Tianyi Cui
25 lines
1.0 KiB
Markdown
25 lines
1.0 KiB
Markdown
<!-- Generated by scripts/gen-doc-graphs.ts - do not edit by hand.
|
|
Run `pnpm run gen-doc-graphs` to regenerate. -->
|
|
|
|
# ACP Snapshot Replay
|
|
|
|
Maintenance mode: curated Mermaid sequence based on the snapshot test harness.
|
|
|
|
This graph explains what a snapshot scenario proves: recorded real-model session logs are replayed keylessly, then ACP stdout is normalized and diffed.
|
|
|
|
```mermaid
|
|
sequenceDiagram
|
|
participant Recorder as Real API recording
|
|
participant Fixture as snapshot fixture
|
|
participant Replay as llm-replay adapter
|
|
participant ACP as acp-agent subprocess
|
|
participant Golden as stdout golden
|
|
Recorder->>Fixture: session.jsonl + workspace inputs
|
|
Fixture->>Replay: recorded StreamChunk script
|
|
Replay->>ACP: deterministic llm/stream chunks
|
|
ACP->>Golden: normalized sessionUpdate stream
|
|
Golden-->>ACP: diff must be empty
|
|
```
|
|
|
|
Future pressure from the fs stack: policy rejection scenarios are valuable because they prove both world state and failed tool-card rendering, not just that replay returns text.
|