DNA-017: source-tag scans (ui vs internal-test)

DNA-017: per-session scan-log file (stop restart truncation)

KNOWN_BASES: add Cactus Needle

UNRECOGNIZED -> UNVERIFIED (UNKNOWN band), not CLEAN

UNRECOGNIZED -> UNVERIFIED (UNKNOWN band), not CLEAN

KNOWN_BASES: add Inkling + ERNIE-text-MoE/Zamba2/PLaMo families (36 total)

KNOWN_BASES: add ERNIE 4.5 text-MoE + Zamba2; extend ZAYA1 patterns

DNA-018: verdict-first result hierarchy + audit trail; reference_count()

DNA-018: verdict-first result hierarchy + audit trail; reference_count()

Fix NAME_MISMATCH false-positive on honestly-labeled models (case + family-label match)

DNA-017: normalize+validate model_id input (reject pasted blobs)

DNA-017: startup+per-scan logging diagnostics, flush every 1min

Deploy: mxbai GGUF false-positive fix + DNA-017 scan telemetry (scan.py)

Deploy: mxbai GGUF false-positive fix + DNA-017 scan telemetry (app.py)

Wave 2: verified-lab recognition + badge + lineage graph

Remove internal board IDs from public engine labels

Wave 1: provenance risk score (DNA-001) + engine panel (DNA-003)

Detect GGUF-only repos by file list (not just name)

Recognize catalogued original architectures (Laguna fix) + filter fix

Mythos/Fable closed-brand labels (Mythos unreleased, Fable released weights-closed)

GGUF support (metadata inspection + source resolution) + Mythos closed-brand

Stage 1: PROVENANCE_SCRUBBING + IMPOSSIBLE_WEIGHT_PROVENANCE + NAMING_MISATTRIBUTION detectors

Add hunyuan_v3 (hy_v3) to KNOWN_BASES

Add mellum to KNOWN_BASES (JetBrains Mellum2-12B-A2.5B)

Qwen3 vocab: add OpenSeek 151851 and SciCore 151670 variants

Use HF_TOKEN env var for gated model access

starcoder2: add gpt_bigcode pattern for tiny_starcoder_py (234K dl)

Add starcoder2: BigCode StarCoder2 family

Add ministral3: fixes false NemotronH match on Mistral-Medium-3.5-128B

Add zaya entry (Zyphra ZAYA1); update to 28 KNOWN_BASES total

Wire free CTAs to HF Space, contact to email

Update stats: 2450+ models, 27 base families, new CN labs in coverage grid

Fix config lift model_type priority: gpt_oss InternVL3.5 wrapper now HIGH

P9: hunyuan_v1, gpt_oss, valley + recursive 2-level config lifting

P2 batch: pangu_pro_moe, iquest_coder, minicpm, step3_5, mimo_v2, llada2-uni, emu3 + fixes

Add emu3; fix qianfan_vl/interns1 patterns (Nemo audit)

Add 8 KNOWN_BASES entries batch 1+2: GLM-4.x, Seed-OSS, Bailing, Kimi, ERNIE-4.5, Qianfan, LLaDA2, InternS1

Add 4 KNOWN_BASES entries: GLM-4.x, Seed-OSS, Bailing-V2/Ling, Kimi-Linear

Wire modelatlas_similar to HF parquet dataset (2,435 models)

Add NemotronH to KNOWN_BASES; handle llm_config nesting pattern

Add GGUF detection with friendly error; clarify scope in UI

Fix: upgrade to Gradio 5.50 (Python 3.13 compat), Ocean theme

Launch modeldna Space: Stage 1 architecture scanner

initial commit