feat(hosted): hostedTenantFromEnv — env→HostedTenant for selfImprove (#227)

selfImprove takes a HostedTenant config; products read hosted config from env.
hostedClientFromEnv already had the env→tenant logic but returns a client, not
the config. Extract hostedTenantFromEnv (hostedClientFromEnv now composes it) so
a product passes hostedTenant: hostedTenantFromEnv({ tenantId }) when collapsing
its loop onto selfImprove — instead of hand-rolling the env map. Fails soft
(undefined) when unconfigured.

chore(release): 0.83.0 (lockstep). Also genericize the 0.81.0 changelog entry.

Author: drewstone

CHANGELOG.md

@@ -4,6 +4,14 @@ All notable changes to `@tangle-network/agent-eval` and its sibling `agent-eval-
 
 ---
 
+## [0.83.0] — 2026-06-05 — hostedTenantFromEnv
+
+### Added
+
+- **`hostedTenantFromEnv` (`/hosted`).** Builds a `HostedTenant` config from env (the input `selfImprove({ hostedTenant })` and `emitLoopProvenance` take), with the same env precedence + overrides as `hostedClientFromEnv` — which now composes it. Returns `undefined` (not an error) when unconfigured, so a product wires `hostedTenant: hostedTenantFromEnv({ tenantId: 'my-agent' })` unconditionally and hosted ingest stays off until the env is set. Removes the env→tenant mapping every product would otherwise hand-roll when collapsing onto `selfImprove`.
+
+---
+
 ## [0.82.0] — 2026-06-05 — selfImprove forwards the full loop surface
 
 ### Changed
@@ -21,7 +29,7 @@ A deterministic offline test that drives `selfImprove` with a mock agent must no
 
 ### Added
 
-- **`aggregateJudgeVerdicts<D>` (root).** Generic judge-ensemble reducer: fan out N uncorrelated judges, mean each rubric dimension over the SURVIVORS, report the inter-rater disagreement spread, sum cost. Replaces the same reduction hand-rolled in legal (`aggregateEnsemble`), creative (`production-loop/judges.ts`), and tax (`judge-ensemble.ts`). Fail-loud: a failed judge (`perDimension: null`) is recorded in `failedJudges`, never folded into a zero; all-failed throws; a failed judge's cost is still summed. Composite reuses `weightedComposite`.
+- **`aggregateJudgeVerdicts<D>` (root).** Generic judge-ensemble reducer: fan out N uncorrelated judges, mean each rubric dimension over the SURVIVORS, report the inter-rater disagreement spread, sum cost. Replaces the same reduction hand-rolled across multiple product agents. Fail-loud: a failed judge (`perDimension: null`) is recorded in `failedJudges`, never folded into a zero; all-failed throws; a failed judge's cost is still summed. Composite reuses `weightedComposite`.
 - **`createTokenRecallChecker` (root).** The deterministic, no-LLM `CorrectnessChecker` — sibling of `createLlmCorrectnessChecker`. A produced item fulfils a requirement when its content is substantive and recalls ≥ `minRecall` of the requirement title's significant tokens. The default completion gate for apps/tests without an LLM judge.
 - **`ErrorCluster` (root + `/analyst`).** The failure-cluster element type is now a named export, so consumers import it instead of deriving `DatasetOverview['error_clusters'][number]`.
 

clients/python/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "agent-eval-rpc"
-version = "0.82.0"
+version = "0.83.0"
 description = "Python RPC client for @tangle-network/agent-eval — judge content against rubrics over HTTP or stdio RPC. Eval logic runs in the Node runtime; this package is a thin wire client."
 readme = "README.md"
 requires-python = ">=3.10"

clients/python/src/agent_eval_rpc/__init__.py

@@ -58,7 +58,7 @@
 try:
     __version__ = version("agent-eval-rpc")
 except PackageNotFoundError:
-    __version__ = "0.82.0"
+    __version__ = "0.83.0"
 
