pi-harness-runtime 1.1.439 → 1.1.441
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/harness/e2e/auto-test-loop.ts +8 -2
- package/harness/e2e/browser-scrapers/base-browser-scraper.ts +2 -0
- package/harness/e2e/browser-scrapers/chrome-profile-manager.ts +2 -0
- package/harness/e2e/browser-scrapers/codex-browser-scraper.ts +2 -0
- package/harness/e2e/tools-detector.ts +2 -0
- package/harness/index.ts +2 -1
- package/harness/key-loader.ts +2 -0
- package/harness/langchain/checkpointer.ts +2 -0
- package/harness/langchain/cron-cli.ts +2 -0
- package/harness/langchain/graph.ts +2 -0
- package/harness/master-planner.ts +1 -1
- package/harness/notification-center.ts +2 -0
- package/harness/output-limit-handler.ts +1 -1
- package/harness/project-detector/detector.ts +2 -4
- package/harness/secret-provider.ts +2 -0
- package/harness/task-graph.ts +460 -113
- package/index.ts +16 -7
- package/package.json +7 -1
- package/packages/a2a-adapter/package.json +1 -1
- package/packages/agent-policy/dist/extension.d.ts.map +1 -1
- package/packages/agent-policy/dist/extension.js +2 -0
- package/packages/agent-policy/dist/extension.js.map +1 -1
- package/packages/agent-policy/package.json +1 -1
- package/packages/agent-policy/src/extension.ts +2 -0
- package/packages/architecture-generator/package.json +1 -1
- package/packages/auth/dist/minimax-browser-auth.d.ts.map +1 -1
- package/packages/auth/dist/minimax-browser-auth.js +23 -1
- package/packages/auth/dist/minimax-browser-auth.js.map +1 -1
- package/packages/auth/package.json +1 -1
- package/packages/auth/src/minimax-browser-auth.ts +30 -5
- package/packages/autonomous-refactor/package.json +1 -1
- package/packages/autonomous-runtime/dist/inbox.d.ts.map +1 -1
- package/packages/autonomous-runtime/dist/inbox.js +2 -0
- package/packages/autonomous-runtime/dist/inbox.js.map +1 -1
- package/packages/autonomous-runtime/package.json +1 -1
- package/packages/autonomous-runtime/src/inbox.ts +2 -0
- package/packages/cache-strategy/package.json +1 -1
- package/packages/capability-registry/package.json +1 -1
- package/packages/checkpoint/package.json +1 -1
- package/packages/cli-plugin-sdk/package.json +1 -1
- package/packages/clipboard/dist/src/clipboard.d.ts.map +1 -1
- package/packages/clipboard/dist/src/clipboard.js +2 -0
- package/packages/clipboard/dist/src/clipboard.js.map +1 -1
- package/packages/clipboard/dist/src/copy-sync.d.ts.map +1 -1
- package/packages/clipboard/dist/src/copy-sync.js +2 -0
- package/packages/clipboard/dist/src/copy-sync.js.map +1 -1
- package/packages/clipboard/dist/src/gist-relay.d.ts.map +1 -1
- package/packages/clipboard/dist/src/gist-relay.js +2 -0
- package/packages/clipboard/dist/src/gist-relay.js.map +1 -1
- package/packages/clipboard/package.json +1 -1
- package/packages/clipboard/src/clipboard.ts +2 -0
- package/packages/clipboard/src/copy-sync.ts +2 -0
- package/packages/clipboard/src/gist-relay.ts +2 -0
- package/packages/clipboard-plugin/dist/types.d.ts +9 -2
- package/packages/clipboard-plugin/dist/types.d.ts.map +1 -1
- package/packages/clipboard-plugin/package.json +1 -1
- package/packages/clipboard-plugin/src/types.ts +9 -2
- package/packages/code-generation/package.json +1 -1
- package/packages/code-review/package.json +1 -1
- package/packages/codex-adapter/package.json +1 -1
- package/packages/codex-watcher/package.json +1 -1
- package/packages/config-capture/package.json +1 -1
- package/packages/connection-runtime/package.json +1 -1
- package/packages/context-compiler/package.json +1 -1
- package/packages/context-discovery/package.json +1 -1
- package/packages/context-manager/package.json +1 -1
- package/packages/contracts/dist/delegation-store.js +1 -4
- package/packages/contracts/dist/workspace-env.js +28 -1
- package/packages/contracts/package.json +1 -1
- package/packages/contracts/src/delegation-store.ts +9 -4
- package/packages/contracts/src/workspace-env.ts +29 -4
- package/packages/cookie-sanitizer/package.json +1 -1
- package/packages/cost-optimizer/package.json +1 -1
- package/packages/dependency-analyzer/package.json +1 -1
- package/packages/django-plugin/package.json +1 -1
- package/packages/doc-generator/package.json +1 -1
- package/packages/durable-review/package.json +1 -1
- package/packages/e2e-tester/package.json +1 -1
- package/packages/evaluation-engine/package.json +1 -1
- package/packages/evaluation-runner/package.json +1 -1
- package/packages/event-bus/package.json +1 -1
- package/packages/event-sink/package.json +1 -1
- package/packages/event-store/dist/src/store.d.ts.map +1 -1
- package/packages/event-store/dist/src/store.js +2 -0
- package/packages/event-store/dist/src/store.js.map +1 -1
- package/packages/event-store/package.json +1 -1
- package/packages/event-store/src/store.ts +3 -0
- package/packages/experience-replay/package.json +1 -1
- package/packages/feedback-collector/package.json +1 -1
- package/packages/file-copy-helper/package.json +1 -1
- package/packages/framework-detector/package.json +1 -1
- package/packages/framework-plugin-sdk/package.json +1 -1
- package/packages/frappe-plugin/package.json +1 -1
- package/packages/generic-web-plugin/dist/analyzer.d.ts.map +1 -1
- package/packages/generic-web-plugin/dist/analyzer.js +2 -0
- package/packages/generic-web-plugin/dist/analyzer.js.map +1 -1
- package/packages/generic-web-plugin/package.json +1 -1
- package/packages/generic-web-plugin/src/analyzer.ts +3 -0
- package/packages/health-monitor/package.json +1 -1
- package/packages/intent-analyzer/package.json +1 -1
- package/packages/jev-judge/package.json +1 -1
- package/packages/knowledge-graph/package.json +1 -1
- package/packages/knowledge-retrieval/dist/mcp-client.d.ts.map +1 -1
- package/packages/knowledge-retrieval/dist/mcp-client.js +0 -1
- package/packages/knowledge-retrieval/dist/mcp-client.js.map +1 -1
- package/packages/knowledge-retrieval/package.json +1 -4
- package/packages/knowledge-retrieval/src/mcp-client.ts +0 -1
- package/packages/laravel-plugin/package.json +1 -1
- package/packages/learning-engine/package.json +1 -1
- package/packages/logging/package.json +1 -1
- package/packages/mcp-adapter/dist/server.d.ts.map +1 -1
- package/packages/mcp-adapter/dist/server.js +2 -0
- package/packages/mcp-adapter/dist/server.js.map +1 -1
- package/packages/mcp-adapter/package.json +1 -1
- package/packages/mcp-adapter/src/server.ts +3 -0
- package/packages/memory-engine/package.json +1 -1
- package/packages/milestone-manager/package.json +1 -1
- package/packages/model-registry/package.json +1 -1
- package/packages/nextjs-plugin/package.json +1 -1
- package/packages/notification/package.json +1 -1
- package/packages/observability/package.json +1 -1
- package/packages/okf-indexer/package.json +1 -1
- package/packages/paperclip-integration/dist/index.d.ts.map +1 -1
- package/packages/paperclip-integration/dist/index.js +2 -0
- package/packages/paperclip-integration/dist/index.js.map +1 -1
- package/packages/paperclip-integration/package.json +1 -1
