modelrig 0.8.1 → 0.10.0
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 +324 -0
- package/README.md +55 -29
- package/data/PROVENANCE.md +14 -0
- package/data/pricing-map.json +26 -0
- package/dist/adapters/attempt-meta-map.d.ts +8 -1
- package/dist/adapters/attempt-meta-map.d.ts.map +1 -1
- package/dist/adapters/attempt-meta-map.js +4 -0
- package/dist/adapters/build-adapters.js +2 -2
- package/dist/adapters/gemini.d.ts +28 -1
- package/dist/adapters/gemini.d.ts.map +1 -1
- package/dist/adapters/gemini.js +38 -9
- package/dist/adapters/grok.d.ts.map +1 -1
- package/dist/adapters/grok.js +53 -4
- package/dist/adapters/openai-compat.d.ts +8 -2
- package/dist/adapters/openai-compat.d.ts.map +1 -1
- package/dist/adapters/openai-compat.js +8 -1
- package/dist/adapters/openai-stream.d.ts +10 -2
- package/dist/adapters/openai-stream.d.ts.map +1 -1
- package/dist/adapters/openai-stream.js +5 -1
- package/dist/adapters/openai.d.ts +10 -1
- package/dist/adapters/openai.d.ts.map +1 -1
- package/dist/adapters/openai.js +57 -7
- package/dist/adapters/responses-api.d.ts +61 -0
- package/dist/adapters/responses-api.d.ts.map +1 -0
- package/dist/adapters/responses-api.js +124 -0
- package/dist/artifacts/artifact.d.ts +31 -12
- package/dist/artifacts/artifact.d.ts.map +1 -1
- package/dist/artifacts/artifact.js +218 -188
- package/dist/artifacts/index.d.ts +15 -5
- package/dist/artifacts/index.d.ts.map +1 -1
- package/dist/artifacts/index.js +19 -5
- package/dist/artifacts/run-context.d.ts +3 -24
- package/dist/artifacts/run-context.d.ts.map +1 -1
- package/dist/artifacts/run-context.js +3 -39
- package/dist/bakeoff/engine.d.ts +29 -0
- package/dist/bakeoff/engine.d.ts.map +1 -1
- package/dist/bakeoff/engine.js +47 -0
- package/dist/bakeoff/eval-cases-rig.d.ts +36 -0
- package/dist/bakeoff/eval-cases-rig.d.ts.map +1 -0
- package/dist/bakeoff/eval-cases-rig.js +27 -0
- package/dist/bakeoff/eval-cases.d.ts +19 -0
- package/dist/bakeoff/eval-cases.d.ts.map +1 -1
- package/dist/bakeoff/eval-cases.js +18 -1
- package/dist/bakeoff/golden-judge.d.ts +47 -0
- package/dist/bakeoff/golden-judge.d.ts.map +1 -0
- package/dist/bakeoff/golden-judge.js +49 -0
- package/dist/bakeoff/groundedness.d.ts +58 -0
- package/dist/bakeoff/groundedness.d.ts.map +1 -0
- package/dist/bakeoff/groundedness.js +165 -0
- package/dist/bakeoff/quality.d.ts +43 -0
- package/dist/bakeoff/quality.d.ts.map +1 -1
- package/dist/bakeoff/quality.js +31 -0
- package/dist/bakeoff/replay.d.ts +11 -0
- package/dist/bakeoff/replay.d.ts.map +1 -1
- package/dist/bakeoff/replay.js +12 -0
- package/dist/bakeoff/report.d.ts +7 -0
- package/dist/bakeoff/report.d.ts.map +1 -1
- package/dist/bakeoff/report.js +26 -0
- package/dist/cli-judge.d.ts +12 -0
- package/dist/cli-judge.d.ts.map +1 -0
- package/dist/cli-judge.js +48 -0
- package/dist/cli.d.ts.map +1 -1
- package/dist/cli.js +22 -42
- package/dist/config.d.ts +85 -29
- package/dist/config.d.ts.map +1 -1
- package/dist/config.js +109 -47
- package/dist/detect/detector.d.ts +4 -2
- package/dist/detect/detector.d.ts.map +1 -1
- package/dist/detect/safe-harbor.d.ts +57 -0
- package/dist/detect/safe-harbor.d.ts.map +1 -0
- package/dist/detect/safe-harbor.js +278 -0
- package/dist/doctor-cli.d.ts +7 -0
- package/dist/doctor-cli.d.ts.map +1 -0
- package/dist/doctor-cli.js +67 -0
- package/dist/doctor-preflight.d.ts +52 -0
- package/dist/doctor-preflight.d.ts.map +1 -0
- package/dist/doctor-preflight.js +112 -0
- package/dist/feedback/coverage.d.ts +11 -8
- package/dist/feedback/coverage.d.ts.map +1 -1
- package/dist/feedback/coverage.js +29 -16
- package/dist/feedback/grade.d.ts +58 -0
- package/dist/feedback/grade.d.ts.map +1 -0
- package/dist/feedback/grade.js +67 -0
- package/dist/gateway/http-protocol.d.ts +20 -1
- package/dist/gateway/http-protocol.d.ts.map +1 -1
- package/dist/gateway/http-protocol.js +1 -1
- package/dist/gateway/serve-client.d.ts +99 -0
- package/dist/gateway/serve-client.d.ts.map +1 -0
- package/dist/gateway/serve-client.js +360 -0
- package/dist/index.d.ts +68 -29
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +167 -46
- package/dist/init/observe-cli.js +1 -1
- package/dist/init/observe.d.ts +2 -2
- package/dist/init/observe.d.ts.map +1 -1
- package/dist/init/observe.js +11 -7
- package/dist/mask/arm.d.ts +112 -0
- package/dist/mask/arm.d.ts.map +1 -0
- package/dist/mask/arm.js +179 -0
- package/dist/mask/index.d.ts +2 -0
