Restore cogtest prompts to match cogbot reference (re-add type field)

Default sample size to 10

Route cogtest/expert synthesis to GPT-OSS 20B when main model is Llama 3.1 8B Instant

Route cogtest/expert synthesis to GPT-OSS 20B when main model is Llama 3.1 8B Instant

Update Anthropic model to Sonnet 4.6 and add reproducible-sample option

Update Anthropic model to Sonnet 4.6 and add reproducible-sample option

Increase run limit to 15 for Haonan Sun

Increase run limit to 15 for Haonan Sun

Add student tokens CSV for assignment mode

Update cogtest prompts: checklist respondent, label-free analyst, merge-by-issue synthesis

Update token usage: load from CSV, use HF_USAGE_REPO/TOKEN secrets, limit=8, show remaining runs

Add huggingface_hub dependency for assignment usage persistence

Add assignment token mode for MY456 students (per-token run limits, persistent usage tracking)

Clean up model dropdown labels: remove (Free) and (On-Demand) tags

Remove GPT-OSS 20B model

Add Groq Llama 3.3 70B (On-Demand) model

Update models: Claude, OpenAI, Groq (Llama 4 Scout, GPT-OSS 20B, Llama 3.1 8B). Remove Winston/vLLM/local.

Hide individual responses in Response Generation tab (keep for Question Testing only)

QAS: summary ordered by rater consensus

QAS: 5 independent raters with aggregated vote counts

Fix QAS parser: item_id casing mismatch

Add QAS-99 (Question Appraisal System) evaluation

Update Expert Review info text

Add Expert Review pipeline (3 experts + synthesis)

Add Question Appraisal System placeholder option

Replace CogBot prompt style with Cognitive Interview default + Question testing expert placeholder

Fix Groq cogtest failure: sequential execution + error visibility

Fix over-diagnosis: reframe analyst taxonomy as reference categories, not checklist

Add impact score (severity × prevalence), sort problems by impact

Replace severity labels with B&C-aligned severity scores and prevalence

Fix confidence rating extraction to handle varied LLM output formats

Fix cogtest pipeline: remove traffic lights, align synthesis with coding, add confidence stats

Add Cognitive Interview pipeline and Claude/OpenAI model support

Upload dashboard.py with huggingface_hub

Upload ess_uk_with_backstories_short_half.csv with huggingface_hub

Upload ess_uk_with_backstories_long_half.csv with huggingface_hub

Upload dashboard_backend.py with huggingface_hub

Upload dashboard.py with huggingface_hub

Upload dashboard_backend.py with huggingface_hub

Restore individual respondent expanders for Question Testing mode only

Remove individual respondent view, keep distribution charts for non-text formats

Add Custom response format option

Restore correct dashboard_backend.py

Upload dashboard_backend.py with huggingface_hub

Upload dashboard.py with huggingface_hub

Update ess_uk_with_backstories_short.csv: restore demographics + add weight/design variables

Update ess_uk_with_backstories_long.csv: restore demographics + add weight/design variables

Update ess_uk_with_backstories.csv: restore demographics + add weight/design variables

Remove unused _v2.csv

Remove unused _v1.csv