Skip to main content
Load an existing transcript and re-synthesize it — optionally with a different synthesizer or additional reflection rounds. Produces a new transcript; the original is never modified.

​
Usage

dissent replay TRANSCRIPT_ID [OPTIONS]

​
Arguments

ArgumentDescription
TRANSCRIPT_IDFull UUID or prefix (minimum 4 characters)

​
Options

OptionDefaultDescription
--synthesizeroriginalModel alias override for synthesis
--rounds0Additional reflection rounds to run before synthesis
--verboseoffShow all rounds, not just synthesis
--no-saveoffSkip saving replay transcript to disk
--outputterminalOutput format: terminal, json, markdown
--file PATH—Write output to file instead of stdout
--ground-truth TEXT—Reference answer to score synthesis against
--ground-truth-file PATH—File containing reference answer

​
Examples

# Re-synthesize with a different model
dissent replay abcd1234 --synthesizer gpt

# Add two more reflection rounds then synthesize
dissent replay abcd1234 --rounds 2

# Re-synthesize and score against a reference answer
dissent replay abcd1234 --synthesizer claude --ground-truth "The correct answer is X"

# Write replay output to markdown
dissent replay abcd1234 --output markdown --file replay.md

​
Modes

Re-synthesize only (--rounds 0, the default): copies the source transcript’s rounds unchanged and runs a fresh synthesis pass. Use this to compare how different synthesizers interpret the same debate. Add rounds (--rounds N): copies the source rounds, runs N additional reflection rounds continuing from the last round, then synthesizes. Each additional round costs one API call per panel model.

​
Notes

  • The new transcript links to the source via metadata.source_transcript_id
  • Partial IDs must be at least 4 characters; use dissent list to find IDs
  • --rounds counts additional rounds, not total rounds