- package/dist/mask/index.d.ts.map +1 -1
- package/dist/mask/index.js +7 -1
- package/dist/mask/mask-policy-cli.d.ts +12 -4
- package/dist/mask/mask-policy-cli.d.ts.map +1 -1
- package/dist/mask/mask-policy-cli.js +34 -9
- package/dist/prompts/manifest.d.ts +128 -0
- package/dist/prompts/manifest.d.ts.map +1 -0
- package/dist/prompts/manifest.js +214 -0
- package/dist/prompts/prompt-fetch.d.ts +84 -0
- package/dist/prompts/prompt-fetch.d.ts.map +1 -0
- package/dist/prompts/prompt-fetch.js +120 -0
- package/dist/prompts/prompt-resolve.d.ts +76 -0
- package/dist/prompts/prompt-resolve.d.ts.map +1 -0
- package/dist/prompts/prompt-resolve.js +93 -0
- package/dist/prompts/slot-resolver.d.ts +91 -0
- package/dist/prompts/slot-resolver.d.ts.map +1 -0
- package/dist/prompts/slot-resolver.js +135 -0
- package/dist/prompts/tools-register.d.ts +60 -0
- package/dist/prompts/tools-register.d.ts.map +1 -0
- package/dist/prompts/tools-register.js +48 -0
- package/dist/prompts/version-hash.d.ts +29 -0
- package/dist/prompts/version-hash.d.ts.map +1 -0
- package/dist/prompts/version-hash.js +54 -0
- package/dist/prompts-export-cli.d.ts +21 -0
- package/dist/prompts-export-cli.d.ts.map +1 -0
- package/dist/prompts-export-cli.js +130 -0
- package/dist/query/grammar.d.ts +2 -2
- package/dist/query/grammar.d.ts.map +1 -1
- package/dist/query/grammar.js +3 -1
- package/dist/query/saved.d.ts +13 -0
- package/dist/query/saved.d.ts.map +1 -1
- package/dist/query/saved.js +67 -1
- package/dist/receipts/types.d.ts +1 -1
- package/dist/receipts/types.d.ts.map +1 -1
- package/dist/receipts/types.js +1 -0
- package/dist/registry/build.d.ts +43 -1
- package/dist/registry/build.d.ts.map +1 -1
- package/dist/registry/build.js +52 -0
- package/dist/registry/cache-policy.d.ts.map +1 -1
- package/dist/registry/cache-policy.js +22 -0
- package/dist/registry/pricing-shape.d.ts +67 -8
- package/dist/registry/pricing-shape.d.ts.map +1 -1
- package/dist/registry/pricing-shape.js +50 -13
- package/dist/registry/pricing.d.ts +6 -5
- package/dist/registry/pricing.d.ts.map +1 -1
- package/dist/registry/pricing.js +9 -5
- package/dist/registry/rows.d.ts.map +1 -1
- package/dist/registry/rows.js +1 -0
- package/dist/registry/select.d.ts +6 -1
- package/dist/registry/select.d.ts.map +1 -1
- package/dist/registry/select.js +11 -0
- package/dist/routes/call-config.d.ts +199 -0
- package/dist/routes/call-config.d.ts.map +1 -0
- package/dist/routes/call-config.js +343 -0
- package/dist/routes/destination-loader.d.ts +16 -0
- package/dist/routes/destination-loader.d.ts.map +1 -0
- package/dist/routes/destination-loader.js +29 -0
- package/dist/routes/draft.d.ts +7 -0
- package/dist/routes/draft.d.ts.map +1 -1
- package/dist/routes/draft.js +12 -0
- package/dist/routes/loader.d.ts.map +1 -1
- package/dist/routes/loader.js +31 -56
- package/dist/routes/objective-loader.d.ts +12 -0
- package/dist/routes/objective-loader.d.ts.map +1 -1
- package/dist/routes/objective-loader.js +82 -0
- package/dist/routes/output-loader.d.ts +42 -0
- package/dist/routes/output-loader.d.ts.map +1 -0
- package/dist/routes/output-loader.js +112 -0
- package/dist/routes/prompt-loader.d.ts +49 -0
- package/dist/routes/prompt-loader.d.ts.map +1 -0
- package/dist/routes/prompt-loader.js +113 -0
- package/dist/routes/resolved-bundle.d.ts +69 -0
- package/dist/routes/resolved-bundle.d.ts.map +1 -1
- package/dist/routes/resolved-bundle.js +76 -0
- package/dist/routes/screen-loader.d.ts +29 -1
- package/dist/routes/screen-loader.d.ts.map +1 -1
- package/dist/routes/screen-loader.js +59 -1
- package/dist/run/accept.d.ts +5 -0
- package/dist/run/accept.d.ts.map +1 -1
- package/dist/run/accept.js +2 -0
- package/dist/run/cost.d.ts +31 -2
- package/dist/run/cost.d.ts.map +1 -1
- package/dist/run/cost.js +57 -2
- package/dist/run/destination.d.ts +39 -0
- package/dist/run/destination.d.ts.map +1 -0
- package/dist/run/destination.js +45 -0
- package/dist/run/dispatch.d.ts +67 -16
- package/dist/run/dispatch.d.ts.map +1 -1
- package/dist/run/dispatch.js +133 -29
- package/dist/run/grounding.d.ts +28 -2
- package/dist/run/grounding.d.ts.map +1 -1
- package/dist/run/grounding.js +54 -10
- package/dist/run/raw-defaults.d.ts.map +1 -1
- package/dist/run/raw-defaults.js +6 -1
- package/dist/run/raw-finalize.d.ts +50 -0
- package/dist/run/raw-finalize.d.ts.map +1 -0
- package/dist/run/raw-finalize.js +266 -0
