Repo README: document 3-tier catalog + mirror policy (ADR-245)

README.md

@@ -1,133 +1,53 @@
 ---
 license: apache-2.0
-base_model: Qwen/Qwen3-0.6B-Base
 tags:
   - text-generation
+  - gguf
+  - smollm2
   - qwen3
   - hinglish
-  - roman-hindi
-  - gguf
-  - qlora
-datasets:
-  - Abhishekcr448/Hinglish-Everyday-Conversations-1M
-  - Roshan32/Hinglish_Dataset_instruction_and_raw
-  - BhabhaAI/Cross-Hindi-Hinglish-chat
-pipeline_tag: text-generation
+  - mirror
 ---
 
-# WriteAmp-Qwen3-0.6B-Hinglish-v3
-
-QLoRA fine-tune of [`Qwen/Qwen3-0.6B-Base`](https://huggingface.co/Qwen/Qwen3-0.6B-Base) (Apache 2.0) for
-**Hinglish** (Roman/Latin-script Hindi) short-continuation. Shipped as a single Q4_K_M GGUF and used by
-[WriteAmp](https://writeamp.site) as the `max` / multilingual tier of its on-device macOS tab-autocomplete
-engine. Replaces the upstream base Qwen3-0.6B for all `max` users and all non-English routing in the
-catalog.
-
-> Base model card, license, and citation: see
-> [`Qwen/Qwen3-0.6B-Base`](https://huggingface.co/Qwen/Qwen3-0.6B-Base) (Apache 2.0, © Alibaba Cloud).
-
-## Intended use
-
-Short, cursor-local continuations of Roman-script Hinglish writing — the kind that flows through text
-fields in mail, chat, social posts, and notes. Designed to run **on-device** on Apple Silicon via the
-[llama.cpp](https://github.com/ggml-org/llama.cpp) runtime vendored by WriteAmp (`b9402`,
-`19e92c33ef974661e4b1e43dd48be231d07be5ed`). **Not** an instruct or chat model — pure base-model
-continuation.
-
-The WriteAmp app wraps the model with a Hindi-Roman directive, repetition guards, no-repeat-n-gram
-size 2, short output budget, and a Devanagari→Latin safety filter. Bare-bones `llama.cpp` use without
-those guards is **not** a supported configuration.
-
-## What's in the box
-
-A single immutable file:
-
-| File | Size | SHA-256 |
-| --- | --- | --- |
-| `WriteAmp-Qwen3-0.6B-Hinglish-v3.Q4_K_M.gguf` | 396,704,544 bytes | `98df7dd1daaf65602fa002733d5c5a32527acd174730b591235f0ed170f17185` |
-
-Architecture: `qwen3`, 28 layers, 1024 hidden dim, 151,936 token vocabulary, BOS/EOS token ID `151643`.
-Q4_K_M quantization. Built against the same llama.cpp revision as the WriteAmp vendored runtime,
-and loads cleanly through `LlamaModelRuntime` with reported `vocabularySize == 151936`.
-
-The on-disk ACPF token profile (`qwen3-v151936.acpf.bin`) shipped by the WriteAmp app opens cleanly
-against this artifact; no profile rebuild is required.
+# WriteAmp's on-device model catalog
 
-## Reproducibility
+This repository hosts WriteAmp's downloadable model artifacts. The catalog has three tiers
+(`mini`, `midi`, `max`); two of them are mirrors of upstream Q4_K_M quantizations, one is a
+custom Hinglish fine-tune.
 
-Exact toolchain and command lines used to produce this artifact live in
-[`tools/hinglish-sft/build-release-q4.sh`](../../tools/hinglish-sft/build-release-q4.sh) and the
-[WriteAmp decision log](../../docs/05-decisions.md) (ADR-243, ADR-244).
+| Tier | File | Size | SHA-256 | Source |
+| --- | --- | --- | --- | --- |
+| `mini` | [`mini/SmolLM2-135M.Q4_K_M.gguf`](resolve/main/mini/SmolLM2-135M.Q4_K_M.gguf?download=true) | 105,454,016 B | `127c52bae983b2d0de09c60d6292743e3239e44239bbe8ccb8cf3241513ff5a0` | Mirror of [`mradermacher/SmolLM2-135M-GGUF`](https://huggingface.co/mradermacher/SmolLM2-135M-GGUF) (Apache 2.0) |
+| `midi` | [`midi/SmolLM2-360M.Q4_K_M.gguf`](resolve/main/midi/SmolLM2-360M.Q4_K_M.gguf?download=true) | 270,590,464 B | `4ac7aa712c43fe7d07ed910298b5eaab932292692954abd8e0703307724e4e04` | Mirror of [`mradermacher/SmolLM2-360M-GGUF`](https://huggingface.co/mradermacher/SmolLM2-360M-GGUF) (Apache 2.0) |
+| `max` | [`WriteAmp-Qwen3-0.6B-Hinglish-v3.Q4_K_M.gguf`](resolve/main/WriteAmp-Qwen3-0.6B-Hinglish-v3.Q4_K_M.gguf?download=true) | 396,704,544 B | `98df7dd1daaf65602fa002733d5c5a32527acd174730b591235f0ed170f17185` | Custom QLoRA fine-tune of [`Qwen/Qwen3-0.6B-Base`](https://huggingface.co/Qwen/Qwen3-0.6B-Base) (Apache 2.0) |
 
