How does the fallback mechanism work? #169
|
Given my current config: Whenever I hit rate limits on my Claude sub, it doesn't fallback into the fallback models specified. Or am I missing something? I tried searching the docs about it but I found no reference. |
Replies: 4 comments 4 replies
|
Yeah, I dont think its working at all |
|
We hit the same issue. The working approach in our OpenCode setup is to put the fallback chain directly on each agent’s Example: {
"fallback": { "enabled": true },
"agents": {
"orchestrator": {
"model": [
"anthropic/claude-opus-4-6",
"google/antigravity-claude-opus-4-6-thinking",
"google/antigravity-gemini-3.1-pro",
"opencode/minimax-m2.5-free",
"opencode/big-pickle"
]
}
}
}The first model is the primary. On a rate-limit/quota/concurrency error, the runtime fallback hook aborts the failed prompt and re-submits the last user message with the next untried model in that agent’s array. So the practical fix is: move each Prompt to have the orchestrator fix this |
|
Answered by @Zaradacht: |
|
But according to authors-preset, the config value for models is string, does this entry actually supports array? |
Answered by @Zaradacht:
fallback.chainsis legacy — use the agentmodelfield as an ordered array instead.