@wenathlan/getry 2.0.165 → 2.0.167

package/docs/distributionlanes.md

@@ -6,7 +6,7 @@ The eleven distribution lanes of the PR14 law and the honest status each lane an
 | --- | --- | --- | --- |
 | npmjs | package.json — @wenathlan/getry | live | the family-npmjs lane publishes the envelope every release rung |
 | GitHub Packages npm | package.json — @wenathlan/getry | live | the family-githubnpm lane publishes the same envelope idempotently |
-| Maven | pom.xml — io.github.wenathlan:getry | live | the family-maven lane deploys the jar |
+| Maven | pom.xml — com.wenathlan:getry | live | the family-maven lane deploys the jar |
 | NuGet | getry.csproj | live | the family-nuget lane packs the envelope |
 | Gradle | build.gradle.kts — com.wenathlan:getry | live | the family-gradle lane publishes the module |
 | Ruby Gems | getry.gemspec | live | the family-rubygems lane pushes the gem |

package/docs/features.md

@@ -16,8 +16,7 @@ every feature below is read from the code of the getry app root and the Sol them
 - cors handling — preflight and cors headers on every route.
 - response formats — chat completion, anthropic message and response shapes built from one pipeline.
 - gateway cli — `cli.ts` scaffolds and inspects gateway configs.
-- gateway probe — `gatewayprobe.ts` runs a manual chat round trip against a deployed host.
-- container smoke — `gatewaysmoke.ts` runs one full offline request cycle against the recorded openaicompat fixture.
+- gateway check — `gatewaycheck.ts` folds both check seats: the offline smoke (one full request cycle against the recorded openaicompat fixture) and the manual probe (a chat round trip against a deployed host).
 
 ## providers
 

package/docs/gateway.md

@@ -29,8 +29,7 @@
 - `oauth.ts` / `accounts.ts` / `recovery.ts` / `project.ts` — the provider oauth family, multi-account manager, recovery and project id discovery.
 - `db.ts` — the database contract: typed rows, in-memory seeds, HTTPS accessors, the health-aware round-robin key picker (`nextkeyrow`) and the better-sqlite3 mirror (onboarding states, labels and the `gateway_versions` route table).
 - `prismaconfig.ts` / `prismashim.ts` / `nodeshim.ts` — prisma schema path resolution and runtime shims.
-- `gatewayprobe.ts` — manual chat round trip against a deployed gateway (`GET /v1/models`, `POST /v1/chat/completions`).
-- `gatewaysmoke.ts` — offline container smoke: one full request cycle against the recorded openaicompat fixture.
+- `gatewaycheck.ts` — the two gateway check seats in one file: the offline container smoke (one full request cycle against the recorded openaicompat fixture) and the manual probe (chat round trip against a deployed gateway, `GET /v1/models`, `POST /v1/chat/completions`).
 - `registernvidiakeys.ts` — registers the 22 nvidia api keys into the `nvidiaKey` table.
 - `cli.ts` / `providercli.ts` — the gateway cli and the provider cli.
 - `versionregistry.ts`, `request.ts`, `search.ts`, `fingerprint.ts`, `constants.ts`, `antigravity.ts` — the antigravity/gemini-cli provider lineage: version resolution, request building, search, fingerprinting, endpoint constants.
@@ -122,4 +121,4 @@ the 7-level thinking ladder (`thinking.ts`), each level bound to a token budget:
 
 ## tests
 
-- 29 test files under `tests/` cover the gateway surface: gateway, gatewayversions (the 5/35 route table law, the sqlite mirror rows, the thinking chain), server (auth, engine, property, utils), llm, modelroute, promptlibrary, keyrotation (the health-aware key round-robin and the masking discipline), routerotation, thinkingladder (the registerable ladder and the tier/budget/chain/lane routing), streamfallback, intelligence, jsonpolicy, family, design, cli, providercli, onboardingstate, sandboxmodality, debonairlane, torture.llm, plus the provider lineage (constants, fingerprint, models, oauth) and the shared fixtures (`gatewayserver.ts`, `providerfixtures.ts`, `researchfixtures.ts`).
+- 29 test files under `test/` cover the gateway surface: gateway, gatewayversions (the 5/35 route table law, the sqlite mirror rows, the thinking chain), server (auth, engine, property, utils), llm, modelroute, promptlibrary, keyrotation (the health-aware key round-robin and the masking discipline), routerotation, thinkingladder (the registerable ladder and the tier/budget/chain/lane routing), streamfallback, intelligence, jsonpolicy, family, design, cli, providercli, onboardingstate, sandboxmodality, debonairlane, torture.llm, plus the provider lineage (constants, fingerprint, models, oauth) and the shared fixtures (`gatewayserver.ts`, `providerfixtures.ts`, `researchfixtures.ts`).

package/docs/getry.md

@@ -0,0 +1,33 @@
+# getry — the dossier
+
+The AI gateway library of the family: the key rotation, the reasoning levels and the provider authentication.
+
+## Identity
+
+- Category: AI gateway library.
+- Competitor space: OpenAI-compatible gateway surfaces, opencode.
+- Consumer law: devthink pulls the gateway from the getry package — the AI logic never duplicates inside a sibling.
+
+## Objective
+
+Mask every upstream provider behind the house name and answer one honest gateway: five provider gateways serving thirty-five OpenAI-compatible routes with the reasoning ladder.
+
+## Features
+
+- Five provider gateways serving thirty-five OpenAI-compatible routes.
+- The reasoning ladder: seven thinking levels the requests ride.
+- Key rotation across providers.
+- Provider authentication seats: google, github, gitlab (the authorize, scope, token, profile, env and pkce seats).
+- The meta model: upstream providers masked behind the house name, free models fused into million-token windows, mid-session model rotation.
+- The inference pipeline: porter (zero-cost rulers), the three-member local jury, the anti-sycophancy rule.
+- Universal parameter autocorrection: never a whitelist, always the correction.
+
+## Interactions
+
+- The chat console with the family chat design (the same tokens, titlebar and dpi ladder).
+- The gateway API surface consumed by devthink and the family.
+
+## Doctrine
+
+- One gateway, one owner: the AI seam lives here; the family imports it.
+- The honesty law: the gateway never pretends a model it did not call.