yf176 commited on
Commit ·
7e4b3af
1
Parent(s): d647d80
Record V7.1 readiness and committee halt
Browse files- docs/V7.1_PAPER_READINESS.md +49 -0
- docs/polyhedraV7.1.md +37 -35
docs/V7.1_PAPER_READINESS.md
ADDED
|
@@ -0,0 +1,49 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# V7.1 Paper-Readiness Status
|
| 2 |
+
|
| 3 |
+
Paper trading simulation only. No real capital. Not financial advice.
|
| 4 |
+
|
| 5 |
+
This document separates implemented behavior from verification evidence and known limitations.
|
| 6 |
+
|
| 7 |
+
## Implemented Code
|
| 8 |
+
|
| 9 |
+
- YOLO UI session list and current-run view.
|
| 10 |
+
- UI-only prior-run replay APIs and replay panel.
|
| 11 |
+
- Recorded YOLO run snapshots under `yolo_runs/`.
|
| 12 |
+
- Evidence ledger artifact writer for YOLO run packages.
|
| 13 |
+
- Run package export includes evidence ledger artifacts when entries exist.
|
| 14 |
+
- YOLO diagnostics for agent blocking, LLM availability failures, provider partial data, and options data gaps.
|
| 15 |
+
- Proposal availability classification:
|
| 16 |
+
- `paper_trade_suggestion`
|
| 17 |
+
- `data_loss`
|
| 18 |
+
- `agent_pending`
|
| 19 |
+
- `true_no_trade_or_watch`
|
| 20 |
+
- Runtime-owned QuantStrategists weights override LLM-emitted `current_weight`.
|
| 21 |
+
- Unknown LLM-emitted committee members are retained for audit but excluded from weighted aggregation.
|
| 22 |
+
|
| 23 |
+
## Local Automated Tests
|
| 24 |
+
|
| 25 |
+
- Focused syntax checks passed with the bundled Codex Python runtime.
|
| 26 |
+
- Targeted V7.1 regression suite passed: `PYTHONPATH=src python -m pytest tests/test_v611_web_ui.py tests/test_yolo_context.py` -> 43 passed.
|
| 27 |
+
- Full local `pytest` was attempted and is not green. Collection is blocked by pre-existing legacy research-agent/test syntax errors and missing legacy dependencies (`mcp`, `sentence_transformers`, `bibtexparser`) in the local Codex runtime.
|
| 28 |
+
|
| 29 |
+
## Live UI Verification
|
| 30 |
+
|
| 31 |
+
- HF UI health and prior deployed YOLO diagnostics were verified during V7.1 work.
|
| 32 |
+
- A fresh live happy-path YOLO completion is not recorded because Gemini returned a rate-limit error during live smoke testing.
|
| 33 |
+
|
| 34 |
+
## Live Slack Verification
|
| 35 |
+
|
| 36 |
+
- Live Slack YOLO baseline/session-delta evidence for V7.1 replay is not yet recorded.
|
| 37 |
+
|
| 38 |
+
## Halted Scope
|
| 39 |
+
|
| 40 |
+
- Distilled analyst committee implementation is halted for V7.1.
|
| 41 |
+
- Reason: adding distilled analyst members now would influence the experiment and make later empirical evaluation harder to interpret.
|
| 42 |
+
- Current generic QuantStrategists members remain V7.0 scaffolding only.
|
| 43 |
+
|
| 44 |
+
## Known Limitations
|
| 45 |
+
|
| 46 |
+
- V7.1 replay is UI replay of recorded run snapshots, not frozen deterministic replay.
|
| 47 |
+
- Historical experiment mode is not implemented in V7.1.
|
| 48 |
+
- No real trades are placed.
|
| 49 |
+
- No new providers, API keys, brokers, schedulers, or workflow agents were added for V7.1.
|
docs/polyhedraV7.1.md
CHANGED
|
@@ -187,6 +187,8 @@ These future additions must remain paper-trading unless a separate real-capital
|
|
| 187 |
|
| 188 |
### FR-3 - Distilled Analyst Committee Direction
|
| 189 |
|
|
|
|
|
|
|
| 190 |
**FR-3.1** The current generic QuantStrategists committee members are V7.0 scaffolding, not the final V7.1+ analyst group.
|
| 191 |
|
| 192 |
**FR-3.2** Future committee members must be created from approved distilled markdown sources.
|
|
@@ -354,46 +356,46 @@ The items above may be promoted into a later empirical-evaluation or product PRD
|
|
| 354 |
|
| 355 |
## Acceptance Criteria
|
| 356 |
|
| 357 |
-
- [
|
| 358 |
-
- [
|
| 359 |
-
- [
|
| 360 |
-
- [
|
| 361 |
-
- [
|
| 362 |
-
- [
|
| 363 |
-
- [
|
| 364 |
-
- [
|
| 365 |
-
- [
|
| 366 |
-
- [
|
| 367 |
-
- [ ] The first distilled markdown member can be added independently before joining aggregation.
|
| 368 |
-
- [
|
| 369 |
-
- [
|
| 370 |
-
- [
|
| 371 |
-
- [
|
| 372 |
-
- [
|
| 373 |
-
- [
|
| 374 |
|
| 375 |
---
|
| 376 |
|
| 377 |
## Implementation Checklist
|
| 378 |
|
| 379 |
-
- [
|
| 380 |
-
- [
|
| 381 |
-
- [
|
| 382 |
-
- [
|
| 383 |
-
- [
|
| 384 |
-
- [
|
| 385 |
-
- [
|
| 386 |
-
- [
|
| 387 |
-
- [
|
| 388 |
-
- [
|
| 389 |
-
- [
|
| 390 |
-
- [
|
| 391 |
-
- [ ] Add committee member registry schema for future distilled markdown members.