- package/packages/paperclip-integration/src/index.ts +2 -0
- package/packages/performance-optimizer/package.json +1 -1
- package/packages/permission-mode/package.json +1 -1
- package/packages/privilege-broker/dist/audit.d.ts.map +1 -1
- package/packages/privilege-broker/dist/audit.js +2 -0
- package/packages/privilege-broker/dist/audit.js.map +1 -1
- package/packages/privilege-broker/package.json +1 -1
- package/packages/privilege-broker/src/audit.ts +2 -0
- package/packages/project-analyzer/package.json +1 -1
- package/packages/project-bootstrap/package.json +1 -1
- package/packages/projection-engine/package.json +1 -1
- package/packages/prompt-compiler/dist/compiler.d.ts.map +1 -1
- package/packages/prompt-compiler/dist/compiler.js +2 -0
- package/packages/prompt-compiler/dist/compiler.js.map +1 -1
- package/packages/prompt-compiler/package.json +1 -1
- package/packages/prompt-compiler/src/compiler.ts +2 -0
- package/packages/prompt-versioning/package.json +1 -1
- package/packages/provider-adapter-sdk/package.json +1 -1
- package/packages/provider-router/package.json +1 -1
- package/packages/provider-selector/package.json +1 -1
- package/packages/providers/package.json +1 -1
- package/packages/qdrant-skills/package.json +1 -1
- package/packages/quota-manager/package.json +1 -1
- package/packages/rate-limiter/package.json +1 -1
- package/packages/react-vite-plugin/package.json +1 -1
- package/packages/relay/package.json +1 -1
- package/packages/release-manager/package.json +1 -1
- package/packages/requirement-compiler/package.json +1 -1
- package/packages/runtime/package.json +1 -1
- package/packages/scheduler/package.json +1 -1
- package/packages/scheduler-adapter/package.json +1 -1
- package/packages/session/package.json +1 -1
- package/packages/session-api/package.json +1 -1
- package/packages/session-export/package.json +1 -1
- package/packages/shared-context/package.json +1 -1
- package/packages/skill-mcp-client/dist/client.d.ts.map +1 -1
- package/packages/skill-mcp-client/dist/client.js +0 -1
- package/packages/skill-mcp-client/dist/client.js.map +1 -1
- package/packages/skill-mcp-client/package.json +1 -4
- package/packages/skill-mcp-client/src/client.ts +0 -1
- package/packages/skill-registry/package.json +1 -1
- package/packages/skills/dist/parser.d.ts.map +1 -1
- package/packages/skills/dist/parser.js +2 -0
- package/packages/skills/dist/parser.js.map +1 -1
- package/packages/skills/dist/skill-versioning.d.ts.map +1 -1
- package/packages/skills/dist/skill-versioning.js +2 -0
- package/packages/skills/dist/skill-versioning.js.map +1 -1
- package/packages/skills/package.json +1 -1
- package/packages/skills/src/parser.ts +3 -0
- package/packages/skills/src/skill-versioning.ts +2 -0
- package/packages/sprint-planner/package.json +1 -1
- package/packages/subscription-engine/package.json +1 -1
- package/packages/task-compiler/package.json +1 -1
- package/packages/tencentdb-memory/package.json +1 -1
- package/packages/tencentdb-sync/package.json +1 -1
- package/packages/test-data-generator/package.json +1 -1
- package/packages/test-generator/package.json +1 -1
- package/packages/todo-bd-sync/package.json +1 -1
- package/packages/token-estimation/package.json +1 -1
- package/packages/token-optimizer/package.json +1 -1
- package/packages/tooling-preference/package.json +1 -1
- package/packages/trajectory/package.json +1 -1
- package/packages/tui/package.json +1 -1
- package/packages/two-phase-review/package.json +1 -1
- package/packages/types/package.json +1 -1
- package/packages/voice-api/package.json +1 -1
- package/packages/webmcp-tester/dist/index.js +0 -2
- package/packages/webmcp-tester/package.json +1 -1
- package/packages/webmcp-tester/src/index.ts +0 -3
- package/packages/wiki/package.json +1 -1
- package/packages/workflow-events/package.json +1 -1
- package/packages/workspace-scanner/package.json +1 -1
- package/packages/worktree/package.json +1 -1
- package/packages/write-review/dist/trigger.d.ts.map +1 -1
- package/packages/write-review/dist/trigger.js +1 -2
- package/packages/write-review/dist/trigger.js.map +1 -1
- package/packages/write-review/package.json +1 -1
- package/packages/write-review/src/trigger.ts +1 -2
- package/scripts/lib/runtime-imports.ts +189 -3
- package/scripts/release-all.ts +5 -3
- package/scripts/telegram-webhook-server.ts +2 -0
- package/scripts/typecheck-packages.mjs +7 -4
- package/scripts/verify-package-contents.ts +161 -2
- package/harness/e2e/tools-detector.test.ts +0 -156
- package/harness/glm-quota-countdown.test.ts +0 -323
- package/harness/langchain/agents.test.ts +0 -39
- package/harness/langchain/checkpointer.test.ts +0 -243
- package/harness/langchain/daemon.test.ts +0 -459
- package/harness/langchain/model-router.test.ts +0 -191
- package/harness/langchain/ping-pong-middleware.test.ts +0 -69
- package/harness/langchain/scoreboard.test.ts +0 -245
- package/harness/langchain/status-line.test.ts +0 -106
- package/harness/langchain/surge.test.ts +0 -467
- package/harness/langchain/widget.test.ts +0 -251
- package/harness/skill-sync.test.ts +0 -57
- package/harness/ssh-detach-interceptor.test.ts +0 -263
- package/harness/ssh-hang-recovery.test.ts +0 -86
- package/packages/a2a-adapter/test/a2a.test.ts +0 -151
- package/packages/agent-policy/test/interlock.test.ts +0 -101
- package/packages/agent-policy/test/policy-retention.test.ts +0 -380
- package/packages/agent-policy/test/policy.test.ts +0 -387
- package/packages/auth/test/auth.test.ts +0 -238
- package/packages/autonomous-refactor/test/refactor.test.ts +0 -202
- package/packages/autonomous-runtime/test/types.test.ts +0 -95
- package/packages/cache-strategy/test/cache.test.ts +0 -137
- package/packages/capability-registry/test/registry.test.ts +0 -204
- package/packages/checkpoint/test/engine.test.ts +0 -265
- package/packages/cli-plugin-sdk/test/cli.test.ts +0 -45
- package/packages/code-generation/test/generator.test.ts +0 -238
- package/packages/code-review/test/engine.test.ts +0 -302
- package/packages/codex-adapter/test/provider.test.ts +0 -142
- package/packages/codex-watcher/test/watch.test.ts +0 -453
- package/packages/context-compiler/test/context-compiler.test.ts +0 -746
- package/packages/context-manager/test/context.test.ts +0 -198
- package/packages/contracts/test/delegation.test.ts +0 -350
- package/packages/contracts/test/workspace-env.test.ts +0 -802
- package/packages/cookie-sanitizer/test/infer.test.ts +0 -110
- package/packages/cookie-sanitizer/test/normalize.test.ts +0 -113
- package/packages/cookie-sanitizer/test/parse.test.ts +0 -204
- package/packages/cookie-sanitizer/test/sync.test.ts +0 -237
- package/packages/cookie-sanitizer/test/watcher.test.ts +0 -27
- package/packages/cost-optimizer/test/cost-optimizer.test.ts +0 -438
- package/packages/dependency-analyzer/test/analyzer.test.ts +0 -49
- package/packages/django-plugin/test/analyzer.test.ts +0 -195
- package/packages/doc-generator/test/docs.test.ts +0 -168
