Replace Gemma 3 1B with Qwen2.5-1.5B q8; both models now Apache 2.0

NOTICE

@@ -1,29 +1,38 @@
-This repository redistributes Google's Gemma models for use by the EchoPeak (山應)
-offline wilderness assistant. The models are unmodified Google releases except as
-stated in README.md.
+This application downloads and runs language models produced by third parties.
+Every model it ships is distributed under the Apache License, Version 2.0, a copy
+of which is included as LICENSE-apache-2.0.txt.
 
 --------------------------------------------------------------------------------
-gemma-4-E2B-it.litertlm
+Gemma 4 E2B  (gemma-4-E2B-it.litertlm)
 --------------------------------------------------------------------------------
 
 Gemma 4 is provided by Google under the Apache License, Version 2.0.
-A copy of the license is included as LICENSE-apache-2.0.txt, and is also
-available at:
 
+    https://ai.google.dev/gemma
     https://ai.google.dev/gemma/apache_2
 
 --------------------------------------------------------------------------------
-gemma3-1b-it-int4.litertlm
+Qwen2.5 1.5B Instruct
+(Qwen2.5-1.5B-Instruct_multi-prefill-seq_q8_ekv4096.litertlm)
 --------------------------------------------------------------------------------
 
-Gemma is provided under and subject to the Gemma Terms of Use found at
-ai.google.dev/gemma/terms
+Qwen2.5-1.5B-Instruct is provided by Alibaba Cloud under the Apache License,
+Version 2.0. Converted to LiteRT-LM format by the litert-community project.
 
-A copy of those terms is included as LICENSE-gemma-terms.txt. Use of this model is
-additionally subject to the Gemma Prohibited Use Policy, included as
-PROHIBITED_USE_POLICY.txt and available at:
+    https://huggingface.co/Qwen/Qwen2.5-1.5B-Instruct
+    https://huggingface.co/litert-community/Qwen2.5-1.5B-Instruct
 
-    https://ai.google.dev/gemma/prohibited_use_policy
+--------------------------------------------------------------------------------
+Embedding model
+--------------------------------------------------------------------------------
+
+EmbeddingGemma 300M is provided by Google and powers on-device semantic search.
+See https://ai.google.dev/gemma for its terms.
+
+--------------------------------------------------------------------------------
 
-Gemma 3 is listed in the Appendix of the Gemma Terms of Use and is therefore
-governed by those terms rather than the Apache License.
+Earlier builds shipped Gemma 3 1B, which falls under the separate Gemma Terms of
+Use rather than Apache 2.0. That model was removed, and with it the obligation to
+distribute the Gemma Terms and the Prohibited Use Policy alongside the weights.
+Re-adding any model listed in the appendix of ai.google.dev/gemma/terms
+(Gemma 1, 1.1, 2, 3, 3n) would reinstate those obligations.

Qwen2.5-1.5B-Instruct_multi-prefill-seq_q8_ekv4096.litertlm

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:faa60663b333290c1496c499828b21d3e3254a788cacd8cce917ce0f761a2dc9
+size 1597931520

README.md

@@ -32,15 +32,15 @@ APK。因此本 repo 以 public 形式轉載，並依授權要求隨附完整條
 
 | 檔案 | 大小 | 授權 | App 內層級 |
 |---|---|---|---|
-| `gemma-4-E2B-it.litertlm` | 約 2.59 GB | **Apache 2.0** | STANDARD（12 GB 級手機） |
-| `Gemma3-1B-IT_multi-prefill-seq_q4_ekv4096.litertlm` | 約 0.5 GB | **Gemma Terms of Use** | LIGHT（6–8 GB 級手機） |
+| `gemma-4-E2B-it.litertlm` | 約 2.59 GB | Apache 2.0 | STANDARD（12 GB 級手機） |
+| `Qwen2.5-1.5B-Instruct_multi-prefill-seq_q8_ekv4096.litertlm` | 約 1.49 GB | Apache 2.0 | LIGHT（6–8 GB 級手機） |
 
-**兩顆模型授權不同，請勿混用條款：**
+**兩顆都是 Apache License 2.0** → [`LICENSE-apache-2.0.txt`](LICENSE-apache-2.0.txt)，
+散布義務僅需附授權全文與出處標示。
 
-- **Gemma 4** 採 Apache License 2.0 → [`LICENSE-apache-2.0.txt`](LICENSE-apache-2.0.txt)
-- **Gemma 3** 列於 Gemma Terms of Use 的 Appendix，採該自訂條款 →
-  [`LICENSE-gemma-terms.txt`](LICENSE-gemma-terms.txt)，並受
-  [`PROHIBITED_USE_POLICY.txt`](PROHIBITED_USE_POLICY.txt) 規範。
+先前的輕量層是 Gemma 3 1B，它適用的是自訂的 Gemma Terms of Use（義務較重，需額外
+傳遞 Agreement 與 Prohibited Use Policy）。該模型已移除，相關義務隨之消失；若日後
+重新加入 Gemma Terms 附錄所列的模型（Gemma 1／1.1／2／3／3n），義務會重新產生。
 
 完整聲明見 [`NOTICE`](NOTICE)。
 

SHA256SUMS.txt

@@ -1,2 +1,2 @@
 181938105e0eefd105961417e8da75903eacda102c4fce9ce90f50b97139a63c  gemma-4-E2B-it.litertlm
-1325ae366d31950f137c9c357b9fa89448b176d76998180c08ceaca78bba98be  Gemma3-1B-IT_multi-prefill-seq_q4_ekv4096.litertlm
+faa60663b333290c1496c499828b21d3e3254a788cacd8cce917ce0f761a2dc9  Qwen2.5-1.5B-Instruct_multi-prefill-seq_q8_ekv4096.litertlm

fetch_licenses.py

@@ -24,9 +24,12 @@ PLAINTEXT = {
     "LICENSE-apache-2.0.txt": "https://www.apache.org/licenses/LICENSE-2.0.txt",
 }
 
-HTML_PAGES = {
-    "LICENSE-gemma-terms.txt": "https://ai.google.dev/gemma/terms",
-    "PROHIBITED_USE_POLICY.txt": "https://ai.google.dev/gemma/prohibited_use_policy",
+# Both shipped models are Apache 2.0, so nothing here needs scraping today. Keep the entries
+# commented rather than deleting the machinery: re-adding any Gemma 1/1.1/2/3/3n model brings
+# these obligations straight back, and the extraction is fiddly enough to be worth preserving.
+HTML_PAGES: dict[str, str] = {
+    # "LICENSE-gemma-terms.txt": "https://ai.google.dev/gemma/terms",
+    # "PROHIBITED_USE_POLICY.txt": "https://ai.google.dev/gemma/prohibited_use_policy",
 }
 
 BLOCK_TAGS = {"p", "li", "h1", "h2", "h3", "h4", "h5", "h6", "tr", "div", "section"}