Documentation
¶
Overview ¶
Package mock ships a credential-free scripted LLM provider for offline testing of agent flows.
The scripted provider replays a JSONL transcript turn-by-turn. Pair it with a recording captured against a real provider to exercise the agent loop without burning API quota. Construct one with NewScripted; there is no registry and no init-time registration — callers wire the returned model.LLM explicitly.
Tool execution at replay time uses the live environment, so the scripted provider faithfully replays the LLM side but not the wider tool surface — fine for testing prompt construction and loop shape, not for bit-exact session reproduction.
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
Types ¶
type RecordedTurn ¶
type RecordedTurn struct {
Request *adkmodel.LLMRequest `json:"request"`
Responses []*adkmodel.LLMResponse `json:"responses"`
}
RecordedTurn is the on-disk shape of a single LLM turn consumed by the scripted provider.
One RecordedTurn is written per JSONL line. Request is a snapshot taken before the inner LLM may have mutated it (Config.Tools is commonly appended to). Responses is the full ordered stream of LLMResponse values yielded for that turn — typically zero or more Partial: true chunks followed by exactly one TurnComplete: true.
Note that adkmodel.LLMRequest.Tools is tagged json:"-" upstream and will silently drop on serialization. That's intentional: the inner LLM provides tool declarations on replay; recorded Tools would be dead weight.
This wire format is shared with core-agent's pkg/recording, whose recording wrapper (recording.NewRecorder) produces transcripts in exactly this shape. If/when the recorder ports to mast, this type and the JSONL read path should be extracted to a mast pkg/recording that both packages consume.
Source Files