- package/packages/durable-review/test/acceptance.test.ts +0 -357
- package/packages/durable-review/test/smoke.test.ts +0 -107
- package/packages/evaluation-engine/test/engine.test.ts +0 -277
- package/packages/evaluation-runner/test/runner.test.ts +0 -168
- package/packages/event-bus/test/bus.test.ts +0 -100
- package/packages/event-store/test/store.test.ts +0 -115
- package/packages/experience-replay/test/replay.test.ts +0 -267
- package/packages/feedback-collector/test/collector.test.ts +0 -142
- package/packages/framework-detector/test/detector.test.ts +0 -75
- package/packages/framework-plugin-sdk/test/manager.test.ts +0 -320
- package/packages/frappe-plugin/test/analyzer.test.ts +0 -250
- package/packages/generic-web-plugin/test/analyzer.test.ts +0 -134
- package/packages/health-monitor/test/health.test.ts +0 -231
- package/packages/intent-analyzer/test/analyzer.test.ts +0 -72
- package/packages/intent-analyzer/test/ping-pong.test.ts +0 -96
- package/packages/jev-judge/tests/auto-continue-context.test.ts +0 -280
- package/packages/jev-judge/tests/continuation-controller.test.ts +0 -1396
- package/packages/jev-judge/tests/jev-provider-chain.test.ts +0 -536
- package/packages/jev-judge/tests/outcome-application.test.ts +0 -132
- package/packages/jev-judge/tests/question-risk.test.ts +0 -129
- package/packages/jev-judge/tests/typesafe-provider.test.ts +0 -640
- package/packages/jev-judge/tests/validation-ledger.test.ts +0 -181
- package/packages/jev-judge/tests/validation-plan.test.ts +0 -170
- package/packages/jev-judge/tests/validation-tie-breaker.test.ts +0 -88
- package/packages/laravel-plugin/test/analyzer.test.ts +0 -220
- package/packages/learning-engine/test/engine.test.ts +0 -331
- package/packages/logging/test/buffer.test.ts +0 -44
- package/packages/logging/test/classification.test.ts +0 -137
- package/packages/logging/test/config.test.ts +0 -97
- package/packages/logging/test/fanout.test.ts +0 -186
- package/packages/logging/test/identity-transport.test.ts +0 -197
- package/packages/logging/test/incidents.test.ts +0 -404
- package/packages/logging/test/ingest-schema.test.ts +0 -184
- package/packages/logging/test/integration.test.ts +0 -140
- package/packages/logging/test/log-http.test.ts +0 -246
- package/packages/logging/test/redact.test.ts +0 -109
- package/packages/logging/test/sampler.test.ts +0 -57
- package/packages/logging/test/server.test.ts +0 -444
- package/packages/logging/test/shipper.test.ts +0 -152
- package/packages/logging/test/spool.test.ts +0 -96
- package/packages/mcp-adapter/test/server.test.ts +0 -91
- package/packages/memory-engine/test/engine.test.ts +0 -456
- package/packages/model-registry/test/registry.test.ts +0 -342
- package/packages/nextjs-plugin/test/analyzer.test.ts +0 -149
- package/packages/notification/test/approval-payload.test.ts +0 -269
- package/packages/notification/test/detail-format.test.ts +0 -106
- package/packages/notification/test/notification.test.ts +0 -375
- package/packages/notification/test/pending-question-lifecycle.test.ts +0 -195
- package/packages/notification/test/pending-question.test.ts +0 -331
- package/packages/notification/test/tooling-proposal-payload.test.ts +0 -212
- package/packages/observability/test/observability.test.ts +0 -1350
- package/packages/performance-optimizer/test/optimizer.test.ts +0 -267
- package/packages/project-analyzer/test/index.test.ts +0 -1719
- package/packages/prompt-compiler/test/prompt-compiler.test.ts +0 -813
- package/packages/prompt-versioning/test/versioning.test.ts +0 -112
- package/packages/provider-adapter-sdk/test/sdk.test.ts +0 -220
- package/packages/provider-router/test/provider-router.test.ts +0 -405
- package/packages/provider-selector/test/selector.test.ts +0 -192
- package/packages/providers/test/providers.test.ts +0 -225
- package/packages/qdrant-skills/test/collection-info.test.ts +0 -165
- package/packages/qdrant-skills/test/embedder-batch.test.ts +0 -184
- package/packages/qdrant-skills/test/embedder.test.ts +0 -97
- package/packages/quota-manager/test/quota.test.ts +0 -432
- package/packages/rate-limiter/test/limiter.test.ts +0 -128
- package/packages/react-vite-plugin/test/analyzer.test.ts +0 -139
- package/packages/relay/test/client.test.ts +0 -446
- package/packages/relay/test/relay-e2e.test.ts +0 -299
- package/packages/relay/test/relay-endpoint.test.ts +0 -278
- package/packages/release-manager/test/release.test.ts +0 -215
- package/packages/requirement-compiler/test/requirement-compiler.test.ts +0 -743
- package/packages/runtime/test/runtime-api.test.ts +0 -169
- package/packages/runtime/test/runtime.test.ts +0 -297
- package/packages/scheduler/test/scheduler.test.ts +0 -125
- package/packages/session/test/manager.test.ts +0 -233
- package/packages/shared-context/test/shared-context.test.ts +0 -114
- package/packages/skill-registry/test/registry.test.ts +0 -508
- package/packages/skills/test/init-skills.test.ts +0 -114
- package/packages/skills/test/remote-skills.test.ts +0 -360
- package/packages/skills/test/remote-source.test.ts +0 -248
- package/packages/skills/test/sanitize.test.ts +0 -250
- package/packages/skills/test/vector-indexing.test.ts +0 -205
- package/packages/skills/test/vector-schema.test.ts +0 -180
- package/packages/skills/test/vector-threshold.test.ts +0 -108
- package/packages/sprint-planner/test/planner.test.ts +0 -187
- package/packages/task-compiler/test/task-compiler.test.ts +0 -628
- package/packages/test-data-generator/test/generator.test.ts +0 -265
- package/packages/test-generator/test/generator.test.ts +0 -156
- package/packages/token-estimation/test/token-estimation.test.ts +0 -207
- package/packages/token-optimizer/test/optimizer.test.ts +0 -141
- package/packages/tooling-preference/tests/changed-files.test.ts +0 -124
- package/packages/tooling-preference/tests/detect.test.ts +0 -188
- package/packages/tooling-preference/tests/proposal.test.ts +0 -107
- package/packages/tooling-preference/tests/quality-check.test.ts +0 -272
- package/packages/tooling-preference/tests/registry.test.ts +0 -51
- package/packages/tooling-preference/tests/store.test.ts +0 -99
- package/packages/trajectory/test/store.test.ts +0 -250
- package/packages/tui/test/harness-footer.test.ts +0 -559
- package/packages/tui/test/harness-header.test.ts +0 -590
- package/packages/two-phase-review/test/classify.test.ts +0 -102
- package/packages/two-phase-review/test/helpers.ts +0 -72
- package/packages/two-phase-review/test/rules.test.ts +0 -150
- package/packages/two-phase-review/test/runtime.test.ts +0 -362
- package/packages/two-phase-review/test/scope.test.ts +0 -186
- package/packages/two-phase-review/test/signals.test.ts +0 -183
- package/packages/two-phase-review/test/state.test.ts +0 -201
- package/packages/wiki/test/basic.test.ts +0 -242
- package/packages/workspace-scanner/test/scanner.test.ts +0 -43
- package/packages/worktree/test/worktree.test.ts +0 -163
|
@@ -1,536 +0,0 @@
|
|
|
1
|
-
/**
|
|
2
|
-
* Unit tests for the Jev provider chain.
