ChatClientRecordingMode
Enums · AiDotNet.Enums
Says whether chat calls are made live, saved to a store, or served back from one.
For Beginners
This decides whether the library talks to a real AI model, saves what it says, or
plays back what it said last time. Use Record once while you have a working API key, then switch
to Replay: from then on your run repeats exactly, instantly and for free, which is what makes it
usable in automated tests. ReplayWithFallback is the middle ground — it reuses saved answers and
only calls the model for something it has not seen before.
How It Works
Recording turns an experiment that depends on a hosted model into one that can be repeated exactly. Record a run once against a real provider, commit the resulting file, and every later run — in a test, on a build server, on a colleague’s machine — replays the identical answers with no network access, no key, and no cost. A benchmark whose model calls are recorded is reproducible; one whose calls are live is not.
Fields
| Field | Summary |
|---|---|
ChatClientRecordingMode None |
Call the model directly and save nothing. |
ChatClientRecordingMode Record |
Call the model and save every request and response into the configured store. |
ChatClientRecordingMode Replay |
Serve every request from the store and fail when no recording matches. |
ChatClientRecordingMode ReplayWithFallback |
Serve requests from the store, calling the model and recording the answer on a miss. |
int value__ |