-Summary of the inputs:
-
-| Input | Source |
-| --- | --- |
-| Base model | [`Qwen/Qwen3-0.6B-Base`](https://huggingface.co/Qwen/Qwen3-0.6B-Base) (Apache 2.0) |
-| LoRA adapter | `tools/hinglish-sft/adapter/big-v2/0003000_adapters.safetensors` (QLoRA, rank 8, q/v only) |
-| Training data | 838K examples built by `build_dataset.py` from the three HF datasets below |
-| llama.cpp revision | `b9402` (`19e92c33ef974661e4b1e43dd48be231d07be5ed`) |
-| Converter | `llama.cpp/convert_hf_to_gguf.py` (official) |
-| Quantizer | `llama-quantize` from the same checkout |
-
-## Training data
-
-Built from three public datasets, mixed in `tools/hinglish-sft/build_dataset.py`. Only the QLoRA
-adapter and the merged + converted GGUF are redistributed — the raw training data is **not** in this
-repository.
-
-- [`Abhishekcr448/Hinglish-Everyday-Conversations-1M`](https://huggingface.co/datasets/Abhishekcr448/Hinglish-Everyday-Conversations-1M) — 1M conversational Hinglish exchanges (primary).
-- [`Roshan32/Hinglish_Dataset_instruction_and_raw`](https://huggingface.co/datasets/Roshan32/Hinglish_Dataset_instruction_and_raw) — 14K persona chat (style mix).
-- [`BhabhaAI/Cross-Hindi-Hinglish-chat`](https://huggingface.co/datasets/BhabhaAI/Cross-Hindi-Hinglish-chat) — 1.7K Hinglish-only (script-filter).
-
-Each source dataset's own Hugging Face page lists its license; we treat the released fine-tune as a
-derivative of the Apache 2.0 base model and the aggregated training-data signals, and we distribute
-under Apache 2.0 consistent with the base. Reviewers should consult the three upstream dataset
-license fields before any downstream redistribution.
-
-## Known limitations
-
-- **0.6B base-model ceiling.** Quality still trails larger base models. Short outputs (2–3 words)
-  are the sweet spot; longer outputs are often truncated or repeated.
-- **Hindi-script coverage is partial.** Fine-tuned for Roman/Latin Hinglish. Devanagari Hindi
-  inputs are passed through to the model; WriteAmp's safety filter suppresses Latin-script leakage
-  in that direction. True Devanagari generation quality is not benchmarked here.
-- **Multilingual breadth is narrow.** The base Qwen3 0.6B supports EN/ES/FR/DE/PT/JA/KO/AR with
-  varying quality; this fine-tune preserves that surface but does not improve it.
-- **No safety filter baked in.** The model is a continuation model; WriteAmp's app wraps it with a
-  directive + repetition guard + script filter. Direct generation use without those guards may
-  produce prompt echoes or low-quality output on edge prompts.
-
-## Quality evidence
-
-Production-decoder bakeoff (`rep1.15, ngram2, maxTokens=6, branchWidth=3, directive=hinglish`,
-through the vendored `LlamaModelRuntime` and the existing `qwen3-v151936.acpf.bin` profile):
-
-```
-SUMMARY model=hinglish-v3-q4-existing-profile n=8 p50=128ms p90=130ms empty=0 loop=0 echo=0 dev=0 ok=8
-```
+Each tier's mirror card (under `mini/` and `midi/`) carries provenance and license details. The
+`max` tier's full description, training data, and known limitations are in
+[`WriteAmp-Qwen3-0.6B-Hinglish-v3.Q4_K_M.gguf`](resolve/main/WriteAmp-Qwen3-0.6B-Hinglish-v3.Q4_K_M.gguf?download=true)'s
+card above.
 
-8/8 prompts produced usable Hinglish, zero Devanagari/CJK leakage, zero prompt echo, zero repetition
-loops. Cold-load p50 ≈ 128 ms, p90 ≈ 130 ms at 6 tokens — inside the existing `max` envelope.
+## Mirroring policy
 
-## License & attribution
+`mini` and `midi` are byte-identical mirrors of `mradermacher`'s community Q4_K_M quantizations.
+The WriteAmp app's catalog keeps `mradermacher`'s URLs as the primary download path (the
+community attribution is preserved), and uses this repo as a cold-start fallback when the primary
+becomes unreachable or returns HTTP 404. See ADR-245 in the WriteAmp decision log for the full
+rationale.
 
-This artifact is released under the **Apache License 2.0**, consistent with the base model
-[`Qwen/Qwen3-0.6B-Base`](https://huggingface.co/Qwen/Qwen3-0.6B-Base) (© Alibaba Cloud).
+The `max` tier is hosted here as the primary (no third-party mirror). It is also the source of
+truth for the `max` catalog slot.
 
-Modifications from the base:
-- QLoRA fine-tune on aggregated Hinglish conversational data (see "Training data" above).
-- Merged + converted to GGUF F16 via `llama.cpp/convert_hf_to_gguf.py` (revision `b9402`).
-- Quantized to Q4_K_M via `llama-quantize` from the same revision.
+## License
 
-The full Apache 2.0 text is reproduced at the head of the base model's `LICENSE` file on Hugging Face
-and applies to this derivative work.
+All artifacts under this repo are distributed under the **Apache License 2.0**, consistent with
+their upstream base models (`HuggingFaceTB/SmolLM2-135M`, `HuggingFaceTB/SmolLM2-360M`,
+`Qwen/Qwen3-0.6B-Base`) and `mradermacher`'s Q4_K_M quantizations of those bases.
 
 ## Citation
 
 ```
-@misc{writeamp_hinglish_v3_2026,
-  title  = {WriteAmp-Qwen3-0.6B-Hinglish-v3: a QLoRA fine-tune of Qwen3-0.6B-Base for on-device Hinglish continuation},
+@misc{writeamp_models_2026,
+  title  = {WriteAmp on-device model catalog: SmolLM2 mirrors + Qwen3-0.6B Hinglish fine-tune},
   author = {DataDab LLP},
   year   = {2026},
   url    = {https://huggingface.co/datadab/writeamp-models}