|
|
3
|
-
*
|
|
4
|
-
* The TypeSafe provider is stubbed at the `fetch` boundary and the OpenRouter
|
|
5
|
-
* provider runs against a local HTTP server, so no network calls to the real
|
|
6
|
-
* services and no credentials are involved.
|
|
7
|
-
*
|
|
8
|
-
* The behaviour under test:
|
|
9
|
-
*
|
|
10
|
-
* 1. A credential belongs to ONE provider. A TypeSafe key is never sent to
|
|
11
|
-
* OpenRouter and an OpenRouter key is never sent to the TypeSafe API.
|
|
12
|
-
* 2. The next provider is only used when the previous one failed with a
|
|
13
|
-
* retryable error. A request contract error (422) is never papered over
|
|
14
|
-
* with a fallback provider, because it is a bug in the request we build.
|
|
15
|
-
*/
|
|
16
|
-
|
|
17
|
-
import { afterEach, describe, expect, it } from "bun:test";
|
|
18
|
-
import { createServer, type Server } from "node:http";
|
|
19
|
-
import { mkdtempSync, rmSync, writeFileSync } from "node:fs";
|
|
20
|
-
import { tmpdir } from "node:os";
|
|
21
|
-
import { join } from "node:path";
|
|
22
|
-
import type { AddressInfo } from "node:net";
|
|
23
|
-
import {
|
|
24
|
-
JevJudge,
|
|
25
|
-
resolveJevProviders,
|
|
26
|
-
OPENROUTER_DEFAULT_MODEL,
|
|
27
|
-
TYPESAFE_DEFAULT_MODEL,
|
|
28
|
-
} from "../src/index.js";
|
|
29
|
-
import { JevProviderError } from "../src/provider-error.js";
|
|
30
|
-
import type { JevProviderConfig } from "../src/types.js";
|
|
31
|
-
|
|
32
|
-
const originalFetch = globalThis.fetch;
|
|
33
|
-
|
|
34
|
-
interface Call {
|
|
35
|
-
url: string;
|
|
36
|
-
authorization: string;
|
|
37
|
-
}
|
|
38
|
-
|
|
39
|
-
let calls: Call[] = [];
|
|
40
|
-
let tempDirs: string[] = [];
|
|
41
|
-
|
|
42
|
-
/**
|
|
43
|
-
* A temporary key directory with the given key files.
|
|
44
|
-
*/
|
|
45
|
-
function makeKeysDir(files: Record<string, string>): string {
|
|
46
|
-
const dir = mkdtempSync(join(tmpdir(), "jev-keys-"));
|
|
47
|
-
tempDirs.push(dir);
|
|
48
|
-
for (const [name, content] of Object.entries(files)) {
|
|
49
|
-
writeFileSync(join(dir, name), content, "utf8");
|
|
50
|
-
}
|
|
51
|
-
return dir;
|
|
52
|
-
}
|
|
53
|
-
|
|
54
|
-
/**
|
|
55
|
-
* Stub the TypeSafe endpoint (the provider under test uses global fetch).
|
|
56
|
-
*/
|
|
57
|
-
function stubTypeSafeFetch(status: number, body: unknown): void {
|
|
58
|
-
globalThis.fetch = (async (url: string, init?: { headers?: Record<string, string> }) => {
|
|
59
|
-
calls.push({
|
|
60
|
-
url,
|
|
61
|
-
authorization: init?.headers?.["Authorization"] ?? "",
|
|
62
|
-
});
|
|
63
|
-
return {
|
|
64
|
-
ok: status >= 200 && status < 300,
|
|
65
|
-
status,
|
|
66
|
-
json: async () => body,
|
|
67
|
-
text: async () => JSON.stringify(body),
|
|
68
|
-
};
|
|
69
|
-
}) as unknown as typeof fetch;
|
|
70
|
-
}
|
|
71
|
-
|
|
72
|
-
/**
|
|
73
|
-
* A local OpenRouter compatible endpoint (the OpenAI SDK does its own
|
|
74
|
-
* networking, so the provider cannot be stubbed at the fetch boundary).
|
|
75
|
-
*/
|
|
76
|
-
async function startOpenRouterServer(options: {
|
|
77
|
-
status?: number;
|
|
78
|
-
content?: string;
|
|
79
|
-
}): Promise<{ baseUrl: string; server: Server; close: () => Promise<void> }> {
|
|
80
|
-
const status = options.status ?? 200;
|
|
81
|
-
const content = options.content ?? JSON.stringify({ answer: { type: "noul", noul: 0.5 } });
|
|
82
|
-
|
|
83
|
-
const server = createServer((req, res) => {
|
|
84
|
-
calls.push({
|
|
85
|
-
url: req.url ?? "",
|
|
86
|
-
authorization: req.headers["authorization"] ?? "",
|
|
87
|
-
});
|
|
88
|
-
const payload = {
|
|
89
|
-
id: "chatcmpl-test",
|
|
90
|
-
object: "chat.completion",
|
|
91
|
-
created: Math.floor(Date.now() / 1000),
|
|
92
|
-
model: "test-model",
|
|
93
|
-
choices: [
|
|
94
|
-
{
|
|
95
|
-
index: 0,
|
|
96
|
-
message: { role: "assistant", content },
|
|
97
|
-
finish_reason: "stop",
|
|
98
|
-
},
|
|
99
|
-
],
|
|
100
|
-
usage: { prompt_tokens: 1, completion_tokens: 1, total_tokens: 2 },
|
|
101
|
-
};
|
|
102
|
-
res.writeHead(status, { "content-type": "application/json" });
|
|
103
|
-
res.end(JSON.stringify(payload));
|
|
104
|
-
});
|
|
105
|
-
|
|
106
|
-
await new Promise<void>((resolve) => server.listen(0, "127.0.0.1", resolve));
|
|
107
|
-
const port = (server.address() as AddressInfo).port;
|
|
108
|
-
|
|
109
|
-
return {
|
|
110
|
-
baseUrl: `http://127.0.0.1:${port}/v1`,
|
|
111
|
-
server,
|
|
112
|
-
close: () =>
|
|
113
|
-
new Promise<void>((resolve) => {
|
|
114
|
-
server.close(() => resolve());
|
|
115
|
-
}),
|
|
116
|
-
};
|
|
117
|
-
}
|
|
118
|
-
|
|
119
|
-
function typesafeProvider(overrides: Partial<JevProviderConfig> = {}): JevProviderConfig {
|
|
120
|
-
return {
|
|
121
|
-
provider: "typesafe",
|
|
122
|
-
apiKey: "typesafe-secret",
|
|
123
|
-
baseUrl: "https://typesafe.invalid",
|
|
124
|
-
model: "test-model",
|
|
125
|
-
...overrides,
|
|
126
|
-
};
|
|
127
|
-
}
|
|
128
|
-
|
|
129
|
-
function openrouterProvider(
|
|
130
|
-
baseUrl: string,
|
|
131
|
-
overrides: Partial<JevProviderConfig> = {}
|
|
132
|
-
): JevProviderConfig {
|
|
133
|
-
return {
|
|
134
|
-
provider: "openrouter",
|
|
135
|
-
apiKey: "openrouter-secret",
|
|
136
|
-
baseUrl,
|
|
137
|
-
model: "test-model",
|
|
138
|
-
...overrides,
|
|
139
|
-
};
|
|
140
|
-
}
|
|
141
|
-
|
|
142
|
-
afterEach(() => {
|
|
143
|
-
