tbxy09/dsh-replay-lab

tbxy09★ 0TypeScriptLast synced: 2026-08-17

Open on GitHub

Controlled request-surface replay and regression workbench for DeepSeek Harness

README excerpt

English 简体中文 DSH Replay Lab (ReplayLab) Replay the request surface, not just the prompt. dsh-replay-lab is a DeepSeek Harness plugin for replaying completed agent turns against different presets or plugins and comparing their request surfaces, trajectories, costs, errors, and outcomes. Use it to reproduce and debug long agent trajectories, repeated tool-call loops, no-progress turns, and preset- or plugin-dependent regressions. Freeze a completed DeepSeek Harness turn, approve one isolated candidate, and compare outcome, trajectory, errors, cost, and the request surface that produced them. Install · Verify · Security · DeepSeek Harness Why Replay Lab exists DeepSeek Harness made the agent runtime programmable, but changing a preset or plugin also changes what the model sees. Replay Lab makes that request surface observable and testable instead of treating every behavior change as a prompt change. The capability lineage has five connected stages: 1. Request surface became a variable. Long tool/context histories produced observable failures such as plain-text tool calls and empty responses after tool results. 2. Narrow and wide surfaces became comparable. DSH's Minimal and Standard p…

View full README on GitHub →
Agentsagent-observabilitydeepseekdeepseek-harnessdshdsh-pluginllm-evaluationreplayagent

Category