- package/dist/run/raw-knobs.d.ts +36 -0
- package/dist/run/raw-knobs.d.ts.map +1 -0
- package/dist/run/raw-knobs.js +99 -0
- package/dist/run/raw-request.d.ts +24 -0
- package/dist/run/raw-request.d.ts.map +1 -0
- package/dist/run/raw-request.js +69 -0
- package/dist/run/raw-schema.d.ts +76 -0
- package/dist/run/raw-schema.d.ts.map +1 -0
- package/dist/run/raw-schema.js +166 -0
- package/dist/run/raw-search-cost.d.ts +48 -0
- package/dist/run/raw-search-cost.d.ts.map +1 -0
- package/dist/run/raw-search-cost.js +78 -0
- package/dist/run/raw-support.d.ts +64 -6
- package/dist/run/raw-support.d.ts.map +1 -1
- package/dist/run/raw-support.js +84 -11
- package/dist/run/raw.d.ts +25 -26
- package/dist/run/raw.d.ts.map +1 -1
- package/dist/run/raw.js +30 -351
- package/dist/run/record-attempt.d.ts +12 -0
- package/dist/run/record-attempt.d.ts.map +1 -1
- package/dist/run/record-attempt.js +30 -12
- package/dist/run/repair-attempt.d.ts +68 -0
- package/dist/run/repair-attempt.d.ts.map +1 -0
- package/dist/run/repair-attempt.js +180 -0
- package/dist/run/resolve-destination.d.ts +49 -0
- package/dist/run/resolve-destination.d.ts.map +1 -0
- package/dist/run/resolve-destination.js +76 -0
- package/dist/run/routing-docs.d.ts.map +1 -1
- package/dist/run/routing-docs.js +10 -5
- package/dist/run/run.d.ts +30 -1
- package/dist/run/run.d.ts.map +1 -1
- package/dist/run/run.js +94 -172
- package/dist/run/telemetry-row.d.ts +92 -0
- package/dist/run/telemetry-row.d.ts.map +1 -0
- package/dist/run/telemetry-row.js +81 -0
- package/dist/run/validate.d.ts +12 -1
- package/dist/run/validate.d.ts.map +1 -1
- package/dist/run/validate.js +31 -0
- package/dist/run/web-search-pricing.d.ts +43 -0
- package/dist/run/web-search-pricing.d.ts.map +1 -0
- package/dist/run/web-search-pricing.js +56 -0
- package/dist/savings/odometer.d.ts +4 -0
- package/dist/savings/odometer.d.ts.map +1 -1
- package/dist/savings/odometer.js +20 -1
- package/dist/savings/swap-execute.d.ts +7 -0
- package/dist/savings/swap-execute.d.ts.map +1 -1
- package/dist/savings/swap-execute.js +30 -0
- package/dist/savings/swaps-client.d.ts +10 -0
- package/dist/savings/swaps-client.d.ts.map +1 -1
- package/dist/screen/fixtures/probe-corpora.d.ts +23 -0
- package/dist/screen/fixtures/probe-corpora.d.ts.map +1 -0
- package/dist/screen/fixtures/probe-corpora.js +69 -0
- package/dist/screen/index.d.ts +21 -0
- package/dist/screen/index.d.ts.map +1 -0
- package/dist/screen/index.js +64 -0
- package/dist/screen/model-screen.d.ts +111 -0
- package/dist/screen/model-screen.d.ts.map +1 -0
- package/dist/screen/model-screen.js +178 -0
- package/dist/screen/patterns.d.ts +51 -0
- package/dist/screen/patterns.d.ts.map +1 -0
- package/dist/screen/patterns.js +151 -0
- package/dist/screen/phi-gate.d.ts +52 -0
- package/dist/screen/phi-gate.d.ts.map +1 -0
- package/dist/screen/phi-gate.js +86 -0
- package/dist/screen/policy.d.ts +71 -0
- package/dist/screen/policy.d.ts.map +1 -0
- package/dist/screen/policy.js +334 -0
- package/dist/screen/probe.d.ts +118 -0
- package/dist/screen/probe.d.ts.map +1 -0
- package/dist/screen/probe.js +132 -0
- package/dist/screen/screen-cli.d.ts +16 -0
- package/dist/screen/screen-cli.d.ts.map +1 -0
- package/dist/screen/screen-cli.js +178 -0
- package/dist/screen/screen.d.ts +99 -0
- package/dist/screen/screen.d.ts.map +1 -0
- package/dist/screen/screen.js +0 -0
- package/dist/screen/sidecar.d.ts +69 -0
- package/dist/screen/sidecar.d.ts.map +1 -0
- package/dist/screen/sidecar.js +145 -0
- package/dist/screen/types.d.ts +43 -2
- package/dist/screen/types.d.ts.map +1 -1
- package/dist/settings/api-fetcher.d.ts +14 -3
- package/dist/settings/api-fetcher.d.ts.map +1 -1
- package/dist/settings/api-fetcher.js +61 -10
- package/dist/settings/grading-guide.d.ts +102 -0
- package/dist/settings/grading-guide.d.ts.map +1 -0
- package/dist/settings/grading-guide.js +23 -0
- package/dist/settings/index.d.ts +3 -2
- package/dist/settings/index.d.ts.map +1 -1
- package/dist/settings/index.js +2 -1
- package/dist/settings/task-settings.d.ts +44 -7
- package/dist/settings/task-settings.d.ts.map +1 -1
- package/dist/settings/task-settings.js +53 -12
- package/dist/status.d.ts +32 -20
- package/dist/status.d.ts.map +1 -1
- package/dist/status.js +51 -30
- package/dist/telemetry/artifact-store.d.ts +5 -6
- package/dist/telemetry/artifact-store.d.ts.map +1 -1
- package/dist/telemetry/artifact-store.js +21 -24