globalThis.fetch = originalFetch;
|
|
144
|
-
calls = [];
|
|
145
|
-
for (const dir of tempDirs) {
|
|
146
|
-
rmSync(dir, { recursive: true, force: true });
|
|
147
|
-
}
|
|
148
|
-
tempDirs = [];
|
|
149
|
-
});
|
|
150
|
-
|
|
151
|
-
describe("resolveJevProviders", () => {
|
|
152
|
-
it("binds each environment credential to its own provider", () => {
|
|
153
|
-
const providers = resolveJevProviders({
|
|
154
|
-
env: { TYPESAFE_API_KEY: "typesafe-secret", OPENROUTER_API_KEY: "openrouter-secret" },
|
|
155
|
-
keysDir: "/nonexistent-keys-dir",
|
|
156
|
-
});
|
|
157
|
-
|
|
158
|
-
expect(providers.length).toBe(2);
|
|
159
|
-
expect(providers[0].provider).toBe("typesafe");
|
|
160
|
-
expect(providers[0].apiKey).toBe("typesafe-secret");
|
|
161
|
-
expect(providers[0].baseUrl).toContain("typesafe.ai");
|
|
162
|
-
expect(providers[0].model).toBe(TYPESAFE_DEFAULT_MODEL);
|
|
163
|
-
expect(providers[1].provider).toBe("openrouter");
|
|
164
|
-
expect(providers[1].apiKey).toBe("openrouter-secret");
|
|
165
|
-
expect(providers[1].baseUrl).toContain("openrouter");
|
|
166
|
-
expect(providers[1].model).toBe(OPENROUTER_DEFAULT_MODEL);
|
|
167
|
-
});
|
|
168
|
-
|
|
169
|
-
it("never puts the TypeSafe key on the OpenRouter provider", () => {
|
|
170
|
-
const providers = resolveJevProviders({
|
|
171
|
-
env: { TYPESAFE_API_KEY: "typesafe-secret" },
|
|
172
|
-
keysDir: "/nonexistent-keys-dir",
|
|
173
|
-
});
|
|
174
|
-
|
|
175
|
-
expect(providers.length).toBe(1);
|
|
176
|
-
expect(providers[0].provider).toBe("typesafe");
|
|
177
|
-
expect(providers.some((p) => p.provider === "openrouter")).toBe(false);
|
|
178
|
-
});
|
|
179
|
-
|
|
180
|
-
it("never puts the OpenRouter key on the TypeSafe provider", () => {
|
|
181
|
-
const providers = resolveJevProviders({
|
|
182
|
-
env: { OPENROUTER_API_KEY: "openrouter-secret" },
|
|
183
|
-
keysDir: "/nonexistent-keys-dir",
|
|
184
|
-
});
|
|
185
|
-
|
|
186
|
-
expect(providers.length).toBe(1);
|
|
187
|
-
expect(providers[0].provider).toBe("openrouter");
|
|
188
|
-
expect(providers[0].apiKey).toBe("openrouter-secret");
|
|
189
|
-
});
|
|
190
|
-
|
|
191
|
-
it("returns no provider when no credential is available", () => {
|
|
192
|
-
const providers = resolveJevProviders({
|
|
193
|
-
env: {},
|
|
194
|
-
keysDir: "/nonexistent-keys-dir",
|
|
195
|
-
});
|
|
196
|
-
|
|
197
|
-
expect(providers.length).toBe(0);
|
|
198
|
-
});
|
|
199
|
-
|
|
200
|
-
it("reads the key files and binds each key to its own provider", () => {
|
|
201
|
-
const keysDir = makeKeysDir({
|
|
202
|
-
"typesafe-api-key.txt": "typesafe-file-secret",
|
|
203
|
-
"openrouter-api-key.txt": "openrouter-file-secret",
|
|
204
|
-
});
|
|
205
|
-
|
|
206
|
-
const providers = resolveJevProviders({ env: {}, keysDir });
|
|
207
|
-
|
|
208
|
-
expect(providers.length).toBe(2);
|
|
209
|
-
expect(providers[0].provider).toBe("typesafe");
|
|
210
|
-
expect(providers[0].apiKey).toBe("typesafe-file-secret");
|
|
211
|
-
expect(providers[0].baseUrl).toContain("typesafe.ai");
|
|
212
|
-
expect(providers[1].provider).toBe("openrouter");
|
|
213
|
-
expect(providers[1].apiKey).toBe("openrouter-file-secret");
|
|
214
|
-
expect(providers[1].baseUrl).toContain("openrouter");
|
|
215
|
-
});
|
|
216
|
-
|
|
217
|
-
it("binds the legacy key file by the shape of the key it holds", () => {
|
|
218
|
-
const keysDir = makeKeysDir({
|
|
219
|
-
"jev-api-key.txt": "sk-or-v1-legacy-openrouter-secret",
|
|
220
|
-
});
|
|
221
|
-
|
|
222
|
-
const providers = resolveJevProviders({ env: {}, keysDir });
|
|
223
|
-
|
|
224
|
-
expect(providers.length).toBe(1);
|
|
225
|
-
// An OpenRouter key must never be sent to the TypeSafe API.
|
|
226
|
-
expect(providers[0].provider).toBe("openrouter");
|
|
227
|
-
expect(providers[0].apiKey).toBe("sk-or-v1-legacy-openrouter-secret");
|
|
228
|
-
expect(providers[0].baseUrl).toContain("openrouter");
|
|
229
|
-
});
|
|
230
|
-
|
|
231
|
-
it("binds a legacy key file that is not an OpenRouter key to TypeSafe", () => {
|
|
232
|
-
const keysDir = makeKeysDir({
|
|
233
|
-
"jev-api-key.txt": "legacy-typesafe-secret",
|
|
234
|
-
});
|
|
235
|
-
|
|
236
|
-
const providers = resolveJevProviders({ env: {}, keysDir });
|
|
237
|
-
|
|
238
|
-
expect(providers.length).toBe(1);
|
|
239
|
-
expect(providers[0].provider).toBe("typesafe");
|
|
240
|
-
expect(providers[0].apiKey).toBe("legacy-typesafe-secret");
|
|
241
|
-
});
|
|
242
|
-
|
|
243
|
-
it("prefers the dedicated key file over the legacy one", () => {
|
|
244
|
-
const keysDir = makeKeysDir({
|
|
245
|
-
"typesafe-api-key.txt": "typesafe-file-secret",
|
|
246
|
-
"jev-api-key.txt": "legacy-typesafe-secret",
|
|
247
|
-
});
|
|
248
|
-
|
|
249
|
-
const providers = resolveJevProviders({ env: {}, keysDir });
|
|
250
|
-
|
|
251
|
-
expect(providers.length).toBe(1);
|
|
252
|
-
expect(providers[0].apiKey).toBe("typesafe-file-secret");
|
|
253
|
-
});
|
|
254
|
-
|
|
255
|
-
it("keeps TypeSafe first when the legacy key is the TypeSafe one", () => {
|
|
256
|
-
const keysDir = makeKeysDir({
|
|
257
|
-
"jev-api-key.txt": "legacy-typesafe-secret",
|
|
258
|
-
"openrouter-api-key.txt": "openrouter-file-secret",
|
|
259
|
-
});
|
|
260
|
-
|
|
261
|
-
const providers = resolveJevProviders({ env: {}, keysDir });
|
|
262
|
-
|
|
263
|
-
// TypeSafe is the primary provider, OpenRouter the fallback.
