
"exit_code": { "type": "integer", "minimum": 1, "maximum": 255 }
|
|
19
|
+
}
|
|
20
|
+
}
|
|
21
|
+
}
|
|
22
|
+
}
|
|
@@ -29,21 +29,7 @@
|
|
|
29
29
|
"planned_trials": { "type": ["integer", "null"], "minimum": 0 },
|
|
30
30
|
"trials": {
|
|
31
31
|
"type": "array",
|
|
32
|
-
"items": {
|
|
33
|
-
"type": "object",
|
|
34
|
-
"additionalProperties": false,
|
|
35
|
-
"required": ["trial_id", "run_id", "task_id", "attempt", "observation_status"],
|
|
36
|
-
"properties": {
|
|
37
|
-
"trial_id": { "type": "string", "minLength": 1 },
|
|
38
|
-
"run_id": { "type": "string", "pattern": "^run_[a-f0-9]{32}$" },
|
|
39
|
-
"task_id": { "type": "string", "minLength": 1 },
|
|
40
|
-
"attempt": { "type": "integer", "minimum": 1 },
|
|
41
|
-
"observation_status": { "enum": ["valid", "invalid"] },
|
|
42
|
-
"reward": { "type": "number" },
|
|
43
|
-
"verifier_result_ref": { "type": "string", "minLength": 1 },
|
|
44
|
-
"invalid_reason": { "type": "string", "minLength": 1 }
|
|
45
|
-
}
|
|
46
|
-
}
|
|
32
|
+
"items": { "$ref": "eval-trial-reference.schema.json" }
|
|
47
33
|
},
|
|
48
34
|
"summary": {
|
|
49
35
|
"type": "object",
|
|
@@ -41,10 +41,15 @@
|
|
|
41
41
|
"source": {
|
|
42
42
|
"type": "object",
|
|
43
43
|
"additionalProperties": false,
|
|
44
|
-
"required": ["source_trial_id", "
|
|
44
|
+
"required": ["source_trial_id", "source_work_id", "source_backend_directory"],
|
|
45
|
+
"oneOf": [
|
|
46
|
+
{ "required": ["source_run_id"], "not": { "required": ["source_run_group_id"] } },
|
|
47
|
+
{ "required": ["source_run_group_id"], "not": { "required": ["source_run_id"] } }
|
|
48
|
+
],
|
|
45
49
|
"properties": {
|
|
46
50
|
"source_trial_id": { "type": "string", "minLength": 1 },
|
|
47
51
|
"source_run_id": { "type": "string", "pattern": "^run_[a-f0-9]{32}$" },
|
|
52
|
+
"source_run_group_id": { "type": "string", "pattern": "^run_group_[a-f0-9]{32}$" },
|
|
48
53
|
"source_work_id": { "type": "string", "pattern": "^work_[a-f0-9]{32}$" },
|
|
49
54
|
"source_backend_directory": { "type": "string", "pattern": "^harbor/work-items/work_[a-f0-9]{32}/epoch-[0-9]{6}$" }
|
|
50
55
|
}
|
|
@@ -10,10 +10,12 @@
|
|
|
10
10
|
"rerun_id": { "type": "string", "pattern": "^rerun_[a-f0-9]{32}$" },
|
|
11
11
|
"eval_id": { "type": "string", "pattern": "^eval_[a-f0-9]{32}$" },
|
|
12
12
|
"rerun_type": { "$ref": "#/$defs/rerunType" },
|
|
13
|
+
"verifier_runtime_id": { "type": "string", "pattern": "^sha256:[a-f0-9]{64}$" },
|
|
13
14
|
"semantics": { "$ref": "#/$defs/semantics" },
|
|
14
15
|
"selector": { "$ref": "#/$defs/selector" },
|
|
15
16
|
"submitted_at": { "type": "string", "format": "date-time" }
|
|
16
17
|
},
|
|
18
|
+
"allOf": [{ "if": { "required": ["verifier_runtime_id"] }, "then": { "properties": { "rerun_type": { "const": "verifier-only" } } } }],
|
|
17
19
|
"$defs": {
|
|
18
20
|
"rerunType": {
|
|
19
21
|
"enum": ["candidate-restart", "candidate-resume", "trajectory-replay", "verifier-only", "collect-only"]
|
|
@@ -25,7 +27,7 @@
|
|
|
25
27
|
"properties": {
|
|
26
28
|
"candidate_action": { "enum": ["restart", "resume", "replay", "none"] },
|
|
27
29
|
"conversation_source": { "enum": ["original-instruction", "native-session", "canonical-trajectory", "none"] },
|
|
28
|
-
"sandbox_source": { "enum": ["clean", "checkpoint", "retained", "none"] },
|
|
30
|
+
"sandbox_source": { "enum": ["clean", "checkpoint", "retained", "recorded-artifacts", "none"] },
|
|
29
31
|
"candidate_executes": { "type": "boolean" }
|
|
30
32
|
}
|
|
31
33
|
},
|
|
@@ -76,21 +76,7 @@
|
|
|
76
76
|
},
|
|
77
77
|
"trials": {
|
|
78
78
|
"type": "array",
|
|
79
|
-
"items": {
|
|
80
|
-
"type": "object",
|
|
81
|
-
"additionalProperties": false,
|
|
82
|
-
"required": ["trial_id", "run_id", "task_id", "attempt", "observation_status"],
|
|
83
|
-
"properties": {
|
|
84
|
-
"trial_id": { "type": "string", "minLength": 1 },
|
|
85
|
-
"run_id": { "type": "string", "pattern": "^run_[a-f0-9]{32}$" },
|
|
86
|
-
"task_id": { "type": "string", "minLength": 1 },
|
|
87
|
-
"attempt": { "type": "integer", "minimum": 1 },
|
|
88
|
-
"observation_status": { "enum": ["valid", "invalid"] },
|
|
89
|
-
"reward": { "type": "number" },
|
|
90
|
-
"verifier_result_ref": { "type": "string", "minLength": 1 },
|
|
91
|
-
"invalid_reason": { "type": "string", "minLength": 1 }
|
|
92
|
-
}
|
|
93
|
-
}
|
|
79
|
+
"items": { "$ref": "eval-trial-reference.schema.json" }
|
|
94
80
|
},
|
|
95
81
|
"local_source_transport": {
|
|
96
82
|
"type": "object",
|