 __all__ = [
     "Client",

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tangle-network/agent-eval",
-  "version": "0.82.0",
+  "version": "0.83.0",
   "description": "Evaluate and improve AI agents from runs, traces, judges, and feedback. Compare candidates, cluster failures, measure lift, and gate releases.",
   "homepage": "https://github.com/tangle-network/agent-eval#readme",
   "repository": {

src/hosted/client.ts

@@ -161,9 +161,17 @@ export function createHostedClient(tenant: HostedTenant): HostedClient {
  * A trailing slash on the endpoint is stripped. Pass `overrides` to supply any
  * field directly (e.g. a fixed `tenantId` per product) — overrides win over env.
  */
-export function hostedClientFromEnv(
+/**
+ * Build a {@link HostedTenant} config from env — the input `selfImprove`'s
+ * `hostedTenant` and `emitLoopProvenance` take. Same env precedence + overrides
+ * as {@link hostedClientFromEnv}; returns `undefined` (not an error) when any of
+ * endpoint / apiKey / tenantId is missing, so a product wires
+ * `hostedTenant: hostedTenantFromEnv({ tenantId: 'my-agent' })` unconditionally
+ * and it stays off until the env is set.
+ */
+export function hostedTenantFromEnv(
   overrides: Partial<HostedTenant> & { env?: Record<string, string | undefined> } = {},
-): HostedClient | undefined {
+): HostedTenant | undefined {
   const env = overrides.env ?? process.env
   const endpoint = (
     overrides.endpoint ??
@@ -177,5 +185,12 @@ export function hostedClientFromEnv(
   if (overrides.fetchImpl) tenant.fetchImpl = overrides.fetchImpl
   if (overrides.timeoutMs !== undefined) tenant.timeoutMs = overrides.timeoutMs
   if (overrides.retries !== undefined) tenant.retries = overrides.retries
-  return createHostedClient(tenant)
+  return tenant
+}
+
+export function hostedClientFromEnv(
+  overrides: Partial<HostedTenant> & { env?: Record<string, string | undefined> } = {},
+): HostedClient | undefined {
+  const tenant = hostedTenantFromEnv(overrides)
+  return tenant ? createHostedClient(tenant) : undefined
 }

src/hosted/from-env.test.ts

@@ -0,0 +1,44 @@
+/**
+ * `hostedTenantFromEnv` is the env→HostedTenant map every product passes to
+ * `selfImprove({ hostedTenant })`. It must fail SOFT (undefined, not throw) when
+ * unconfigured, so a product wires it unconditionally and it stays off until the
+ * env is set.
+ */
+
+import { describe, expect, it } from 'vitest'
+
+import { hostedClientFromEnv, hostedTenantFromEnv } from './client'
+
+describe('hostedTenantFromEnv', () => {
+  const fullEnv = {
+    TANGLE_INGEST_URL: 'https://orchestrator.example/v1/',
+    TANGLE_API_KEY: 'k-123',
+    TANGLE_TENANT_ID: 'acme',
+  }
+
+  it('builds a tenant from env and strips the trailing slash', () => {
+    const t = hostedTenantFromEnv({ env: fullEnv })
+    expect(t).toEqual({
+      endpoint: 'https://orchestrator.example/v1',
+      apiKey: 'k-123',
+      tenantId: 'acme',
+    })
+  })
+
+  it('returns undefined (not an error) when any required field is missing', () => {
+    expect(hostedTenantFromEnv({ env: {} })).toBeUndefined()
+    expect(
+      hostedTenantFromEnv({ env: { TANGLE_INGEST_URL: 'x', TANGLE_API_KEY: 'y' } }),
+    ).toBeUndefined()
+  })
+
+  it('overrides win over env (e.g. a fixed per-product tenantId)', () => {
+    const t = hostedTenantFromEnv({ env: fullEnv, tenantId: 'my-agent' })
+    expect(t?.tenantId).toBe('my-agent')
+  })
+
+  it('hostedClientFromEnv composes it — undefined env ⇒ no client', () => {
+    expect(hostedClientFromEnv({ env: {} })).toBeUndefined()
+    expect(hostedClientFromEnv({ env: fullEnv })).toBeDefined()
+  })
+})

src/hosted/index.ts

@@ -14,6 +14,7 @@ export {
   type HostedClient,
   type HostedTenant,
   hostedClientFromEnv,
+  hostedTenantFromEnv,
 } from './client'
 export {
   type EvalRunCellScore,