- package/dist/telemetry/attempt-meta.d.ts +76 -1
- package/dist/telemetry/attempt-meta.d.ts.map +1 -1
- package/dist/telemetry/attempt-meta.js +33 -1
- package/dist/telemetry/exporter-setup.d.ts +2 -0
- package/dist/telemetry/exporter-setup.d.ts.map +1 -1
- package/dist/telemetry/exporter-setup.js +6 -1
- package/dist/telemetry/exporter.d.ts +2 -2
- package/dist/telemetry/exporter.d.ts.map +1 -1
- package/dist/telemetry/exporter.js +22 -41
- package/dist/telemetry/grade-mappers.d.ts +25 -0
- package/dist/telemetry/grade-mappers.d.ts.map +1 -0
- package/dist/telemetry/grade-mappers.js +64 -0
- package/dist/telemetry/grades-sql.d.ts +56 -0
- package/dist/telemetry/grades-sql.d.ts.map +1 -0
- package/dist/telemetry/grades-sql.js +156 -0
- package/dist/telemetry/ingest-protocol.d.ts +55 -48
- package/dist/telemetry/ingest-protocol.d.ts.map +1 -1
- package/dist/telemetry/ingest-protocol.js +48 -83
- package/dist/telemetry/org-identity.d.ts +44 -0
- package/dist/telemetry/org-identity.d.ts.map +1 -0
- package/dist/telemetry/org-identity.js +98 -0
- package/dist/telemetry/route-registration.d.ts +48 -0
- package/dist/telemetry/route-registration.d.ts.map +1 -0
- package/dist/telemetry/route-registration.js +77 -0
- package/dist/telemetry/sinks.d.ts +10 -1
- package/dist/telemetry/sinks.d.ts.map +1 -1
- package/dist/telemetry/sinks.js +52 -6
- package/dist/telemetry/store-migrations.d.ts.map +1 -1
- package/dist/telemetry/store-migrations.js +64 -0
- package/dist/telemetry/store.d.ts +9 -6
- package/dist/telemetry/store.d.ts.map +1 -1
- package/dist/telemetry/store.js +8 -50
- package/dist/tenancy/api-key-format.d.ts +45 -1
- package/dist/tenancy/api-key-format.d.ts.map +1 -1
- package/dist/tenancy/api-key-format.js +50 -1
- package/dist/types/provider.d.ts +23 -3
- package/dist/types/provider.d.ts.map +1 -1
- package/dist/types/provider.js +26 -0
- package/dist/types/route.d.ts +154 -2
- package/dist/types/route.d.ts.map +1 -1
- package/dist/types/telemetry.d.ts +73 -48
- package/dist/types/telemetry.d.ts.map +1 -1
- package/dist/validate-cli.d.ts.map +1 -1
- package/dist/validate-cli.js +32 -0
- package/dist/verify/verify.d.ts +5 -0
- package/dist/verify/verify.d.ts.map +1 -1
- package/dist/verify/verify.js +2 -0
- package/dist/watch/evaluate.d.ts +84 -1
- package/dist/watch/evaluate.d.ts.map +1 -1
- package/dist/watch/evaluate.js +28 -0
- package/dist/watch/monitor.d.ts +8 -2
- package/dist/watch/monitor.d.ts.map +1 -1
- package/dist/watch/monitor.js +13 -5
- package/dist/watch/regret-monitor-cli.d.ts +16 -0
- package/dist/watch/regret-monitor-cli.d.ts.map +1 -0
- package/dist/watch/regret-monitor-cli.js +137 -0
- package/dist/watch/regret-monitor.d.ts +168 -0
- package/dist/watch/regret-monitor.d.ts.map +1 -0
- package/dist/watch/regret-monitor.js +304 -0
- package/dist/watch/upgrade.d.ts +62 -0
- package/dist/watch/upgrade.d.ts.map +1 -0
- package/dist/watch/upgrade.js +57 -0
- package/docs/agent-skill/SKILL.md +181 -9
- package/docs/how-it-fits.md +3 -2
- package/docs/instrumentation.md +219 -46
- package/docs/llms.txt +158 -35
- package/docs/managed-serving.md +150 -0
- package/docs/mcp-oracle.md +322 -0
- package/docs/migration-playbook.md +5 -4
- package/docs/quickstart.md +263 -89
- package/package.json +18 -4
- package/registry/leaderboard.json +74 -0
- package/registry/registry.json +232 -112
- package/screen.d.ts +16 -0
- package/templates/call-disposition-qa.json +77 -0
- package/templates/catalog-cleansing.json +77 -0
- package/templates/compliance-review.json +76 -0
- package/templates/cs-next-action.json +76 -0
- package/templates/document-extraction.json +78 -0
- package/templates/financial-classification.json +75 -0
- package/templates/lead-qualification.json +76 -0
- package/templates/medical-classification.json +71 -0
- package/templates/ticket-triage.json +76 -0
- package/dist/feedback/feedback.d.ts +0 -35
- package/dist/feedback/feedback.d.ts.map +0 -1
- package/dist/feedback/feedback.js +0 -39
|
@@ -0,0 +1,124 @@
|
|
|
1
|
+
"use strict";
|
|
2
|
+
/**
|
|
3
|
+
* Responses-API dispatch path (Native Web Search lane, spec N1–N2) — the SHARED
|
|
4
|
+
* request/response mapping for providers whose native `web_search` tool lives only
|
|
5
|
+
* on `/v1/responses` (OpenAI + xAI/Grok today; same shape). This is a NEW dispatch
|
|
6
|
+
* path, NOT an extension of the Chat Completions `tools` block — see research §2.