|
| 392 |
-
- [
|
| 393 |
-
- [
|
| 394 |
-
- [
|
| 395 |
-
- [
|
| 396 |
-
- [
|
| 397 |
- [ ] Record live Slack YOLO baseline and session-delta smoke-test evidence.
|
| 398 |
- [ ] Record live UI YOLO baseline, session-delta, replay, and download-package smoke-test evidence.
|
| 399 |
- [ ] Preserve extension points for a later empirical-evaluation track without presenting them as completed V7.1 behavior.
|
|
|
|
| 187 |
|
| 188 |
### FR-3 - Distilled Analyst Committee Direction
|
| 189 |
|
| 190 |
+
**V7.1 halt status:** implementation of distilled analyst committee members is halted for V7.1. Reason: adding or changing distilled analyst members now would influence the experiment and make later empirical evaluation harder to interpret. The generic V7.0 QuantStrategists committee remains scaffolding only.
|
| 191 |
+
|
| 192 |
**FR-3.1** The current generic QuantStrategists committee members are V7.0 scaffolding, not the final V7.1+ analyst group.
|
| 193 |
|
| 194 |
**FR-3.2** Future committee members must be created from approved distilled markdown sources.
|
|
|
|
| 356 |
|
| 357 |
## Acceptance Criteria
|
| 358 |
|
| 359 |
+
- [x] UI can select a prior YOLO run in an existing session.
|
| 360 |
+
- [x] UI replay displays the selected run's recorded workflow progress.
|
| 361 |
+
- [x] UI replay displays the selected run's Finley output.
|
| 362 |
+
- [x] UI replay labels baseline vs session-delta runs.
|
| 363 |
+
- [x] UI replay does not rerun agents.
|
| 364 |
+
- [x] UI replay does not claim frozen/deterministic replay.
|
| 365 |
+
- [x] Download package includes evidence ledger when evidence entries exist.
|
| 366 |
+
- [x] Evidence ledger records used evidence, excluded evidence, and data gaps.
|
| 367 |
+
- [x] Evidence ledger does not invent missing evidence.
|
| 368 |
+
- [x] QuantStrategists member registry requirement is documented for distilled markdown members.
|
| 369 |
+
- [ ] The first distilled markdown member can be added independently before joining aggregation. Halted for V7.1 because it will influence the experiment.
|
| 370 |
+
- [x] LLM committee votes cannot override runtime-owned member weights.
|
| 371 |
+
- [x] Unknown LLM-emitted committee members cannot silently join weighted aggregation.
|
| 372 |
+
- [x] Paper-readiness checklist separates implemented behavior, local tests, live Slack checks, live UI checks, and limitations.
|
| 373 |
+
- [x] arXiv-facing claims distinguish current architecture claims from future or separately verified empirical claims.
|
| 374 |
+
- [x] No new env vars, providers, brokers, schedulers, or workflow agents are added without explicit approval.
|
| 375 |
+
- [x] Paper-trading disclaimer remains visible in UI output.
|
| 376 |
|
| 377 |
---
|
| 378 |
|
| 379 |
## Implementation Checklist
|
| 380 |
|
| 381 |
+
- [x] Add UI replay selector for runs inside the selected session.
|
| 382 |
+
- [x] Add backend run-list API using existing session/run records.
|
| 383 |
+
- [x] Add backend run-detail API using recorded artifacts and state.
|
| 384 |
+
- [x] Render selected prior run progress without executing agents.
|
| 385 |
+
- [x] Render selected prior run Finley output without executing agents.
|
| 386 |
+
- [x] Preserve current live-run view alongside replay view.
|
| 387 |
+
- [x] Add evidence ledger artifact writer for YOLO runs.
|
| 388 |
+
- [x] Include evidence ledger artifacts in run package export.
|
| 389 |
+
- [x] Add tests for UI-only replay not triggering agent execution.
|
| 390 |
+
- [x] Add tests for replay baseline/delta labels.
|
| 391 |
+
- [x] Add tests for evidence ledger package inclusion.
|
| 392 |
+
- [x] Add tests for evidence ledger data-gap entries.
|
| 393 |
+
- [ ] Add committee member registry schema for future distilled markdown members. Halted for V7.1 because it will influence the experiment.
|
| 394 |
+
- [x] Refactor committee aggregation so runtime weights override any LLM-emitted weight.
|
| 395 |
+
- [x] Add tests that LLM output cannot override member weights.
|
| 396 |
+
- [x] Add tests that unknown committee members cannot silently create weighted aggregation members.
|
| 397 |
+
- [x] Add paper-readiness checklist or status artifact for arXiv drafting.
|
| 398 |
+
- [x] Record local automated test command and result after repairing the Python/runtime environment.
|
| 399 |
- [ ] Record live Slack YOLO baseline and session-delta smoke-test evidence.
|
| 400 |
- [ ] Record live UI YOLO baseline, session-delta, replay, and download-package smoke-test evidence.
|
| 401 |
- [ ] Preserve extension points for a later empirical-evaluation track without presenting them as completed V7.1 behavior.
|