Replaying a VoltAgent run with the model server off — plus the AI SDK v4/v5 mismatch that got me first #1413
xizhuomengcontin
started this conversation in
Show and tell
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Verified against @voltagent/core 0.1.86 today: a VoltAgent
Agentrun can be recorded from outside the process and replayed with no model server present.Disclosure: I work on the recording tool (OrcaReplay, Apache-2.0, no paid tier).
First, the thing that cost me ten minutes
A fresh
npm i @voltagent/core @voltagent/vercel-ai @ai-sdk/openai airesolves@ai-sdk/openaito v4, while@voltagent/vercel-aibundles AI SDK 5. The result is a runtime error that does not name the real cause:npm i "@ai-sdk/openai@^2"fixes it (2.0.127 here). Worth a line in the quickstart or a peer-dependency constraint — the error points at the model, not at the provider package version, so the natural next move is to blame your model id.The run
Nothing is installed into VoltAgent — no hook, no custom provider. The AI SDK provider reads
OPENAI_BASE_URLand the recorder sets it for the child process.What it is for
Re-running an agent while you change instructions, tools or memory: the model's responses stay fixed, so the difference you observe is your change and not the sampler. Also CI without keys or spend, and chasing one bad run repeatedly for free.
Scope
egress=blockedon replay means model-provider egress. Recorded tool calls still execute for real, so it is not a sandbox. And it captures the HTTP going to the model, not VoltAgent's internals — not a substitute for VoltOps.All reactions