23 lines
2.6 KiB
Markdown
23 lines
2.6 KiB
Markdown
# Investigation v10: restore full research history
|
|
|
|
Current successor: [v11 stable case cache](stable-case-cache-v1.md) preserves
|
|
this full-history policy and adds deterministic aliases. Saved v10 requests
|
|
remain unchanged; newly created project contexts now use v11.
|
|
|
|
At the user's request, new project-backed investigation contexts use `investigation_project_v10` and restore the full stored research history to every model request. This rolls back the v8/v9 evidence window and v9 active-note truncation. It does not introduce provider sessions or caches, alter research budgets, restart cases or run a paid test.
|
|
|
|
- Every evidence ID and payload is included, including old, negative/empty, qualified, other-hypothesis and already assessed results.
|
|
- Every hypothesis's notes and questions remain available, with original resolved states. The selected proposal stays complete. Other proposals retain the previously shortened title/mechanism presentation; their notes/questions are no longer omitted.
|
|
- No evidence-index placeholder substitutes for actual contents. No new prompt offers the `recall` action; it is unnecessary with full evidence present.
|
|
- The v7 lossless source/DB tables, shared provenance, source-file consolidation and contiguous code blocks remain.
|
|
- Interpretation, hypothesis-generation, diagnosis and handoff selection/compression policies remain unchanged. The embedded interpretation-schema fix remains.
|
|
- Existing v8/v9 contexts, output schemas, prompts and recorded recall decisions are retained only for deterministic replay/resume. A frozen in-flight request keeps its original contract; subsequent newly created contexts use v10. Unused recall selections are not passed to v10.
|
|
|
|
The change affects model views only; canonical contexts and package contents remain append-only. `docs/architecture.md` is unchanged. Controller confidence updates, scheduling and provider APIs are unchanged.
|
|
|
|
## Verification
|
|
|
|
Tests cover complete content/history restoration, cross-hypothesis and empty findings, unchanged output schema relative to v7, absence of recall in v10, scoped FILE aliases, legacy recall recovery, and the normal investigation/early-diagnosis lifecycle. All 16 stored investigation contexts from the interrupted selected-context Bug 17 trial still reconstruct their original v9 requests exactly. Their projected v10 views restore the complete v7 working evidence/history, with only the retained short alternative proposals and removal of obsolete recall bookkeeping.
|
|
|
|
The larger input size is intentional: the user requested restoration of the working history before further design changes. No paid LLM test was run.
|