docs(2275): remove non-existent RetryConfig.Builder methods from README [skip-runtime-e2e] (#180)

The README's 'Retry Configuration' section advertised three RetryConfig.Builder
methods that do not exist in the public API:

  - initialDelayMs(int)            — actual Builder has initialDelay(Duration)
  - maxDelayMs(int)                — actual Builder has maxDelay(Duration)
  - retryableStatusCodes(Set<Integer>) — does not exist at all; the retry
                                         policy is hard-coded in
                                         RetryExecutor.isRetryable
                                         (5xx + 429 + connect/timeout, with
                                         401/403/PolicyViolation/Config
                                         exceptions always terminal)

The third method was the most load-bearing in the #2275 context. A user
reading the README would expect to be able to configure which statuses retry.
If they looked for the method and didn't find it, they might write a
workaround that bypassed retryExecutor entirely, reintroducing the storm
that PR #178 added regression coverage for.

Decision: delete the misleading example, not implement the methods.
Implementing retryableStatusCodes would be a feature addition (a new
configuration surface + serialized semantics + tests), not a doc fix.

This PR replaces the offending block with:
  - one paragraph documenting the ACTUAL retry contract from
    RetryExecutor.isRetryable
  - a citation back to #2275 (the regression test from PR #178 locks in
    that 401 stays terminal)
  - a list of every Builder method that ACTUALLY exists (enabled,
    maxAttempts, initialDelay, maxDelay, multiplier) with default values
  - a working example using initialDelay(Duration.ofMillis(100)) /
    maxDelay(Duration.ofSeconds(5)) so the code-block compiles against the
    real API

CHANGELOG entry under [Unreleased] notes this is a documentation-only
fix — no public-API or behavior change.

Lineage:
  - Hostile review of PR #178 (regression test for issue #2275) flagged
    this as HIGH doc-debt. PR #178 author punted.
  - This follow-up PR closes the finding.

Out of scope:
  - Adding retryableStatusCodes or any per-status configuration to the
    Builder (that's a feature).
  - Changing RetryExecutor.isRetryable behavior.
  - Version bump.

Signed-off-by: Saurabh Jain <saurabh.jain@getaxonflow.com>

Author: saurabhjain1592

CHANGELOG.md

@@ -5,6 +5,21 @@ All notable changes to the AxonFlow Java SDK will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [Unreleased]
+
+### Fixed
+
+- **README "Retry Configuration" section.** Removed three `RetryConfig.Builder`
+  methods that do not exist on the public API (`initialDelayMs(int)`,
+  `maxDelayMs(int)`, `retryableStatusCodes(Set<Integer>)`). The example now
+  uses the actual `initialDelay(Duration)` / `maxDelay(Duration)` builders, and
+  the surrounding prose documents the real retry contract from
+  `RetryExecutor.isRetryable`: retries fire on connect/timeout, `5xx`, and
+  `429`; `401`/`403` and other `4xx` are always terminal. Locked in by the
+  regression test added for
+  [getaxonflow/axonflow-enterprise#2275](https://github.com/getaxonflow/axonflow-enterprise/issues/2275)
+  in PR #178. Documentation-only — no code or behavior change.
+
 ## [8.1.0] - 2026-05-19 — `X-Client-ID` header on every outbound request (v9 identity)
 
 **Companion release to the v9 identity cleanup on the platform (Epic #2230).**

README.md

@@ -337,19 +337,35 @@ try {
 
 ## Retry Configuration
 
-The SDK includes automatic retry with exponential backoff:
+The SDK includes automatic retry with exponential backoff. The retry policy
+itself is not configurable per HTTP status — retries fire on connect/timeout
+errors, `5xx` server errors, and `429 Too Many Requests`. Authentication
+failures (`401`/`403`), policy violations, and other `4xx` client errors are
+always terminal and never retried (see `RetryExecutor.isRetryable`). This
+contract is locked in as a regression test for
+[getaxonflow/axonflow-enterprise#2275](https://github.com/getaxonflow/axonflow-enterprise/issues/2275)
+— a retry storm on `401` against a misconfigured agent.
+
+The `RetryConfig.Builder` exposes the following knobs:
+
+- `enabled(boolean)` — turn retries off entirely (defaults to `true`).
+- `maxAttempts(int)` — total attempts including the first, `1`–`10` (default `3`).
+- `initialDelay(Duration)` — base delay before the second attempt (default `1s`).
+- `maxDelay(Duration)` — cap on the exponential backoff (default `30s`).
+- `multiplier(double)` — backoff multiplier, ≥ `1.0` (default `2.0`).
 
 ```java
+import java.time.Duration;
+
 AxonFlowConfig config = AxonFlowConfig.builder()
     .endpoint("https://agent.getaxonflow.com")
     .clientId("your-client-id")
-            .clientSecret("your-client-secret")
+    .clientSecret("your-client-secret")
     .retryConfig(RetryConfig.builder()
         .maxAttempts(3)
-        .initialDelayMs(100)
-        .maxDelayMs(5000)
+        .initialDelay(Duration.ofMillis(100))
+        .maxDelay(Duration.ofSeconds(5))
         .multiplier(2.0)
-        .retryableStatusCodes(Set.of(429, 500, 502, 503, 504))
         .build())
     .build();
 ```