|
|
7
|
+
*
|
|
8
|
+
* DOCTRINE (inherited from CE-9):
|
|
9
|
+
* - ABSENT ⇒ BYTE-IDENTICAL. This path is entered ONLY for a grounded request
|
|
10
|
+
* (`req.grounding === true`). A non-grounded request stays on the adapter's
|
|
11
|
+
* existing `chat.completions.create` call, unchanged. The per-adapter branch
|
|
12
|
+
* lives in openai.ts / grok.ts (Step 2); this module owns only the Responses shape.
|
|
13
|
+
* - FORWARD-VERBATIM + CLASSIFY-TEACHABLY: forward the web_search tool + sampling/
|
|
14
|
+
* reasoning under their Responses-dialect names; let the provider enforce; map a
|
|
15
|
+
* rejection onto an EXISTING FailureClass with a teaching fixHint (no new class).
|
|
16
|
+
* - CITATIONS: map `response` citations onto the `url_citation` shape already used
|
|
17
|
+
* by the inject rung (src/run/grounding.ts) — do NOT invent a second type.
|
|
18
|
+
*/
|
|
19
|
+
Object.defineProperty(exports, "__esModule", { value: true });
|
|
20
|
+
exports.buildResponsesParams = buildResponsesParams;
|
|
21
|
+
exports.mapResponsesResult = mapResponsesResult;
|
|
22
|
+
const sampling_1 = require("./sampling");
|
|
23
|
+
const raw_support_1 = require("../run/raw-support");
|
|
24
|
+
/** Parity with the Chat Completions default output cap (openai.ts / grok.ts). */
|
|
25
|
+
const DEFAULT_MAX_OUTPUT_TOKENS = 32000;
|
|
26
|
+
/**
|
|
27
|
+
* Build the `responses.create` params for a grounded request. Appends the
|
|
28
|
+
* `web_search` tool; forwards sampling/reasoning. Callers MUST only invoke this when
|
|
29
|
+
* `req.grounding === true` (the guard already enforced provider support pre-dispatch).
|
|
30
|
+
*
|
|
31
|
+
* Reasoning maps 1:1 for OpenAI (`level` → `reasoning.effort`); for grok the neutral
|
|
32
|
+
* four-level `level` is coarsened to xAI's `low|high` via `coarsenGrokReasoningEffort`
|
|
33
|
+
* (the SAME disclosed mapping the Chat Completions path uses — never a silent drop).
|
|
34
|
+
*/
|
|
35
|
+
function buildResponsesParams(req) {
|
|
36
|
+
const input = [
|
|
37
|
+
...(req.systemPrompt !== "" ? [{ role: "system", content: req.systemPrompt }] : []),
|
|
38
|
+
{ role: "user", content: req.userPrompt },
|
|
39
|
+
];
|
|
40
|
+
const reasoning = req.reasoning !== undefined
|
|
41
|
+
? {
|
|
42
|
+
reasoning: {
|
|
43
|
+
effort: req.candidate.provider === "grok"
|
|
44
|
+
? (0, raw_support_1.coarsenGrokReasoningEffort)(req.reasoning.level)
|
|
45
|
+
: req.reasoning.level,
|
|
46
|
+
},
|
|
47
|
+
}
|
|
48
|
+
: {};
|
|
49
|
+
return {
|
|
50
|
+
model: req.candidate.model,
|
|
51
|
+
input,
|
|
52
|
+
// Native web-search server tool, left at defaults. As of openai@7.x the
|
|
53
|
+
// `web_search` tool type also accepts `filters`, `external_web_access`, and
|
|
54
|
+
// `search_context_size` (a token/context-budget knob) — NOT wired here on
|
|
55
|
+
// purpose (they need product knobs); available for a follow-up mission.
|
|
56
|
+
tools: [{ type: "web_search" }],
|
|
57
|
+
max_output_tokens: (0, sampling_1.effectiveMaxOutputTokens)(req, DEFAULT_MAX_OUTPUT_TOKENS),
|
|
58
|
+
// temperature / top_p share the Chat Completions dialect names on the
|
|
59
|
+
// Responses API — reuse the one sampling mapper (absent ⇒ no keys added).
|
|
60
|
+
...(0, sampling_1.openAISamplingFields)(req),
|
|
61
|
+
...reasoning,
|
|
62
|
+
};
|
|
63
|
+
}
|
|
64
|
+
/**
|
|
65
|
+
* Map a raw `responses.create` result onto ModelRig's normalized shape: output text,
|
|
66
|
+
* AdapterUsage (cache-read is a SUBSET of `tokensIn`, per the WS-A invariant), and
|
|
67
|
+
* citations in the `url_citation` shape. A zero-citation grounded response is
|
|
68
|
+
* SURFACED (empty array), never thrown (spec N2).
|
|
69
|
+
*/
|
|
70
|
+
function mapResponsesResult(response) {
|
|
71
|
+
const r = (typeof response === "object" && response !== null ? response : {});
|
|
72
|
+
const items = Array.isArray(r.output) ? r.output : [];
|
|
73
|
+
// Text: prefer the SDK convenience aggregate, else concatenate output_text parts.