|
|
264
|
-
expect(providers.length).toBe(2);
|
|
265
|
-
expect(providers[0].provider).toBe("typesafe");
|
|
266
|
-
expect(providers[0].apiKey).toBe("legacy-typesafe-secret");
|
|
267
|
-
expect(providers[1].provider).toBe("openrouter");
|
|
268
|
-
expect(providers[1].apiKey).toBe("openrouter-file-secret");
|
|
269
|
-
});
|
|
270
|
-
});
|
|
271
|
-
|
|
272
|
-
describe("JevJudge - provider chain", () => {
|
|
273
|
-
it("uses the first provider when it succeeds", async () => {
|
|
274
|
-
stubTypeSafeFetch(200, {
|
|
275
|
-
answers: { answer: { type: "noul", noul: 0.9 } },
|
|
276
|
-
model: "test-model",
|
|
277
|
-
});
|
|
278
|
-
|
|
279
|
-
const judge = new JevJudge({
|
|
280
|
-
providers: [typesafeProvider(), openrouterProvider("http://127.0.0.1:1/v1")],
|
|
281
|
-
});
|
|
282
|
-
|
|
283
|
-
const result = await judge.evaluate("state", {
|
|
284
|
-
answer: { type: "noul", instructions: "Is it true?" },
|
|
285
|
-
});
|
|
286
|
-
|
|
287
|
-
expect(result.provider).toBe("typesafe");
|
|
288
|
-
expect(result.decisions.answer.probability).toBe(0.9);
|
|
289
|
-
expect(calls.length).toBe(1);
|
|
290
|
-
expect(calls[0].url).toContain("typesafe.invalid");
|
|
291
|
-
});
|
|
292
|
-
|
|
293
|
-
it("falls back to the next provider on a server error", async () => {
|
|
294
|
-
stubTypeSafeFetch(500, { detail: "server error" });
|
|
295
|
-
const openrouter = await startOpenRouterServer({});
|
|
296
|
-
|
|
297
|
-
try {
|
|
298
|
-
const judge = new JevJudge({
|
|
299
|
-
providers: [typesafeProvider(), openrouterProvider(openrouter.baseUrl)],
|
|
300
|
-
});
|
|
301
|
-
|
|
302
|
-
const result = await judge.evaluate("state", {
|
|
303
|
-
answer: { type: "noul", instructions: "Is it true?" },
|
|
304
|
-
});
|
|
305
|
-
|
|
306
|
-
expect(result.provider).toBe("openrouter");
|
|
307
|
-
expect(calls.length).toBe(2);
|
|
308
|
-
expect(calls[0].url).toContain("typesafe.invalid");
|
|
309
|
-
expect(calls[1].url).toContain("/chat/completions");
|
|
310
|
-
} finally {
|
|
311
|
-
await openrouter.close();
|
|
312
|
-
}
|
|
313
|
-
});
|
|
314
|
-
|
|
315
|
-
it("falls back on an auth error", async () => {
|
|
316
|
-
stubTypeSafeFetch(401, { detail: "invalid token" });
|
|
317
|
-
const openrouter = await startOpenRouterServer({});
|
|
318
|
-
|
|
319
|
-
try {
|
|
320
|
-
const judge = new JevJudge({
|
|
321
|
-
providers: [typesafeProvider(), openrouterProvider(openrouter.baseUrl)],
|
|
322
|
-
});
|
|
323
|
-
|
|
324
|
-
const result = await judge.evaluate("state", {
|
|
325
|
-
answer: { type: "noul", instructions: "Is it true?" },
|
|
326
|
-
});
|
|
327
|
-
|
|
328
|
-
expect(result.provider).toBe("openrouter");
|
|
329
|
-
expect(calls.length).toBe(2);
|
|
330
|
-
} finally {
|
|
331
|
-
await openrouter.close();
|
|
332
|
-
}
|
|
333
|
-
});
|
|
334
|
-
|
|
335
|
-
it("does NOT fall back on a contract error", async () => {
|
|
336
|
-
stubTypeSafeFetch(422, { detail: "Input should be a valid dictionary" });
|
|
337
|
-
|
|
338
|
-
const judge = new JevJudge({
|
|
339
|
-
providers: [typesafeProvider(), openrouterProvider("http://127.0.0.1:1/v1")],
|
|
340
|
-
});
|
|
341
|
-
|
|
342
|
-
const error = (await judge
|
|
343
|
-
.evaluate("state", { answer: { type: "noul", instructions: "Is it true?" } })
|
|
344
|
-
.catch((e: unknown) => e)) as JevProviderError;
|
|
345
|
-
|
|
346
|
-
expect(error).toBeInstanceOf(JevProviderError);
|
|
347
|
-
expect((error as JevProviderError).status).toBe(422);
|
|
348
|
-
expect((error as JevProviderError).retryable).toBe(false);
|
|
349
|
-
// The second provider was never asked: a contract error is a bug, not
|
|
350
|
-
// a provider outage.
|
|
351
|
-
expect(calls.length).toBe(1);
|
|
352
|
-
expect(calls[0].url).toContain("typesafe.invalid");
|
|
353
|
-
});
|
|
354
|
-
|
|
355
|
-
it("sends the credential of the provider it belongs to", async () => {
|
|
356
|
-
stubTypeSafeFetch(500, { detail: "server error" });
|
|
357
|
-
const openrouter = await startOpenRouterServer({});
|
|
358
|
-
|
|
359
|
-
try {
|
|
360
|
-
const judge = new JevJudge({
|
|
361
|
-
providers: [
|
|
362
|
-
typesafeProvider({ apiKey: "typesafe-secret" }),
|
|
363
|
-
openrouterProvider(openrouter.baseUrl, { apiKey: "openrouter-secret" }),
|
|
364
|
-
],
|
|
365
|
-
});
|
|
366
|
-
|
|
367
|
-
await judge.evaluate("state", {
|
|
368
|
-
answer: { type: "noul", instructions: "Is it true?" },
|
|
369
|
-
});
|
|
370
|
-
|
|
371
|
-
expect(calls[0].authorization).toBe("Bearer typesafe-secret");
|
|
372
|
-
expect(calls[1].authorization).toBe("Bearer openrouter-secret");
|
|
373
|
-
} finally {
|
|
374
|
-
await openrouter.close();
|
|
375
|
-
}
|
|
376
|
-
});
|
|
377
|
-
|
|
378
|
-
it("reports the last error when every provider fails", async () => {
|
|
379
|
-
stubTypeSafeFetch(503, { detail: "service unavailable" });
|
|
380
|
-
const openrouter = await startOpenRouterServer({ status: 500 });
|
|
381
|
-
|
|
382
|
-
try {
|
|
383
|
-
const judge = new JevJudge({
|
|
384
|
-
providers: [typesafeProvider(), openrouterProvider(openrouter.baseUrl)],
|
|
385
|
-
});
|
|
386
|
-
|
|
387
|
-
const error = (await judge
|
|
388
|
-
.evaluate("state", { answer: { type: "noul", instructions: "Is it true?" } })
|
|
389
|
-
.catch((e: unknown) => e)) as JevProviderError;
|
|
390
|
-
|
|
391
|
-
expect(error).toBeInstanceOf(JevProviderError);
|
|
392
|
-
// The status of the LAST provider (500), not of the first one (503).
