Skip to content
AI usage dashboard

synthetic-contract expectations (manual calculation, all synthetic)

This directory contains synthetic test data with syn- IDs. It was created when local detection reported not_found; it contains no native samples. Layout: <root>/tmp/<project_id>/chats/<sessionId>.jsonl (ChatRecord JSONL).

Contents (eight lines)

  1. user syn-u-1: first-line detection fields, with one of four types and complete uuid/sessionId/timestamp.
  2. assistant syn-a-1, primary: usageMetadata {prompt 1000, candidates 50, cached 400, thoughts 10, toolUsePrompt 5, total 1050}.
  3. assistant syn-a-2, isSidechain=true → sub_agent: {prompt 2000, candidates 100, cached 0, total 2100}.
  4. assistant syn-a-3, agentId=“agent-x” → sub_agent: {prompt 500, candidates 25, total 525}.
  5. tool_result without usageMetadata: valid shape, no event.
  6. system/session_model: daemon recovery binding, no event.
  7. system/goal_state with cross-turn cumulative goal.tokensUsed=4,000,000,000: explicitly ignored, because adding it would duplicate usage.
  8. system/chat_compression: compression checkpoint control record, explicitly ignored.

Expectations

  • Three events: one primary and two sub_agent; records_seen=8, files[0].status=“complete”.
  • Keys qwen:syn-a-1, qwen:syn-a-2, qwen:syn-a-3; schema_version=“0.5.0” passed from each record.version; session_id=“syn-sess-1”; provider_id=NULL because ChatRecord has no provider.
  • input_total=promptTokenCount, total_tokens=reported totalTokenCount, cachedContentTokenCount→input_cache_read; thoughts/toolUsePrompt are not added to any field.
  • Summary (2026-01-05): call_count=3, input_total_known=1000+2000+500=3500, output_total_known=50+100+25=175, cache_read_known=400+0=400 (a3 omits it; no filled zero), total_tokens_known=1050+2100+525=3675, cache_write_known=None.
  • goal_state’s 4,000,000,000 appears in no summary field.