|
|
74
|
+
const aggregate = typeof r.output_text === "string" ? r.output_text : "";
|
|
75
|
+
const text = aggregate !== ""
|
|
76
|
+
? aggregate
|
|
77
|
+
: items
|
|
78
|
+
.filter((it) => it.type === "message")
|
|
79
|
+
.flatMap((it) => it.content ?? [])
|
|
80
|
+
.filter((c) => c.type === "output_text" && typeof c.text === "string")
|
|
81
|
+
.map((c) => c.text)
|
|
82
|
+
.join("");
|
|
83
|
+
// Citations: url_citation annotations on output_text parts, deduped by URL,
|
|
84
|
+
// first-seen title wins. Reuses the inject rung's UrlCitation shape.
|
|
85
|
+
const seen = new Set();
|
|
86
|
+
const citations = [];
|
|
87
|
+
for (const it of items) {
|
|
88
|
+
if (it.type !== "message")
|
|
89
|
+
continue;
|
|
90
|
+
for (const part of it.content ?? []) {
|
|
91
|
+
if (part.type !== "output_text")
|
|
92
|
+
continue;
|
|
93
|
+
for (const ann of part.annotations ?? []) {
|
|
94
|
+
if (ann.type !== "url_citation" || typeof ann.url !== "string" || ann.url === "")
|
|
95
|
+
continue;
|
|
96
|
+
if (seen.has(ann.url))
|
|
97
|
+
continue;
|
|
98
|
+
seen.add(ann.url);
|
|
99
|
+
citations.push({
|
|
100
|
+
type: "url_citation",
|
|
101
|
+
url: ann.url,
|
|
102
|
+
...(typeof ann.title === "string" && ann.title !== "" ? { title: ann.title } : {}),
|
|
103
|
+
});
|
|
104
|
+
}
|
|
105
|
+
}
|
|
106
|
+
}
|
|
107
|
+
// usage: input_tokens already INCLUDES cached_tokens → tokensIn is subset-correct
|
|
108
|
+
// (WS-A). The Responses API reports no cache-write on these providers ⇒ 0/omit.
|
|
109
|
+
const usage = {
|
|
110
|
+
tokensIn: r.usage?.input_tokens ?? 0,
|
|
111
|
+
tokensOut: r.usage?.output_tokens ?? 0,
|
|
112
|
+
tokensCached: r.usage?.input_tokens_details?.cached_tokens ?? 0,
|
|
113
|
+
servedTier: "standard",
|
|
114
|
+
};
|
|
115
|
+
// searchCalls: count the web_search_call output items (spec N3). None ⇒ undefined
|
|
116
|
+
// (the raw success path assumes 1, disclosed).
|
|
117
|
+
const searchCalls = items.filter((it) => it.type === "web_search_call").length;
|
|
118
|
+
return {
|
|
119
|
+
text,
|
|
120
|
+
usage,
|
|
121
|
+
citations,
|
|
122
|
+
...(searchCalls > 0 ? { searchCalls } : {}),
|
|
123
|
+
};
|
|
124
|
+
}
|
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
/**
|
|
2
2
|
* Artifact namespace — Expansion E2 chunk B (spec §2.2, §9).
|
|
3
3
|
*
|
|
4
|
-
* `save/load/link
|
|
4
|
+
* `save/load/link` over a durable work product's METADATA. The
|
|
5
5
|
* serialized value is hashed (sha256, node:crypto — no new dep) and the bytes
|
|
6
6
|
* are DISCARDED; nothing but the metadata row is persisted (the 00030
|
|
7
7
|
* counter-annotation). Lineage v0: artifacts saved within a run get
|
|
@@ -16,14 +16,24 @@
|
|
|
16
16
|
import type { ArtifactStore } from "../telemetry/artifact-store";
|
|
17
17
|
import type { EvalCaseOrigin } from "../types/telemetry";
|
|
18
18
|
import { type ScrubMode } from "./scrub";
|
|
19
|
+
import { type TaskSettingsResolver } from "../settings/task-settings";
|
|
19
20
|
import type { ContentUploadInput, ContentUploadOutcome } from "./content-uploader";
|
|
21
|
+
import { type Manifest } from "../prompts/manifest";
|
|
20
22
|
/** The upload transport `save` fires content through — injected so the SDK's
|
|
21
23
|
* networked implementation and the tests share one seam. */
|
|
22
24
|
export type ArtifactContentUploader = (input: ContentUploadInput) => Promise<ContentUploadOutcome>;
|
|
23
25
|
/** Per-rig content-custody wiring passed to the artifact api (E2-C). Absent =
|
|
24
26
|
* metadata + hash only, today's behaviour. */
|
|
25
27
|
export interface ArtifactContentDeps {
|
|
26
|
-
|
|
28
|
+
/** Per-task scrub resolver (settings-plane §1.5; P15 CP-3b). The control-plane
|
|
29
|
+
* task/account setting WINS — scrub is an account setting, not a client env var