|
|
393
|
-
expect((error as JevProviderError).status).toBe(500);
|
|
394
|
-
expect((error as JevProviderError).provider).toBe("openrouter");
|
|
395
|
-
// Both providers were contacted (the OpenRouter SDK retries a 5xx
|
|
396
|
-
// on its own, so the exact number of calls is not fixed).
|
|
397
|
-
expect(calls.length).toBeGreaterThanOrEqual(2);
|
|
398
|
-
expect(calls.some((call) => call.url.includes("/chat/completions"))).toBe(true);
|
|
399
|
-
} finally {
|
|
400
|
-
await openrouter.close();
|
|
401
|
-
}
|
|
402
|
-
});
|
|
403
|
-
|
|
404
|
-
it("does NOT fall back on a bad request (400)", async () => {
|
|
405
|
-
stubTypeSafeFetch(400, { detail: "Invalid request." });
|
|
406
|
-
|
|
407
|
-
const judge = new JevJudge({
|
|
408
|
-
providers: [typesafeProvider(), openrouterProvider("http://127.0.0.1:1/v1")],
|
|
409
|
-
});
|
|
410
|
-
|
|
411
|
-
const error = (await judge
|
|
412
|
-
.evaluate("state", { answer: { type: "noul", instructions: "Is it true?" } })
|
|
413
|
-
.catch((e: unknown) => e)) as JevProviderError;
|
|
414
|
-
|
|
415
|
-
expect((error as JevProviderError).status).toBe(400);
|
|
416
|
-
expect((error as JevProviderError).retryable).toBe(false);
|
|
417
|
-
expect(calls.length).toBe(1);
|
|
418
|
-
});
|
|
419
|
-
|
|
420
|
-
it("falls back on a rate limit (429)", async () => {
|
|
421
|
-
stubTypeSafeFetch(429, { detail: "slow down" });
|
|
422
|
-
const openrouter = await startOpenRouterServer({});
|
|
423
|
-
|
|
424
|
-
try {
|
|
425
|
-
const judge = new JevJudge({
|
|
426
|
-
providers: [typesafeProvider(), openrouterProvider(openrouter.baseUrl)],
|
|
427
|
-
});
|
|
428
|
-
|
|
429
|
-
const result = await judge.evaluate("state", {
|
|
430
|
-
answer: { type: "noul", instructions: "Is it true?" },
|
|
431
|
-
});
|
|
432
|
-
|
|
433
|
-
expect(result.provider).toBe("openrouter");
|
|
434
|
-
expect(calls.length).toBe(2);
|
|
435
|
-
} finally {
|
|
436
|
-
await openrouter.close();
|
|
437
|
-
}
|
|
438
|
-
});
|
|
439
|
-
|
|
440
|
-
it("falls back on a network failure", async () => {
|
|
441
|
-
globalThis.fetch = (async () => {
|
|
442
|
-
throw Object.assign(new Error("fetch failed"), { name: "APIConnectionError" });
|
|
443
|
-
}) as unknown as typeof fetch;
|
|
444
|
-
const openrouter = await startOpenRouterServer({});
|
|
445
|
-
|
|
446
|
-
try {
|
|
447
|
-
const judge = new JevJudge({
|
|
448
|
-
providers: [typesafeProvider(), openrouterProvider(openrouter.baseUrl)],
|
|
449
|
-
});
|
|
450
|
-
|
|
451
|
-
const result = await judge.evaluate("state", {
|
|
452
|
-
answer: { type: "noul", instructions: "Is it true?" },
|
|
453
|
-
});
|
|
454
|
-
|
|
455
|
-
expect(result.provider).toBe("openrouter");
|
|
456
|
-
} finally {
|
|
457
|
-
await openrouter.close();
|
|
458
|
-
}
|
|
459
|
-
});
|
|
460
|
-
|
|
461
|
-
it("gives every provider entry its own credential when the base URL is shared", async () => {
|
|
462
|
-
// The first entry fails with a retryable error, so the second entry is
|
|
463
|
-
// used. The clients are cached per provider identity: the second entry
|
|
464
|
-
// must NOT reuse the client (and the credential) of the first one.
|
|
465
|
-
let call = 0;
|
|
466
|
-
globalThis.fetch = (async (url: string, init?: { headers?: Record<string, string> }) => {
|
|
467
|
-
call += 1;
|
|
468
|
-
calls.push({
|
|
469
|
-
url,
|
|
470
|
-
authorization: init?.headers?.["Authorization"] ?? "",
|
|
471
|
-
});
|
|
472
|
-
const ok = call > 1;
|
|
473
|
-
return {
|
|
474
|
-
ok,
|
|
475
|
-
status: ok ? 200 : 503,
|
|
476
|
-
json: async () => ({ answers: { answer: { type: "noul", noul: 0.9 } } }),
|
|
477
|
-
text: async () => "{}",
|
|
478
|
-
};
|
|
479
|
-
}) as unknown as typeof fetch;
|
|
480
|
-
|
|
481
|
-
const judge = new JevJudge({
|
|
482
|
-
providers: [
|
|
483
|
-
typesafeProvider({ apiKey: "first-secret" }),
|
|
484
|
-
typesafeProvider({ apiKey: "second-secret" }),
|
|
485
|
-
],
|
|
486
|
-
});
|
|
487
|
-
|
|
488
|
-
const result = await judge.evaluate("state", {
|
|
489
|
-
answer: { type: "noul", instructions: "Is it true?" },
|
|
490
|
-
});
|
|
491
|
-
|
|
492
|
-
expect(result.provider).toBe("typesafe");
|
|
493
|
-
expect(calls.length).toBe(2);
|
|
494
|
-
expect(calls[0].authorization).toBe("Bearer first-secret");
|
|
495
|
-
expect(calls[1].authorization).toBe("Bearer second-secret");
|
|
496
|
-
});
|
|
497
|
-
|
|
498
|
-
it("keeps the legacy single provider configuration working", async () => {
|
|
499
|
-
stubTypeSafeFetch(200, {
|
|
500
|
-
answers: { answer: { type: "noul", noul: 0.7 } },
|
|
501
|
-
model: "test-model",
|
|
502
|
-
});
|
|
503
|
-
|
|
504
|
-
const judge = new JevJudge({
|
|
505
|
-
apiKey: "legacy-key",
|
|
506
|
-
baseUrl: "https://api.typesafe.ai",
|
|
507
|
-
});
|
|
508
|
-
|
|
509
|
-
const result = await judge.evaluate("state", {
|
|
510
|
-
answer: { type: "noul", instructions: "Is it true?" },
|
|
511
|
-
});
|
|
512
|
-
|
|
513
|
-
expect(result.provider).toBe("typesafe");
|
|
514
|
-
expect(calls.length).toBe(1);
|
|
515
|
-
expect(calls[0].authorization).toBe("Bearer legacy-key");
|
|
516
|
-
});
|
|
517
|
-
|
|
518
|
-
it("refuses to run without a provider", () => {
|
|
519
|
-
expect(() => new JevJudge({})).toThrow(/no provider configured/i);
|
|
520
|
-
});
|
|
521
|
-
|
|
522
|
-
it("refuses a legacy OpenRouter key without an explicit base URL", () => {
|
|
523
|
-
expect(
|
|
524
|
-
() => new JevJudge({ apiKey: "sk-or-v1-legacy-openrouter-secret" })
|
|
525
|
-
).toThrow(/openrouter/i);
|
|
526
|
-
});
|
|
527
|
-
|
|
528
|
-
it("drops a provider with an empty api key", () => {
|
|
529
|
-
expect(
|
|
530
|
-
() =>
|
|
531
|
-
new JevJudge({
|
|
532
|
-
providers: [typesafeProvider({ apiKey: "" })],
|
|
533
|
-
})
|
|
534
|
-
).toThrow(/no usable provider/i);
|
|
535
|
-
});
|
|
536
|
-
});
|
|
@@ -1,132 +0,0 @@
|
|
|
1
|
-
/**
|
|
2
|
-
* The structured outcome of the evaluation.
