Summary
After upgrading OpenClaw to 2026.4.14, our deployment became unable to reply because the configured context engine lossless-claw failed during initialization with an id validation error.
The key error is:
Context engine "lossless-claw" factory returned an invalid ContextEngine: info.id must match registered id "lossless-claw".
This affected multiple lanes, caused embedded agents to fail before reply, and made model fallback irrelevant because the failure happened prior to model execution.
We had to perform a rollback to 2026.4.12 to restore service.
Environment
- OpenClaw version that breaks: 2026.4.14
- Rolled back to working version: 2026.4.12
- Context engine configured as:
slots.contextEngine = "lossless-claw"
- Plugin involved:
lossless-claw
What happened
After the upgrade, agents stopped being able to respond. Logs showed repeated failures across multiple lanes, for example:
lane task error: lane=session:agent:agent1:telegram:direct:<redacted> durationMs=10749 error="Error: Context engine \"lossless-claw\" factory returned an invalid ContextEngine: info.id must match registered id \"lossless-claw\"."
lane task error: lane=main durationMs=3297 error="Error: Context engine \"lossless-claw\" factory returned an invalid ContextEngine: info.id must match registered id \"lossless-claw\"."
Embedded agent failed before reply: Context engine "lossless-claw" factory returned an invalid ContextEngine: info.id must match registered id "lossless-claw".
There were also model fallback attempts (glm-4.7 -> glm-4.5-air -> gpt-5.4 -> glm-5-turbo), but they all failed with the same underlying error because the crash occurs before model execution.
Important observation
This does not look like a model problem. It appears to be a context-engine validation / compatibility regression.
The failure affects agent initialization itself, including lanes similar to:
lane=main
lane=session:agent:agent1:telegram:direct:<redacted>
lane=session:agent:agent2:telegram:direct:<redacted>
Recovery / workaround
We restored service in one of these ways:
- Rollback OpenClaw from 2026.4.14 to 2026.4.12
- Alternatively, switching from:
slots.contextEngine = "lossless-claw"
to:
slots.contextEngine = "legacy"
Using legacy instead of lossless-claw avoided the failure.
Expected behavior
lossless-claw should continue to initialize correctly on 2026.4.14, or OpenClaw should provide a clearer compatibility error / migration path instead of breaking agent startup.
Actual behavior
lossless-claw fails validation because the returned ContextEngine.info.id does not match the registered plugin id lossless-claw, which prevents agents from initializing and replying.
Request
Could you please confirm whether this is:
- a regression in OpenClaw 2026.4.14,
- a compatibility break with the current
lossless-claw,
- or a required migration/update on the plugin side?
If there is already a compatible lossless-claw version or a migration step required for 2026.4.14, that would be very helpful to know.
Summary
After upgrading OpenClaw to 2026.4.14, our deployment became unable to reply because the configured context engine
lossless-clawfailed during initialization with an id validation error.The key error is:
This affected multiple lanes, caused embedded agents to fail before reply, and made model fallback irrelevant because the failure happened prior to model execution.
We had to perform a rollback to 2026.4.12 to restore service.
Environment
slots.contextEngine = "lossless-claw"lossless-clawWhat happened
After the upgrade, agents stopped being able to respond. Logs showed repeated failures across multiple lanes, for example:
There were also model fallback attempts (
glm-4.7->glm-4.5-air->gpt-5.4->glm-5-turbo), but they all failed with the same underlying error because the crash occurs before model execution.Important observation
This does not look like a model problem. It appears to be a context-engine validation / compatibility regression.
The failure affects agent initialization itself, including lanes similar to:
lane=mainlane=session:agent:agent1:telegram:direct:<redacted>lane=session:agent:agent2:telegram:direct:<redacted>Recovery / workaround
We restored service in one of these ways:
slots.contextEngine = "lossless-claw"to:
slots.contextEngine = "legacy"Using
legacyinstead oflossless-clawavoided the failure.Expected behavior
lossless-clawshould continue to initialize correctly on 2026.4.14, or OpenClaw should provide a clearer compatibility error / migration path instead of breaking agent startup.Actual behavior
lossless-clawfails validation because the returnedContextEngine.info.iddoes not match the registered plugin idlossless-claw, which prevents agents from initializing and replying.Request
Could you please confirm whether this is:
lossless-claw,If there is already a compatible
lossless-clawversion or a migration step required for 2026.4.14, that would be very helpful to know.