|
|
30
|
+
* (`MODELRIG_ARTIFACTS_SCRUB` was deleted). null for local-only / direct rigs
|
|
31
|
+
* that have no control plane, and for a run with no task identity — then
|
|
32
|
+
* `defaultScrub` governs. */
|
|
33
|
+
readonly resolver: TaskSettingsResolver | null;
|
|
34
|
+
/** Fallback scrub mode when no resolver answers for the task. The content-on
|
|
35
|
+
* default (`detect`). */
|
|
36
|
+
readonly defaultScrub: ScrubMode;
|
|
27
37
|
readonly upload: ArtifactContentUploader;
|
|
28
38
|
}
|
|
29
39
|
export interface ArtifactHandle {
|
|
@@ -55,13 +65,14 @@ export interface ArtifactSaveOptions {
|
|
|
55
65
|
readonly supersedes?: ArtifactHandle;
|
|
56
66
|
readonly expiresAt?: string;
|
|
57
67
|
}
|
|
58
|
-
|
|
59
|
-
|
|
60
|
-
|
|
61
|
-
|
|
62
|
-
readonly
|
|
63
|
-
readonly
|
|
64
|
-
readonly
|
|
68
|
+
/** Options for {@link ArtifactApi.saveManifest} — the grouping keys the manifest
|
|
69
|
+
* artifact inherits from its prompt (task/step), mirroring a normal save. */
|
|
70
|
+
export interface ArtifactManifestOptions {
|
|
71
|
+
/** Artifact name; defaults to `<prompt name>.manifest` / `"manifest"`. */
|
|
72
|
+
readonly name?: string;
|
|
73
|
+
readonly task?: string;
|
|
74
|
+
readonly stepKey?: string;
|
|
75
|
+
readonly attemptNo?: number;
|
|
65
76
|
}
|
|
66
77
|
export interface ArtifactPromoteOptions {
|
|
67
78
|
readonly note?: string;
|
|
@@ -80,12 +91,20 @@ export interface ArtifactApi {
|
|
|
80
91
|
* a handle. Returns null when the namespace is disabled or a zeroRetention run
|
|
81
92
|
* refuses it (fail-closed). */
|
|
82
93
|
save(value: unknown, opts: ArtifactSaveOptions): ArtifactHandle | null;
|
|
94
|
+
/** PSR-1: attach a **structure-only** assembly manifest (spec §2.1) to a
|
|
95
|
+
* `prompt` artifact. Validates the manifest, saves it as a linked free-text
|
|
96
|
+
* `type: "manifest"` artifact (ZERO-DDL — the `type` column is open), and
|
|
97
|
+
* links `prompt → manifest` with a `derived_from` edge so the manifest joins
|
|
98
|
+
* the existing prompt lineage and stays addressable for PSR-2 ingest. Returns
|
|
99
|
+
* the manifest handle, or null when the manifest is invalid (a counted,
|
|
100
|
+
* teaching warn — never throws) or custody/zeroRetention refuses the save.
|
|
101
|
+
* A no-op unless explicitly called, so `save` stays byte-identical (the emit
|
|
102
|
+
* is inert without a manifest). */
|
|
103
|
+
saveManifest(prompt: ArtifactHandle | string, manifest: Manifest, opts?: ArtifactManifestOptions): ArtifactHandle | null;
|
|
83
104
|
/** Mark an artifact consumed by the current run (feeds derived_from lineage). */
|
|
84
105
|
load(ref: ArtifactHandle | string): void;
|
|
85
106
|
/** Explicit lineage edge the runtime cannot infer. */
|
|
86
107
|
link(from: ArtifactHandle | string, to: ArtifactHandle | string, kind: "consumed" | "derived_from"): void;
|
|
87
|
-
/** Record an evaluation against an artifact. */
|
|
88
|
-
evaluate(ref: ArtifactHandle | string, opts: ArtifactEvaluateOptions): void;
|
|
89
108
|
/** Promote an artifact into the org's durable eval suite (E4 chunk A) — a
|
|
90
109
|
* regression case replayable via `bakeoff --from-eval-cases`. `corrected` set
|
|
91
110
|
* ⇒ a golden-corrected pair; absent ⇒ a failure-case. The `task` binds the
|
|
@@ -94,7 +113,7 @@ export interface ArtifactApi {
|
|
|
94
113
|
* Returns the new case id on success. */
|
|
95
114
|
promoteToEval(ref: ArtifactHandle | string, opts?: ArtifactPromoteOptions): string | null;
|
|
96
115
|
/** Run `fn` inside a single store transaction and return its value (H8), so a
|
|
97
|
-
* step's several save/link
|
|
116
|
+
* step's several save/link writes commit with ONE fsync. On a
|
|
98
117
|
