[codex] Skip unreadable events during pagination

[neubig]

Summary

• skip unreadable persisted event payloads while paginating/searching/counting conversation events
• log a warning with the conversation id and event index when an event file is missing, empty, non-UTF-8, or fails Pydantic validation
• add regression coverage for a zero-byte event file between valid events

Root Cause

A zero-byte persisted event file can remain in the on-disk EventLog after an interrupted write. The event search path reads by index for pagination performance, so one unreadable event currently raises and makes /events/search return 500. The event websocket uses the same search path for replay, so affected conversations repeatedly disconnect even though the backend itself is healthy.

Validation

• uv run pytest tests/agent_server/test_event_service.py -q
• uv run pre-commit run --files openhands-agent-server/openhands/agent_server/event_service.py tests/agent_server/test_event_service.py

---

Agent Server images for this PR

• GHCR package: https://github.com/OpenHands/agent-sdk/pkgs/container/agent-server

Variants & Base Images

Variant	Architectures	Base Image	Docs / Tags
java	amd64, arm64	eclipse-temurin:17-jdk	Link
python	amd64, arm64	nikolaik/python-nodejs:python3.13-nodejs22-slim	Link
golang	amd64, arm64	golang:1.21-bookworm	Link

Pull (multi-arch manifest)

# Each variant is a multi-arch manifest supporting both amd64 and arm64
docker pull ghcr.io/openhands/agent-server:a163ce7-python

Run

docker run -it --rm \
  -p 8000:8000 \
  --name agent-server-a163ce7-python \
  ghcr.io/openhands/agent-server:a163ce7-python

All tags pushed for this build

ghcr.io/openhands/agent-server:a163ce7-golang-amd64
ghcr.io/openhands/agent-server:a163ce740049a0cb3a85b27a5e93bd9908756352-golang-amd64
ghcr.io/openhands/agent-server:codex-tolerate-corrupt-event-files-golang-amd64
ghcr.io/openhands/agent-server:a163ce7-golang_tag_1.21-bookworm-amd64
ghcr.io/openhands/agent-server:a163ce7-golang-arm64
ghcr.io/openhands/agent-server:a163ce740049a0cb3a85b27a5e93bd9908756352-golang-arm64
ghcr.io/openhands/agent-server:codex-tolerate-corrupt-event-files-golang-arm64
ghcr.io/openhands/agent-server:a163ce7-golang_tag_1.21-bookworm-arm64
ghcr.io/openhands/agent-server:a163ce7-java-amd64
ghcr.io/openhands/agent-server:a163ce740049a0cb3a85b27a5e93bd9908756352-java-amd64
ghcr.io/openhands/agent-server:codex-tolerate-corrupt-event-files-java-amd64
ghcr.io/openhands/agent-server:a163ce7-eclipse-temurin_tag_17-jdk-amd64
ghcr.io/openhands/agent-server:a163ce7-java-arm64
ghcr.io/openhands/agent-server:a163ce740049a0cb3a85b27a5e93bd9908756352-java-arm64
ghcr.io/openhands/agent-server:codex-tolerate-corrupt-event-files-java-arm64
ghcr.io/openhands/agent-server:a163ce7-eclipse-temurin_tag_17-jdk-arm64
ghcr.io/openhands/agent-server:a163ce7-python-amd64
ghcr.io/openhands/agent-server:a163ce740049a0cb3a85b27a5e93bd9908756352-python-amd64
ghcr.io/openhands/agent-server:codex-tolerate-corrupt-event-files-python-amd64
ghcr.io/openhands/agent-server:a163ce7-nikolaik_s_python-nodejs_tag_python3.13-nodejs22-slim-amd64
ghcr.io/openhands/agent-server:a163ce7-python-arm64
ghcr.io/openhands/agent-server:a163ce740049a0cb3a85b27a5e93bd9908756352-python-arm64
ghcr.io/openhands/agent-server:codex-tolerate-corrupt-event-files-python-arm64
ghcr.io/openhands/agent-server:a163ce7-nikolaik_s_python-nodejs_tag_python3.13-nodejs22-slim-arm64
ghcr.io/openhands/agent-server:a163ce7-golang
ghcr.io/openhands/agent-server:a163ce740049a0cb3a85b27a5e93bd9908756352-golang
ghcr.io/openhands/agent-server:codex-tolerate-corrupt-event-files-golang
ghcr.io/openhands/agent-server:a163ce7-golang_tag_1.21-bookworm
ghcr.io/openhands/agent-server:a163ce7-java
ghcr.io/openhands/agent-server:a163ce740049a0cb3a85b27a5e93bd9908756352-java
ghcr.io/openhands/agent-server:codex-tolerate-corrupt-event-files-java
ghcr.io/openhands/agent-server:a163ce7-eclipse-temurin_tag_17-jdk
ghcr.io/openhands/agent-server:a163ce7-python
ghcr.io/openhands/agent-server:a163ce740049a0cb3a85b27a5e93bd9908756352-python
ghcr.io/openhands/agent-server:codex-tolerate-corrupt-event-files-python
ghcr.io/openhands/agent-server:a163ce7-nikolaik_s_python-nodejs_tag_python3.13-nodejs22-slim

About Multi-Architecture Support

• Each variant tag (e.g., a163ce7-python) is a multi-arch manifest supporting both amd64 and arm64
• Docker automatically pulls the correct architecture for your platform
• Individual architecture tags (e.g., a163ce7-python-amd64) are also available if needed

Issue

Fixes #3761.

[github-actions]

Python API breakage checks — ✅ PASSED

Result: ✅ PASSED

Action log

[github-actions]

REST API breakage checks (OpenAPI) — ✅ PASSED

Result: ✅ PASSED

Action log

[github-actions]

Coverage Report •

File	Stmts	Miss	Cover	Missing
openhands-agent-server/openhands/agent_server
event_service.py	567	113	80%	103–104, 134, 137–138, 142–143, 150, 154, 160, 170–174, 177–180, 251, 272–273, 347, 401, 421, 428, 452–453, 457, 465, 468, 478, 510, 521, 528, 534, 588, 590, 594–596, 600, 609–610, 612, 616, 622, 624, 671, 701, 704, 759, 780, 888, 1000–1003, 1007, 1016, 1018–1019, 1023, 1037–1042, 1044, 1071, 1076–1079, 1083–1086, 1094–1097, 1113, 1130–1133, 1179–1180, 1182–1189, 1191–1192, 1201–1202, 1204–1205, 1212–1213, 1215–1216, 1236, 1242, 1248, 1257–1258
TOTAL	31387	13756	56%