Merge branch 'upstream' into concedo_experimental

# Conflicts:
#	.github/workflows/ai-issues.yml
#	CONTRIBUTING.md
#	docs/autoparser.md
#	docs/ops.md
#	docs/ops/Metal.csv
#	ggml/src/ggml-cann/aclnn_ops.cpp
#	ggml/src/ggml-cann/ggml-cann.cpp
#	ggml/src/ggml-cpu/CMakeLists.txt
#	ggml/src/ggml-hexagon/ggml-hexagon.cpp
#	ggml/src/ggml-hexagon/htp/CMakeLists.txt
#	ggml/src/ggml-hexagon/htp/hex-dma.h
#	ggml/src/ggml-hexagon/htp/hex-utils.h
#	ggml/src/ggml-hexagon/htp/htp-ctx.h
#	ggml/src/ggml-hexagon/htp/htp-msg.h
#	ggml/src/ggml-hexagon/htp/htp_iface.idl
#	ggml/src/ggml-hexagon/htp/hvx-base.h
#	ggml/src/ggml-hexagon/htp/main.c
#	ggml/src/ggml-hip/CMakeLists.txt
#	models/templates/Apriel-1.6-15b-Thinker-fixed.jinja
#	models/templates/deepseek-ai-DeepSeek-R1-Distill-Qwen-32B.jinja
#	models/templates/deepseek-ai-DeepSeek-V3.1.jinja
#	models/templates/llama-cpp-deepseek-r1.jinja
#	models/templates/meetkai-functionary-medium-v3.1.jinja
#	scripts/fetch_server_test_models.py
#	scripts/snapdragon/adb/run-cli.sh
#	scripts/snapdragon/adb/run-completion.sh
#	scripts/snapdragon/adb/run-mtmd.sh
#	scripts/snapdragon/adb/run-tool.sh
#	tests/test-chat-auto-parser.cpp
#	tests/test-chat-peg-parser.cpp
#	tests/test-chat.cpp
#	tools/cli/cli.cpp
#	tools/server/README.md

tools/server/tests/unit/test_chat_completion.py

@@ -51,6 +51,27 @@ def test_chat_completion(model, system_prompt, user_prompt, max_tokens, re_conte
     assert choice["finish_reason"] == finish_reason
 
 
+def test_chat_completion_cached_tokens():
+    global server
+    server.n_slots = 1
+    server.start()
+    seq = [
+        ("1 2 3 4 5 6", 77, 0),
+        ("1 2 3 4 5 6", 77, 76),
+        ("1 2 3 4 5 9", 77, 51),
+        ("1 2 3 9 9 9", 77, 47),
+    ]
+    for user_prompt, n_prompt, n_cache in seq:
+        res = server.make_request("POST", "/chat/completions", data={
+            "max_tokens": 8,
+            "messages": [
+                {"role": "system", "content": "Test"},
+                {"role": "user", "content": user_prompt},
+            ],
+        })
+        assert res.body["usage"]["prompt_tokens"] == n_prompt
+        assert res.body["usage"]["prompt_tokens_details"]["cached_tokens"] == n_cache
+
 @pytest.mark.parametrize(
     "system_prompt,user_prompt,max_tokens,re_content,n_prompt,n_predicted,finish_reason",
     [
@@ -210,6 +231,7 @@ def test_completion_with_response_format(response_format: dict, n_predicted: int
 def test_completion_with_json_schema(jinja: bool, json_schema: dict, n_predicted: int, re_content: str):
     global server
     server.jinja = jinja
+    server.debug = True
     server.start()
     res = server.make_request("POST", "/chat/completions", data={
         "max_tokens": n_predicted,

tools/server/tests/unit/test_compat_anthropic.py

@@ -63,8 +63,10 @@ def test_anthropic_messages_basic():
     assert "text" in res.body["content"][0], "Text content block missing 'text' field"
     assert res.body["stop_reason"] in ["end_turn", "max_tokens"], f"Invalid stop_reason: {res.body.get('stop_reason')}"
     assert "usage" in res.body, "Missing 'usage' field"
+    assert "cache_read_input_tokens" in res.body["usage"], "Missing usage.cache_read_input_tokens"
     assert "input_tokens" in res.body["usage"], "Missing usage.input_tokens"
     assert "output_tokens" in res.body["usage"], "Missing usage.output_tokens"
+    assert isinstance(res.body["usage"]["cache_read_input_tokens"], int), "cache_read_input_tokens should be integer"
     assert isinstance(res.body["usage"]["input_tokens"], int), "input_tokens should be integer"
     assert isinstance(res.body["usage"]["output_tokens"], int), "output_tokens should be integer"
     assert res.body["usage"]["output_tokens"] > 0, "Should have generated some tokens"