|
|
3
|
-
*
|
|
4
|
-
* Jev has to be able to reject a proposed action, and the outcome of the
|
|
5
|
-
* evaluation may only tighten the decision the thresholds allow.
|
|
6
|
-
*/
|
|
7
|
-
|
|
8
|
-
import { describe, it, expect } from "bun:test";
|
|
9
|
-
import {
|
|
10
|
-
applyDecisionPolicy,
|
|
11
|
-
isDecisionOutcome,
|
|
12
|
-
tightenOutcome,
|
|
13
|
-
type DecisionContext,
|
|
14
|
-
} from "../src/auto-continue.js";
|
|
15
|
-
|
|
16
|
-
const context: DecisionContext = {
|
|
17
|
-
question: {
|
|
18
|
-
requestId: "q-1",
|
|
19
|
-
text: "Should I continue with the remaining tasks?",
|
|
20
|
-
kind: "routine",
|
|
21
|
-
requiresExplicitAuthorization: false,
|
|
22
|
-
destructive: false,
|
|
23
|
-
external: false,
|
|
24
|
-
reversible: true,
|
|
25
|
-
},
|
|
26
|
-
currentTask: "Build the toggle",
|
|
27
|
-
completionRatio: 0.5,
|
|
28
|
-
totalTasks: 4,
|
|
29
|
-
completedTasks: 2,
|
|
30
|
-
blockers: [],
|
|
31
|
-
nextTasks: ["Build the toggle"],
|
|
32
|
-
validationStatus: "no pending final tier",
|
|
33
|
-
waitedMinutes: 6,
|
|
34
|
-
deadlineMinutes: 30,
|
|
35
|
-
};
|
|
36
|
-
|
|
37
|
-
describe("the structured outcome of the evaluation", () => {
|
|
38
|
-
it("recognizes the known outcomes", () => {
|
|
39
|
-
expect(isDecisionOutcome("approve")).toBe(true);
|
|
40
|
-
expect(isDecisionOutcome("proceed_with_caution")).toBe(true);
|
|
41
|
-
expect(isDecisionOutcome("wait")).toBe(true);
|
|
42
|
-
expect(isDecisionOutcome("reject")).toBe(true);
|
|
43
|
-
});
|
|
44
|
-
|
|
45
|
-
it("ignores an unknown outcome", () => {
|
|
46
|
-
expect(isDecisionOutcome("maybe")).toBe(false);
|
|
47
|
-
expect(isDecisionOutcome(undefined)).toBe(false);
|
|
48
|
-
});
|
|
49
|
-
});
|
|
50
|
-
|
|
51
|
-
describe("tightenOutcome", () => {
|
|
52
|
-
it("keeps a rejection a rejection", () => {
|
|
53
|
-
expect(tightenOutcome("approve", "reject")).toBe("reject");
|
|
54
|
-
expect(tightenOutcome("wait", "reject")).toBe("reject");
|
|
55
|
-
});
|
|
56
|
-
|
|
57
|
-
it("keeps a wait a wait", () => {
|
|
58
|
-
expect(tightenOutcome("approve", "wait")).toBe("wait");
|
|
59
|
-
});
|
|
60
|
-
|
|
61
|
-
it("turns an approval into a cautious approval", () => {
|
|
62
|
-
expect(tightenOutcome("approve", "proceed_with_caution")).toBe(
|
|
63
|
-
"proceed_with_caution",
|
|
64
|
-
);
|
|
65
|
-
});
|
|
66
|
-
|
|
67
|
-
it("never lets an approval beat the thresholds", () => {
|
|
68
|
-
expect(tightenOutcome("wait", "approve")).toBe("wait");
|
|
69
|
-
expect(tightenOutcome("proceed_with_caution", "approve")).toBe(
|
|
70
|
-
"proceed_with_caution",
|
|
71
|
-
);
|
|
72
|
-
});
|
|
73
|
-
|
|
74
|
-
it("keeps the threshold outcome without a proposal", () => {
|
|
75
|
-
expect(tightenOutcome("approve", undefined)).toBe("approve");
|
|
76
|
-
});
|
|
77
|
-
});
|
|
78
|
-
|
|
79
|
-
describe("the policy of a rejection", () => {
|
|
80
|
-
it("keeps a rejection", () => {
|
|
81
|
-
const result = applyDecisionPolicy("reject", context);
|
|
82
|
-
|
|
83
|
-
expect(result.outcome).toBe("reject");
|
|
84
|
-
expect(result.appliedRule).toBeUndefined();
|
|
85
|
-
});
|
|
86
|
-
|
|
87
|
-
it("never turns a rejection into an approval", () => {
|
|
88
|
-
const result = applyDecisionPolicy("reject", context, {
|
|
89
|
-
confidence: 0.99,
|
|
90
|
-
});
|
|
91
|
-
|
|
92
|
-
expect(result.outcome).toBe("reject");
|
|
93
|
-
});
|
|
94
|
-
});
|
|
95
|
-
|
|
96
|
-
describe("the policy of unknown risk facts", () => {
|
|
97
|
-
it("waits when nothing is known about the action", () => {
|
|
98
|
-
const result = applyDecisionPolicy(
|
|
99
|
-
"approve",
|
|
100
|
-
{
|
|
101
|
-
...context,
|
|
102
|
-
question: {
|
|
103
|
-
requestId: "q-1",
|
|
104
|
-
text: "Should I continue?",
|
|
105
|
-
kind: "routine",
|
|
106
|
-
},
|
|
107
|
-
},
|
|
108
|
-
{ confidence: 0.99 },
|
|
109
|
-
);
|
|
110
|
-
|
|
111
|
-
expect(result.outcome).toBe("wait");
|
|
112
|
-
expect(result.appliedRule).toBe("unknown-risk-facts");
|
|
113
|
-
});
|
|
114
|
-
|
|
115
|
-
it("still blocks a clarification first", () => {
|
|
116
|
-
const result = applyDecisionPolicy(
|
|
117
|
-
"approve",
|
|
118
|
-
{
|
|
119
|
-
...context,
|
|
120
|
-
question: {
|
|
121
|
-
requestId: "q-1",
|
|
122
|
-
text: "Do you mean the toggle or the switch?",
|
|
123
|
-
kind: "clarification",
|
|
124
|
-
},
|
|
125
|
-
},
|
|
126
|
-
{ confidence: 0.99 },
|
|
127
|
-
);
|
|
128
|
-
|
|
129
|
-
expect(result.outcome).toBe("wait");
|
|
130
|
-
expect(result.appliedRule).toBe("clarification-never-approved");
|
|
131
|
-
});
|
|
132
|
-
});
|