Adds a canonical machine-readable probability signal endpoint that distils
a completed simulation report into a structured prediction thesis.
New: backend/app/services/signal_extractor.py
- SignalExtractor.extract() calls chat_json() (3 attempts, temp 0.1, step 0.05)
against the report markdown and returns a validated MiroSignal dataclass
- Validates and normalises all fields: p_yes clamped to [0.01, 0.99],
confidence/action enums enforced, action recomputed from p_yes when invalid
- _trim_report() keeps the tail (conclusions) for long reports to stay within
token limits
- _salvage() fallback_parser recovers a minimal signal from partial LLM output
using regex probability extraction
New endpoint: POST /api/report/<report_id>/signal
- 404 if report not found
- 400 if report not yet completed or content is empty
- 422 if LLM fails after all retry attempts
- Returns canonical signal with thesis.{p_yes, confidence, action, regime,
summary, drivers, invalidators} — schema_version 1.1
New: backend/tests/services/test_signal_extractor.py
- 27 tests covering happy path, field validation/normalisation, report
trimming, _salvage fallback, and LLM failure propagation
- No real API calls — LLMClient fully mocked
Closes#277
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
30 tests covering _clean_response_text (think tags, markdown fences),
_fix_truncated_json (unclosed braces/brackets/strings), _try_fix_json
(near-valid JSON recovery), and chat_json retry behaviour (max_attempts,
temperature backoff, fallback_parser, ValueError after all attempts fail,
finish_reason==length truncation repair, API exception handling).
All tests use mocked OpenAI client — no real API calls.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Background:
- With many nodes, D3 force simulation tick callback updates all nodes/edges/labels every frame
- Original implementation called simulation.restart() in drag start event on mousedown
- Even after simulation converged, clicking a node to view details would restart simulation, causing lag
Solution:
- Distinguish between "click" and "drag" using 3px movement threshold
- On drag start: only record initial position, do not restart simulation
- On drag event: detect movement exceeding threshold before marking as actual drag and restarting simulation
- Pure click operations no longer trigger simulation restart, keeping graph static
Bug fix:
- Fixed issue where nodes became undraggable after initial optimization
- Cause: incorrectly used if(!event.active) check in drag event
- event.active equals 1 during drag event, causing restart() to never execute
- Removed that condition, using custom _isDragging flag to control logic instead
- Changed the title for the environment configuration status icon to "环境搭建" for clarity.
- Added a creation time display in the project detail modal for better context.
- Introduced a divider for the playback section to improve visual separation.
- Enhanced navigation buttons with step indicators for clearer user progression.
- Added a hint for playback limitations to inform users about simulation requirements.
- Updated CSS styles for new elements and improved overall modal aesthetics.
- Changed Python version specification to indicate compatibility with versions ≥3.11 and ≤3.12.
- Renamed "Contact Us" section to "Join the Conversation" for improved engagement.
- Added a centered image of the QQ group to the contact section for better visibility.
- Updated card header to include status icons indicating project availability for different functionalities.
- Added a visual indicator for additional files when more than three are present.
- Improved card footer layout by grouping date and time, and enhancing progress display with status colors.
- Introduced new CSS styles for status icons and improved overall card aesthetics.
LoryGlory
666ghj