* disabled/inert namespace `fn` runs directly (no store, no transaction). */
|
|
99
118
|
batch<T>(fn: () => T): T;
|
|
100
119
|
/** Await every in-flight content upload (E2-C). Content upload is fire-and-
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"artifact.d.ts","sourceRoot":"","sources":["../../src/artifacts/artifact.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;GAcG;AAGH,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,6BAA6B,CAAC;AACjE,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,oBAAoB,CAAC;AAGzD,OAAO,EAAgC,KAAK,SAAS,EAAE,MAAM,SAAS,CAAC;
|
|
1
|
+
{"version":3,"file":"artifact.d.ts","sourceRoot":"","sources":["../../src/artifacts/artifact.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;GAcG;AAGH,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,6BAA6B,CAAC;AACjE,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,oBAAoB,CAAC;AAGzD,OAAO,EAAgC,KAAK,SAAS,EAAE,MAAM,SAAS,CAAC;AAEvE,OAAO,EAAsB,KAAK,oBAAoB,EAAE,MAAM,2BAA2B,CAAC;AAC1F,OAAO,KAAK,EAAE,kBAAkB,EAAE,oBAAoB,EAAE,MAAM,oBAAoB,CAAC;AACnF,OAAO,EAAoB,KAAK,QAAQ,EAAE,MAAM,qBAAqB,CAAC;AAEtE;4DAC4D;AAC5D,MAAM,MAAM,uBAAuB,GAAG,CAAC,KAAK,EAAE,kBAAkB,KAAK,OAAO,CAAC,oBAAoB,CAAC,CAAC;AAEnG;8CAC8C;AAC9C,MAAM,WAAW,mBAAmB;IAClC;;;;iCAI6B;IAC7B,QAAQ,CAAC,QAAQ,EAAE,oBAAoB,GAAG,IAAI,CAAC;IAC/C;6BACyB;IACzB,QAAQ,CAAC,YAAY,EAAE,SAAS,CAAC;IACjC,QAAQ,CAAC,MAAM,EAAE,uBAAuB,CAAC;CAC1C;AAyBD,MAAM,WAAW,cAAc;IAC7B,QAAQ,CAAC,EAAE,EAAE,MAAM,CAAC;IACpB,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,OAAO,EAAE,MAAM,CAAC;CAC1B;AAED,MAAM,WAAW,mBAAmB;IAClC,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,IAAI,CAAC,EAAE,MAAM,CAAC;IACvB;;;;;iFAK6E;IAC7E,QAAQ,CAAC,OAAO,CAAC,EAAE,MAAM,CAAC;IAC1B;gFAC4E;IAC5E,QAAQ,CAAC,SAAS,CAAC,EAAE,MAAM,CAAC;IAC5B,QAAQ,CAAC,cAAc,CAAC,EAAE,MAAM,CAAC;IACjC,yFAAyF;IACzF,QAAQ,CAAC,UAAU,CAAC,EAAE,MAAM,CAAC;IAC7B,kFAAkF;IAClF,QAAQ,CAAC,MAAM,CAAC,EAAE,MAAM,GAAG,IAAI,CAAC;IAChC,gFAAgF;IAChF,QAAQ,CAAC,UAAU,CAAC,EAAE,cAAc,CAAC;IACrC,QAAQ,CAAC,SAAS,CAAC,EAAE,MAAM,CAAC;CAC7B;AAED;6EAC6E;AAC7E,MAAM,WAAW,uBAAuB;IACtC,0EAA0E;IAC1E,QAAQ,CAAC,IAAI,CAAC,EAAE,MAAM,CAAC;IACvB,QAAQ,CAAC,IAAI,CAAC,EAAE,MAAM,CAAC;IACvB,QAAQ,CAAC,OAAO,CAAC,EAAE,MAAM,CAAC;IAC1B,QAAQ,CAAC,SAAS,CAAC,EAAE,MAAM,CAAC;CAC7B;AAED,MAAM,WAAW,sBAAsB;IACrC,QAAQ,CAAC,IAAI,CAAC,EAAE,MAAM,CAAC;IACvB;;gDAE4C;IAC5C,QAAQ,CAAC,IAAI,CAAC,EAAE,MAAM,CAAC;IACvB,qEAAqE;IACrE,QAAQ,CAAC,SAAS,CAAC,EAAE,cAAc,GAAG,MAAM,CAAC;IAC7C;uCACmC;IACnC,QAAQ,CAAC,MAAM,CAAC,EAAE,cAAc,CAAC;CAClC;AAED,MAAM,WAAW,WAAW;IAC1B;;mCAE+B;IAC/B,IAAI,CAAC,KAAK,EAAE,OAAO,EAAE,IAAI,EAAE,mBAAmB,GAAG,cAAc,GAAG,IAAI,CAAC;IACvE;;;;;;;;uCAQmC;IACnC,YAAY,CAAC,MAAM,EAAE,cAAc,GAAG,MAAM,EAAE,QAAQ,EAAE,QAAQ,EAAE,IAAI,CAAC,EAAE,uBAAuB,GAAG,cAAc,GAAG,IAAI,CAAC;IACzH,iFAAiF;IACjF,IAAI,CAAC,GAAG,EAAE,cAAc,GAAG,MAAM,GAAG,IAAI,CAAC;IACzC,sDAAsD;IACtD,IAAI,CAAC,IAAI,EAAE,cAAc,GAAG,MAAM,EAAE,EAAE,EAAE,cAAc,GAAG,MAAM,EAAE,IAAI,EAAE,UAAU,GAAG,cAAc,GAAG,IAAI,CAAC;IAC1G;;;;;6CAKyC;IACzC,aAAa,CAAC,GAAG,EAAE,cAAc,GAAG,MAAM,EAAE,IAAI,CAAC,EAAE,sBAAsB,GAAG,MAAM,GAAG,IAAI,CAAC;IAC1F;;iFAE6E;IAC7E,KAAK,CAAC,CAAC,EAAE,EAAE,EAAE,MAAM,CAAC,GAAG,CAAC,CAAC;IACzB;;kFAE8E;IAC9E,YAAY,IAAI,OAAO,CAAC,IAAI,CAAC,CAAC;IAC9B;;;;;;0DAMsD;IACtD,mBAAmB,IAAI,OAAO,CAAC,IAAI,CAAC,CAAC;IACrC;8EAC0E;IAC1E,QAAQ,IAAI;QACV,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;QACvB,QAAQ,CAAC,aAAa,EAAE,MAAM,CAAC;QAC/B,QAAQ,CAAC,oBAAoB,EAAE,MAAM,CAAC;QACtC,8CAA8C;QAC9C,QAAQ,CAAC,eAAe,EAAE,MAAM,CAAC;QACjC,6EAA6E;QAC7E,QAAQ,CAAC,eAAe,EAAE,MAAM,CAAC;QACjC,yDAAyD;QACzD,QAAQ,CAAC,cAAc,EAAE,MAAM,CAAC;QAChC;;0EAEkE;QAClE,QAAQ,CAAC,cAAc,EAAE,MAAM,CAAC;QAChC,oEAAoE;QACpE,QAAQ,CAAC,aAAa,EAAE,MAAM,CAAC;KAChC,CAAC;CACH;AA2CD,wBAAgB,iBAAiB,CAAC,IAAI,EAAE;IACtC,QAAQ,CAAC,KAAK,EAAE,aAAa,CAAC;IAC9B,QAAQ,CAAC,OAAO,EAAE,OAAO,CAAC;IAC1B,oEAAoE;IACpE,QAAQ,CAAC,OAAO,CAAC,EAAE,mBAAmB,CAAC;CACxC,GAAG,WAAW,CAuXd"}
|