Hi community,
We are encountering a page-blocking "412 Precondition Failed" error when attempting to use or preview any Committed version of our DSL-enabled Agentforce Service Agent (agentDSLEnabled set to true).
Symptoms:
- Draft editor works fine: We can make changes to the agent Draft in the Agentforce Builder without issues.
- Committed versions fail: As soon as we select any Committed version (e.g., Version 1 Committed) in the Builder, the Preview chat panel fails to connect and displays: "Something went wrong. Refresh and try again."
- Console/API Error: Inspecting the browser console shows that the background session call returns:
- Endpoint: /services/data/v61.0/einstein/ai-agent/v1/agents/{agentId}/sessions (also fails with the same error on the external sessions API via Postman).
- Response Code: 412 Precondition Failed
- Error Message: {"message":"Unable to load agent config: Invalid Config"}
Has anyone else run into this issue where the compiler or orchestrator fails to load/resolve the config only for Committed versions? Is this related to Data Cloud retriever target URIs or sandbox metadata sync issues?
Any advice or workarounds would be highly appreciated. Thanks!
#Agentforce
Hi,
Things worth checking given your setup:
1. Since this only fails on Committed versions and not Draft, compare what's different structurally — specifically check if your DSL config references anything environment-specific (Data Cloud retriever target URIs, endpoint variables like the {{DOMAIN}} placeholder visible in your Postman body) that might resolve correctly in the live Draft context but fail to resolve/bind correctly once "frozen" into a Committed version.
2. Check whether the {{DOMAIN}} placeholder in your instanceConfig.endpoint is actually being substituted with a real value before the request goes out — if that's still literally the placeholder string when hitting the API (rather than your actual instance domain), that alone would produce exactly this kind of config validation failure.
3. If you're testing this in a sandbox, confirm metadata sync between the sandbox and any connected Data Cloud org is fully complete — partial/lagging metadata sync between environments is a known category of "config looks fine in UI but fails at runtime" issues in Agentforce, even if I can't point to a specific KB article confirming this exact error string.
If nothing works,
1. File a case with Salesforce Support, explicitly including both pieces of evidence — the native Live Test failure screenshot AND the Postman reproduction — since showing it fails through Salesforce's own UI (not just a custom API call) makes this much harder to dismiss as a client-side config issue on your end.
2. Reference the exact error payload (AgentApiException, requestId, timestamp) from your screenshot — that requestId in particular will let their engineering team pull server-side logs for that specific failed request.
3. Ask specifically whether this is a known issue with agentDSLEnabled + Committed versions, since that's a fairly new/narrow combination and may not be broadly tested yet.
